Thin-Walled Structures 55 (2012) 2236

Contents lists available at SciVerse ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Finite element numerical evaluation of elliptical hollow section steel

columns in re
Tom Scullion n, Faris Ali, Ali Nadjai
FireSERT, School of Built Environment, University of Ulster, Jordanstown, UK

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

Article history: To complement previous experimental studies on the performance of unprotected Elliptical Hollow
Received 31 October 2011 Section (EHS) steel columns, under the hydrocarbon re curve, a numerical analysis has been
Received in revised form performed and outlined in this paper. A three-dimensional Finite Element Method (FEM) model has
20 February 2012
been developed and calibrated against 12 experiments, comprising of six unrestrained and six axially
Accepted 20 February 2012
Available online 28 March 2012
restrained EHS columns of two slenderness, lz 40.1 and lz 50.8. The EHS temperature proles and
axial displacements, measured under three different loading levels, (aL 0.3, 0.45 and 0.6) were utilised
Keywords: in the calibration process. The mechanical and thermal properties for carbon steel at elevated
Elliptical Hollow Section temperatures, detailed in the Eurocode standard EC3 Part 12, design of steel structures [1] have been
Columns
applied. Axial displacement charts illustrate that a close agreement between the FEM model and the
Fire resistance
experiment results was achieved, while highlighting how critical the thermal expansion coefcient and
Axial restraint
Local buckling geometric imperfection was during the calibration process. Ultimately, the paper will detail appropriate
Finite Element Method recommendations for the thermal analysis of unlled Elliptical Hollow Section steel columns, of steel
Thermal analysis grade S355J2H, and provide the platform for comprehensive parametric re investigations to
Eurocode commence.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction
The Elliptical Hollow Section (EHS) is the newest member of
the Steel Hollow Section (SHS) family, and has been the focus of
numerous studies over the last number years. This new hollow
section type provides an exciting alternative to the rectangular
and circular counterparts, for structural design engineers and
architects. The continuously changing outside curvature not only
provides elegant lines for aesthetic appeal, but numerous
moments of inertia, that can be used to enhance and pioneer
new structural designs.
Until recently, research into this new structural shape have
been limited to a few studies in the 19 60s and early 1970s, which
observed the buckling behaviour of oval cylindrical shells, man-
ufactured from polyester lm. These investigations included
stability under compression [2], and a combination of bending
moments under axial force [3]. The downside to these studies
realised that their plastic resistance could not be achieved. It was
recognised that to enhance the structural design guidance for
this new section, a number of in-depth investigations would be
n
Corresponding author.
E-mail addresses: scullion-t1@email.ulster.ac.uk (T. Scullion),
f.ali@ulster.ac.uk (F. Ali), a.nadjai@ulster.ac.uk (A. Nadjai).
required. One of the rst comprehensive studies was the cross
section classication by Gardner and Chan [4], which grouped
each EHS cross section into four classes. Grouping into these
classes, identied the resistance and rotation capacity limits of
the cross section, before failure occurred by local buckling. This
study concluded that the class limits for Circular Hollow Sections
(CHS) were deemed applicable, if an equivalent diameter for the
EHS was obtained. Further studies in this comprehensive experi-
mental programme by Chan and Gardner, explored the exural
buckling [5], bending strength [6] and compressive resistance [7]
response of member length EHS columns. The elastic buckling of
Elliptical Hollow Sections in compression was also investigated by
Ruiz-Teran and Gardner [8] and analysed by Silvestre [9].
Filling the Elliptical Hollow Section void with concrete is a
practice widely used in Circular (CHS) and Rectangular Hollow
Sections (RHS), this provides additional structural qualities to the
section; these qualities include increased stiffness, compressive
strength and additional re resistance. Yang et al. [10] investigated
the structural properties of concrete-lled elliptical hollow sections
(CFEHS) while suggesting a proposed design formula for member
design. Zhao and Packer [11] also investigated concrete lled EHS
stub columns. This investigation veried the suitability of circular
equivalent diameters, but in addition proposed effective RHS width
and depth dimensions, as local buckling observations resembled
plate buckling rather than cylindrical shell buckling.

0263-8231/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tws.2012.02.012
 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236 23

Nomenclature T thickness of EHS wall

TF furnace temperatures (FEM)
Dl vertical axial displacement TS steel temperatures (FEM)
Dt time step increment u free thermal expansion
Dya,t increase of steel temperature V section volume per unit length
A cross-sectional area of column steel v Poissons ratio
Am unit surface area exposed to re x geometric imperfection
Ca,t dpecic heat capacity at temperature a thermal expansion coefcient
Ea modulus of elasticity at ambient temperature ac convection coefcient
fy yield strength aL axial load ratio
H outside dimension of EHS on minor axis e strain
h_ net net heat ux em steel member surface emissivity
h_ net,c convective net heat ux ef furnace surface emissivity
h_ net,r radiative net heat ux Yf furnace temperatures
hc convection heat transfer coefcient Yf,t furnace gas temperature with respect to time
hcom combined heat transfer coefcient Yg gas temperatures
hr radiative heat transfer coefcient Ym steel member surface temperature
I second moment of area Ym,t steel member surface temperature with respect
L original column length to time
Le effective length Yr radiation induced temperatures
Pa load applied to column lB Euler buckling value
Pcr critical buckling load of column lz slenderness about the z-axis
qc convective heat ux per unit area ra density of carbon steel at ambient temperature
qr radiative heat ux per unit area s Stefan Boltzmann constant
qtot total (Convection and radiation) heat ux per scr elastic critical buckling stress
unit area F radiation conguration factor
r radius of gyration

Only a few viable Finite Element (FE) studies on Elliptical
Hollow Sections have been carried out to date, the most signi-
cant performed by Zhu and Wilkinson [12]. They investigated the
local buckling behaviour of the EHS, while verifying that the
equivalent CHS diameters proposed by Gardner and Chan [13]
was a viable approach to adopt. Espinos et al. [14] produced a
Finite Element model to understand the behaviour of CFEHS, with
proposed design guidance. The FEM model used in Espinoss
investigation was calibrated with historical temperature proles
from concrete-lled circular hollow sections (CFCHS). This inves-
tigation however, will use the experimental data obtained from
the rst ever studies into the re resistance of unrestrained [15]
and axially restrained [16] unlled EHS steel columns performed
by Scullion et al., to calibrate the FEM model.
2. The experimental investigation
A hydrocarbon re based experimental investigation, comprising
of 12 unlled pin-ended Elliptical Hollow Section steel columns, was
conducted at the FireSERT facilities of the University of Ulster. The
two part experimental programme tested two different section sizes,
250  125  8.0 mm and 200  100  8.0 mm with slenderness
lz 40.1 and lz 50.8, namely EHS-A and EHS-B, respectively. The
rst stage tested six columns, three of each slenderness, unrestrained
and under three different loading levels (aL 0.3, 0.45, 0.6). These rst
tests demonstrated the unique local and global buckling modes of the
EHS columns under compressive loads, full details of the results are
available in Scullion et al. [15]. The second stage of the experimental
program applied axial restraint to the EHS columns thermal expan-
sion. As columns are rarely used in isolation, the application of this
axial restraint represents a more realistic condition of column
boundary constraints, found in construction. This study illustrated
the additional axial forces that are present when restraint is applied
and demonstrated the accelerated failure rates of the EHS columns.
Full listings of the experimental results and comparisons with the
unrestrained EHS columns are documented in Scullion et al. [16].
3. Finite element analysis
Fire testing of structural elements is expensive, through an
accumulation of running costs and test specimen expenditure. As
test specimens have only a one test life time, the advantages of using
Finite Element (FE) models become even more benecial. The Finite
Element Method (FEM) is a numerical modelling technique that nds
approximate solutions for partial and integral equations. With the
advancements in computer technology both in FE software and
computing processing speeds, the ability for solving complex elasti-
city and structural analysis problems has become extremely applic-
able. However, an FEM model for thermal analysis can only be
successful and credible, if the correct input parameters and time
temperature varying values for material properties are correctly
detailed. In the absence of tensile coupon testing or experimental
data, the most reliable source to date for obtaining thermal and
mechanical properties for structural carbon steel is from the struc-
tural Eurocode (EC) publications. This investigation will use these EC
material parameters to develop and calibrate an FE model, detailing
the unlled Elliptical Hollow Section steel column under various
loading and restraint levels, subjected to re.
The Finite Element Analysis (FEA) software used for this verica-
tion process was the software package TNO-DIANA Version 9.4,
with model geometry and mesh created using the windows based
graphical user interface (GUI) Midas FXVersion 3.0.0. A three-
dimensional numerical model was created, utilising the staggered
heat ow-stress analysis modules of DIANA. The model rst demon-
strated and validated the maximum compressive capacity resistance
of the columns at room temperatures, before a structural thermal
analysis commenced. The requirement for a heat ow-stress working
domain environment determined the Finite Element (FE) selection.
24 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236
During thermal analysis the eight-node isoparametric brick element
HX8HT was merged with the four nodes of the three-dimensional
(3D) linear boundary ow element BQ4HT. This boundary element
was used to describe the thermal convection and radiation to the
brick element surface. The temperature values resulting from the
heat ow analysis was automatically tabulated into a time
temperature form, this was then used for the structural analysis.
During the structural analysis the boundary element BQ4HT was
deactivated, and the potential ow HX8HT element was converted to
the three-dimensional 20 node isoparametric solid brick structural
element CHX60 [17].
4. The Eurocode parameters
The Elliptical Hollow Sections are manufactured to technical
delivery conditions of the European standard EN10210-1:2006
[18], the tolerances and dimensions of the section are detailed in
EN10210-2:2006 [19]. The Eurocode 3 BS EN 1993 contains 20
parts detailing the structural design recommendations for steel
structures. EC3 Part 11 [20] details the general design rules for
steel structures with EC3 Part 12 [21] outlining the structural
re design rules for steel structures.
4.1. Steel mechanical properties at cold temperatures
The parameters required to model steel under cold tempera-
tures are the modulus of elasticity (Ea ), yield strength (f y ),
Poissons ratio (v) and density (ra ). The EC publications recom-
mend values for these parameters for structural carbon steel. A
value of 210 kN/mm2 was used for the modulus of elasticity as
recommended by EC3 Part 11 Section 3.26 [20]. Table B3 of the
technical delivery conditions BS EN 10210-1:2006 [18] species
for Steel Hollow Sections (SHS) with a thickness (T) less than
16 mm have a minimum yield strength of 355 N/mm2, dened
also in the EHS manufacturers datasheet [22]. The density of hot
rolled steel has been recommended as 7850 kg/m3 by EC3 Part
12 Section 3.2.2 [21], with a Poisson Ratio of 0.3 assigned for hot
rolled steel by EC3 Part 11 [20].
4.2. Thermal and mechanical properties of steel at elevated
temperatures
The thermal parameters required for structural carbon steel, at
elevated temperatures include, specic heat capacity, conductiv-
ity and thermal expansion. These parameters are appropriately
detailed in Section 3.4 of EC3 Part 12 [21]. The reduction values,
required for the modulus of elasticity and yield strength, are
presented in Table 3.1 Section 3.4 of EC3 Part 12 [21].
4.2.1. Density and Poissons ratio of structural carbon steel
at temperature
The density of steel is considered independent of temperature
by EC3 Part 12 Section 3.2.2 [21], therefore the constant ambient
value was used. Like density, any variation in the Poissons ratio
has been discounted under elevated temperatures, as the percen-
tage change in Poissons ratio is negligible compared to other
inuencing parameters.
5. FEM model development
Two FEM models have been created detailing the geometry of
both EHS section types, EHS-A and EHS-B, used in the experi-
mental investigation, with the EC material properties for struc-
tural carbon steel assigned.
5.1. FEM model development at room temperatures
The FEM model was initially evaluated under room tempera-
ture conditions with theoretically hand calculations. This vali-
dated boundary end constraints, critical buckling mode shape,
and maximum compressive strength capacity, calculated using
the EC method.
5.1.1. Geometry and mesh application
Dia and Lam [23] investigated element sizes suggesting a
range of between 5 and 10 mm for steel. This range demonstrated
a good compromise, to maximise model accuracy with experi-
mental results and minimise computation runtime for cold
analysis. However, thermal analysis is more computation exten-
sive therefore this range was increased at column ends with the
smaller elements used in areas of interest. A number of mesh
variations were used throughout this investigation (Fig. 1), for
cold analysis mesh type 1 was selected. This mesh type used a
uniformly distributed seed along the length of the column and
three element layers for the section thickness. The mesh seed was
graded so mesh density was increased in the tighter corners of the
EHS column.
5.1.2. Maximum compressive capacity
EC3 Part 11 Section 6.3 [20] was applied to calculate the
buckling resistance of the EHS columns. This is the current
method for obtaining maximum compressive capacity for all hot
nished hollow section columns of class 1, 2 or 3, and was veried
by Chan and Gardner [5] to be suitable for EHS member columns.
To calculate the maximum compressive strength, a buckling curve
and instability safety margin must be selected. Buckling curve a
is the appropriate selection for hot rolled steel of grade S355J2H.
For the purpose of this investigation an instability safety margin
gM1 1.1 was used. A value of 1.1 is considered a suitable value for
steel structures and buildings with loads of wide scatter. Table 1
details the calculated maximum capacity loads for EHS-A and
EHS-B using these parameters.
5.1.3. Euler buckling analysis
Using the Euler buckling analysis module of DIANA, a buckling
value lB that represents the ratio Pcr =Pa was obtained. Where P cr
is the critical buckling load of a perfect column (Eq. (1)) and Pa the
applied load.
p2 Ea I
Pcr 1
Le 2
Using this buckling value lB the FEM model was veried with
the theoretical critical load using Eq. (2). Fig. 2 illustrates the
overall global buckling mode shape observed for EHS-B, this is the
lowest buckling mode and the most critical collapse mode. The
FEM model buckling values for both sections are compared with
theory in Table 1. The error tolerance indicates a margin of
accuracy; determining that a satisfactory correlation between
FEM model results and theoretical values has been obtained.
Pcr P a  lB 2
5.1.4. Elastic critical buckling stress of member length EHS columns
The critical buckling stress (scr ) of Eq. (3) is dependent on the
modulus of elasticity Ea of the material and the slenderness ratio
Le =r) of the column [24].
P cr p2 Ea
scr 3
A Le =r2
The buckling value obtained from the FEM Euler buckling
analysis and Eq. (4), was used to determine the critical buckling
 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236 25

Mesh Type 1 - Top

Mesh Type 2 - Top

Mesh Type 1 (8640) Mesh Type 2 (1440) Mesh Type 3 (4972) Mesh Type 3 - Top

Fig. 1. FEM mesh density variations.

Table 1
Critical buckling loads at cold temperature.

EC Max. kB Pcr,FEM Pcr,Th Error Tolerance

Cap.
(kN) (kN) (kN) Pcr,Th/Pcr,FEM

EHS-A 1333 4.042 5388 5815 1.079

EHS-B 1000 2.745 2745 2853 1.039

stress. The FEM results are compared with the theoretical values
in Table 2.
Dl
scr Ea :e Ea :lB : 4
l

5.1.5. Geometric imperfection

The buckling mode shape for a pin-ended column is sin px=L.
The geometrical imperfection x is distributed over the column
length with the maximum displacement at the mid span. The
load-displacement chart of Fig. 3 illustrates that applying an
imperfection of 5 mm, an amplitude of L=360, is permitted to a
perfect pin ended 1800 mm length EHS column, before the EC
Fig. 2. Pin ended column with Euler buckling mode 1.
maximum capacity exceeds the section elastic range.

5.1.6. Accidental eccentricity

The experiments used a dual hydraulic ram system, to produce Table 2
the compressive loads, this loading method from the two rams Elastic critical buckling stress at cold temperature.
can create a potential for horizontal loading offset. To calibrate
the FEM model with the experimental data, these immeasurable Displacement (Dl) rcr,FEM rcr,Th Error Tolerance
(mm) (N/mm2) (N/mm2) rcr,Th/rcr,FEM
qualities of eccentricity were accounted for by increasing the
geometric imperfection of the EHS column. EHS-A 2.534 1195 1290 1.079
EHS-B 2.397 767 799 1.042
5.2. FEM model development at high temperatures

5.2.1. Temperature development in unprotected steel (EC method) heat transfer by convection and radiation, described in EC1-1-2
To determine the temperature rise of an unprotected steel Section 3.1, shown in Eq. (5).
member, the net heat ux (h_ net ) to which the member surface is
exposed requires calculation. The net heat ux is a combination of h_ net h_ net,c h_ net,r 5
26 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236

1600

1400

1200
Axial Compressive Load (kN)

1000 EHS-A Imperfection 5 mm

EHS-A Imperfection 10 mm
800 EHS-A Imperfection 15 mm

EHS-A Imperfection 20 mm
600
EHS-B Imperfection 5 mm

400 EHS-B Imperfection 10 mm

EHS-B Imperfection 15 mm
200
EHS-B Imperfection 20 mm

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Total Axial Displacement (mm)

Fig. 3. Elastic range for pin ended EHS column with geometric imperfection.

The convective net heat ux is dened by the linear relation-
ship shown in Eq. (6), and the radiative net heat ux by Eq. (7). A
conguration factor (F) of 1.0 is used for members completely
engulfed in re, this is true for furnace experiments. The surface
emissivity (em ) of the member is normally assumed as 0.8, if not
otherwise stated. The gas temperatures (Yg ) and radiation tem-
peratures (Yr ) will be replaced with one parameter called furnace
temperatures (Yf ). The emissivity of re (ef ) is 1.0 and the Stefan
Boltzmann constant (s) is 5:67E8 W=m2 K 4 . With temperatures in
Kelvin, the net heat ux now becomes Eq. (8).
h_ net,c ac Yg Ym
h_ net,r F:em :ef :s:Yr 2734 Ym 2734 
h_ net ac Yf ,t Ym,t em :s:Y4f ,t Y4m,t
The temperaturetime curve (TTC) equations for gas tempera-
tures in a compartment with corresponding coefcients of heat
transfer by convection are outline in EC1-1-2 Section 3.2. The
hydrocarbon TTC for furnace gas temperatures is shown in Eq. (9),
with a corresponding coefcient of heat transfer by convection
(ac ) of 50 W/m2K.
Yf ,t 108010:325e0:167t 0:675e2:5t 20
A method for determining the increase in the member tem-
perature for unprotected steel members is dened in equa-
tion 4.25 of EC3 Part 12, [21] this temperature development
with respect to time is dened as Eq. (10).
Am =V _
Dya,t ksh h net Dt
ca ra
Replacing h_ net with the convection and radiation summation of
Eq. (8) produces Eq. (12).
fac Yf ,t Ym,t em :s:Y4f ,t Y4m,t gDt Dya,t :Am 1 :V:ca,t :ra 12
Resolving Eq. (12) determines the temperature development of
the member surface, (Ym,t ) through heat transfer by convection
and radiation induced by furnace temperatures, surrounding the
member. An iterative based process is required to resolve this
equation, a time step of 5 s was used, as recommended by EC3
[21]. The furnace temperatures (Yf ,t ) were calculated using the
appropriate TTC with coefcient of heat transfer by convection
6
(ac ) detailed in EC1 Part 12, Section 3.2. [25] Fig. 4 illustrates
the calculated temperature development for section EHS-B, under
7 the hydrocarbon and standard temperaturetime curves using the
aforementioned method.
8
5.2.2. Modelling the boundary heat transfer conditions
To represent the convection and radiation heat transfer con-
ditions of the furnace, the DIANA boundary ow element BQ4HT
was used to model the heat ux at the boundary of the EHS
elements. This heat ux per unit area (qc ) at the boundary is
represented by the linear Eq. (13).
qc hc T F T S 13
9 Parameters T S and T F are the temperatures of the steel and
furnace, respectively, and hc is the convection coefcient in
W/m2K. Eq. (13) models the convective heat transfer of Eq. (6).
To model the radiative heat transfer using the BQ4HT linear
boundary element, the radiative quadratic equation dened in
Eq. (7), must be represented in the linear form as shown in
Eq. (14) [26].
10
qr hr T F T S 14

Similar to other hollow section types, for an EHS fully The radiative heat transfer coefcient (hr ) is therefore dened
embedded in re, the shadow effect does not inuence tempera- as Eq. (15), to enable this linear form.
ture development, therefore ksh is unity. Rearranging Eq. (10)
hr em :s:T F 2 T S 2 T F T S 15
(conservation of energy) denes Eq. (11).
A combined heat transfer coefcient is a summation of
h_ net Dt Dya,t :A1
m :V:c a :ra 11 the radiative heat transfer and the convective heat transfer
 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236 27

1200

1000

800
Temperature (C)

600

Hydrocarbon TTC

400 Standard TTC

Hydrocarbon EC steel
Standard EC steel
200 Hydrocarbon FEM EHS-B
Standard FEM EHS-B

0
0.00 5.00 10.00 15.00 20.00 25.00 30.00
Time (Minutes)

Fig. 4. Eurocode steel temperature development for EHS-B.

Table 3
Combined heat transfer coefcients (hcom) for each re curve, W/m2K.

Temperature (1C) 20 50 100 200 300 400 500 600 700 800 900 1000

Hydrocarbon TTC 55 92 117 147 170 195 223 256 293 341 387 442
Standard TTC 30 48 60 79 97 119 143 173 209 260 321 400
Exp Furnace EHS-A 30 30 31 34 36 38 40 42 45 47 49 51
Exp Furnace EHS-B 32 33 35 38 41 45 48 52 55 58 62 65

1100

1000

900

800

700
Temperature (C)

600

500

400 FEM EHS-A MID Node

FEM EHS-B MID Node

300
FEM Hotspot Node

200 EXP EHS-A3 MID AVG TC

EXP EHS-B3 MID AVG TC

100
EXP EHS Hotspot AVG TC
0
0 2 4 6 8 10 12 14 16 18 20
Time (Minutes)

Fig. 5. Experiment and FEM model steel temperature development.

28 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236

coefcients, Eq. (16). Table 3 details the combined heat transfer coefcients in
W/m2K used to represent a hydrocarbon and ISO standard re
hcom hr hc 16 for the FEM model.

To model the non-linear boundary conditions the total heat

ux per unit area (qtot ) was represented by Eq. (17), with hcom a 5.2.3. Validation of FEM temperature prole with the EC method
function of boundary temperature. As FEM thermal analysis is computation intensive, mesh type
2 was selected (Fig. 1). Decreasing the element numbers helps to
qtot hcom T F T S 17 facilitate in reducing analysis run time. A comparison with cold
analysis results of mesh type 1, conrmed that model accuracy
was not signicantly compromised by reducing the element
number. For thermal analysis, the linear boundary ow element
BQ4HT encapsulated the external element faces of the
section. The EHS column was uniformly heated using an external
potential heat load representing the hydrocarbon furnace gas
temperatures, assigned to these boundary elements. Selecting the
hydrocarbon combined heat transfer coefcients of Table 3 the
FEM model temperature prole was compared with the tempera-
ture development predicted by the EC method, detailed in
Section 5.2.1. Fig. 4 illustrates that the temperature development
of the centroid node of the EHS-B FEM column correlates closely
with EC predicted temperature development for the same col-
umn. Exchanging the hydrocarbon TTC with the standard TTC
and appropriate combined heat transfer coefcients, the tem-
perature development of EHS column under the standard heating
regime was also validated. Fig. 4 demonstrates for EHS-B section,
that an excellent correlation between the EC predicted tempera-
tures and the FEM model for the standard TTC heating regime was
obtained.

5.2.4. FEM model calibration with experiment temperature prole

During the experiments, twelve thermocouples measured the
steel surface average temperature. Although due care was taken
to ensure uniform heating distribution was applied to the EHS
column during the test, accidental thermal eccentricities
occurred. These temperature anomalies in steel temperature
resulted from burner alignment and ventilation draught. An
additional thermocouple was used to monitor the heating eccen-
tricity which resulted from gas burner alignment; this created a
hotspot on the column at an approximate height of 600 mm.
To calibrate the FEM model with the experimental data, mesh
type 3 was used. This mesh type was divided to allow a denser
mesh in the area of the hotspot, where plastic deformation was
observed. The convection boundary mesh at this area was
Fig. 6. Typical FEM temperature prole with calibrated hotspot.
allocated a separate calibrated convection coefcient parameter

30

25
Axial Displacement (mm)

20

15

10

Constant 1.2E-5 C-1

5 Eurocode

0
0 200 400 600 800 1000 1200
Temperature (C)

Fig. 7. EHS-B Free thermal expansion.

 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236 29

to match temperatures observed at the hotspot during the
experiments. The remaining section surface area was customized
to the average steel surface experiment temperatures, using a
variable convection coefcient parameter, specic to the furnace
and section type. The same variable convection coefcient para-
meters could not be used for both section types, as the furnace
size was not scaled for each section. The proximity of the section
sides to the furnace walls was closer for the larger sections,
creating small differences in convection and radiation distribution
within the furnace. However, for the hotspot area, a constant
convection coefcient of 110 W/m2K was deemed to satisfy the
experimental results, for both sections. Table 3 outlines these
calibrated combined convection parameters. The experiment
average steel temperature proles were obtained by using these
convection parameters, paralleled with the hydrocarbon TTC of
Eq. (9). Fig. 5 illustrates the calibrated steel surface temperatures
at the hotspot and average steel surface temperatures from the
experiment results and FEM model. Fig. 6 displays the FEM
transient heat prole observed for section EHS-B, the area of the
hotspot is evident in this illustration.
5.3. FEM staggered heat-ow stress analysis (unrestrained)
Applying the corresponding compressive loads to the cali-
brated thermal model and buckling imperfection factors, the
stability of the column was analysed under both temperature
and compressive loads, and compared with the experimental
ndings.
5.3.1. Validating FEM model with free thermal expansion theory
When the EHS column was exposed to elevated temperatures,
the column increased in length; this thermal elongation for
carbon steel is described in Section 3.4.1 of EC3 Part 12. An
approximation of the thermal expansion can be calculated using
Eq. (18). Before any compressive loads were applied to the FEM
model, the thermal elongation of the model was compared with
theoretical hand calculations to verify functionality. To simplify
the hand calculations a constant thermal expansion coefcient (a)
of 1.2E  05 1C  1 was used to validate the FEM model, with the
theoretical equation (Fig. 7). Table 4 shows these comparisons
illustrating an error tolerance of between 3% and 6%.
u a:L:DT s 18

Table 4 5.3.2. Comparison of the unrestrained experimental results with

Thermal Elongation with constant coefcient 1.2E-05 1C  1. FEM model
The rst set of experiments tested six EHS columns of two
Ts D Ts d (FEM) d (Theory) Error different section types, unrestrained and with three different
20 0 0.00 0.00 1.00
loading levels, 60%, 45% and 30% of maximum capacity. Initially,
100 80 1.84 1.73 0.94 the EC parameters with their respective adjustments at tempera-
200 180 4.04 3.89 0.96 ture were used for the FEM model description, and compared
300 280 6.33 6.05 0.96 with the experimental ndings. Fig. 8 details the FEM axial
400 380 8.56 8.21 0.96
displacements in the time domain for the unrestrained EHS-A
500 480 10.77 10.37 0.96
600 580 12.95 12.53 0.97 column with 45% loading. At rst a 5 mm geometric imperfection
700 680 15.14 14.69 0.97 was applied to the FEM model. The imperfection was increased in
800 780 17.46 16.85 0.96 5 mm steps to enable calibration with the experiment data. From
900 880 19.61 19.01 0.97 Fig. 8 it is apparent that increasing the geometric imperfection by
1000 980 21.75 21.17 0.97
5 mm has a signicant effect in aligning failure times, but not the

14
13
12
11
10
9
8
7
Displacement (mm)

6
5
4
3
2
1
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
-1
-2
-3 EHS-A2-U Experiment FEM_A_U_45_5_EC FEM_A_U_45_10_EC
-4 FEM_A_U_45_15_EC FEM_A_U_45_20_EC FEM_A_U_45_25_EC
-5 FEM_A_U_45_30_EC FEM_A_U_45_30_EC -1.2E-05
-6
Time (Minutes)

Fig. 8. Calibration of EHS-A2-U with experimental data.

30 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236

maximum axial displacement of the experimental data. The axial
displacement is inuenced more by the thermal elongation
coefcient. From the experimental results, any measurement of
axial displacement is delayed until the steel temperature is
approximately 100 1C. This delay in measurement relates to
slackness and bedding within in the test rig, during the start of
the heating process. Therefore, delaying the onset of thermal
expansion, and replacing the variable thermal expansion
coefcient with a constant value of 1.2E-05 1C  1, has helped to
align the FEM model with the maximum axial displacement.
For the EHS-A section type, an imperfection of approximately
30 mm and a constant thermal expansion coefcient of 1.2E-
05 1C  1 produced a close approximation with experimental
results, an exception for EHS-A1 applied as an imperfection
increase to 32.5 mm was required, Fig. 9. For the more slender
section type EHS-B, a geometric imperfection of 25 mm for all

10

5
Displacement (mm)

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
-1

-2

-3

-4 EHS-A1-U Experiment EHS-A2-U Experiment EHS-A3-U Experiment

-5 FEM_A_U_60_30 FEM_A_U_45_30 FEM_A_U_30_32_5

-6
Time (Minutes)

Fig. 9. Unrestrained EHS-A axial displacement development (time domain).

10

7
6

5
Displacement (mm)

2
1

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
-1

-2
-3

-4 EHS-B1-U Experiment EHS-B2-U Experiment EHS-B3-U Experiment

FEM_B_U_60_25 FEM_B_U_45_25 FEM_B_U_30_25
-5
-6
Time (Minutes)

Fig. 10. Unrestrained EHS-B axial displacement development (time domain).

 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236 31

Table 5
Unrestrained experiment and FEM model comparisons.

Load ratio Imp. (mm) Time (min) Temperature (1C) Max. displacement (mm)

(aL) x Exp. FEM Error Exp. FEM Error Exp. FEM Error

EHS-A-U 0.60 30.0 10.20 10.16 0.996 389 455 1.170 5.12 6.18 1.207
0.45 30.0 12.58 12.67 1.007 504 532 1.056 6.75 6.69 0.991
0.30 32.5 14.83 15.50 1.045 586 618 1.055 8.52 8.39 0.984
EHS-B-U 0.60 25.0 10.25 9.42 0.919 460 477 1.037 5.69 6.05 1.063
0.45 25.0 11.50 10.92 0.950 519 532 1.025 7.08 7.09 1.001
0.30 25.0 14.08 13.75 0.977 597 627 1.050 8.37 8.34 0.996

10

5
Displacement (mm)

0
0 100 200 300 400 500 600 700
-1

-2

-3 EHS-A1-U Experiment EHS-A2-U Experiment EHS-A3-U Experiment

-4 FEM_A_U_60_30 FEM_A_U_45_30 FEM_A_U_30_32_5

-5
Steel Average Temperature (C)

Fig. 11. Unrestrained EHS-A axial displacement development (temperature domain).

three loading levels illustrated close agreement with experiment
axial displacements, Fig. 10. Table 5 details the comparisons for
each experiment with corresponding model results and error
tolerances dened.
heating-up program does not inuence the ultimate failure
temperature of the section.
5.5. FEM staggered heat-ow stress analysis (restrained)

During the second stage of the experimental investigation, the

5.4. The heating-up program
thermal expansion of the EHS column was restrained. The test rig
was designed to impose an axial restraint on the thermal expan-
Fig. 11 and Fig. 12 illustrate the axial displacements in the
sion of the EHS columns. This applied stiffness was measured as
temperature domain, for EHS-A-U and EHS-B-U, respectively. As
61.2 kN/mm. Spring elements were used to model the test rig
the EHS columns are not uniformly heated the axial displacement
stiffness, restraining the translational movements of the end
is plotted with respect to the average steel temperature increase,
nodes. A loading and unloading conguration was used to dene
not truly reecting the maximum temperatures observed at the
the springs compression and tension stiffness. When the springs
hotspot. Fig. 13 shows that removing the hotspot and uniformly
were in compression the stiffness of the rig was applied, this
heating the EHS-B column, illustrates that column failure occurs
occurred during the thermal expansion of the section. As the EHS
at a higher steel temperature. This indicates that it is important to
column contracted, the spring stiffness was removed, therefore no
consider localised heating rather than uniform heating when
spring tension was observed.
determining the steel temperature on column failure.
During the experiments, thermal creep is a phenomenon that
may also inuence the failure temperature of the EHS columns. 5.5.1. Comparison of the restrained experimental results with FEM
Twilt [27] details that thermal creep is related to the heating-up model
program directly affecting the collapse and deformation beha- Using the same geometric imperfections, 30 mm for EHS-A and
viour of a steel element. The FEM model in its present form is not 25 mm for EHS-B, and constant thermal coefcient for the unrest-
capable of replicating thermal creep, so any variation of the rained models, the FEM results were compared with the
32 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236

10

5
Displacement (mm)

0
0 100 200 300 400 500 600 700
-1

-2

-3
EHS-B1-U Experiment EHS-B2-U Experiment EHS-B3-U Experiment
-4 FEM_B_U_60_25 FEM_B_U_45_25 FEM_B_U_30_25

-5
Steel Average Temperature (C)

Fig. 12. Unrestrained EHS-B axial displacement development (temperature domain).

10

5
Displacement (mm)

0
0 100 200 300 400 500 600 700
-1

-2

-3
EHS-B1-U Experiment FEM_B_B_U_60_25_HS FEM_B_U_60_25_UH
-4

-5
Steel Average Temperature (C)

Fig. 13. Uniform heating versus hotspot (temperature domain).

restrained experiment data. Fig. 14 and Fig. 15 show the axial
displacements for the restrained EHS columns of EHS-A-R and
EHS-B-R, respectively. The FEM model provided an extremely
close approximate solution for the more slender column type
EHS-B, for the three ranges of loading. However, for the stockier
column type EHS-A, the accuracy of the approximate solution
deviated as the loading level increased from 30% to 60%. Table 6
outlines the experimental results with the FEM model counter-
parts with error tolerance. Fig. 16 and Fig. 17 compare induced
restraint forces between the experiment and FEM model results.
 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236 33

2
Displacement (mm)

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
-1

-2

-3

EHS-A1-R Experiment EHS-A2-R Experiment EHS-A3-R Experiment

-4
FEM_A_R_60_30 FEM_A_R_45_30 FEM_A_R_30_30

-5
Time (Minutes)

Fig. 14. Restrained EHS-A axial displacement development (time domain).

3
Displacement (mm)

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
-1

-2

EHS-B1-R EHS-B2-R EHS-B3-R

-3
FEM_B_R_60_25 FEM_B_R_45_25 FEM_B_R_30_25

-4
Time (Minutes)

Fig. 15. Restrained EHS-B displacement development (time domain).

The restraint force is a multiple of axial displacement by the 6. Conclusions

imposed spring stiffness. It is evident here that thermal expansion
gradient differs from the experiment and the FEM model. Repla-
cing the constant thermal coefcient with the EC thermal expan-
sion parameters, and introducing a slack of 1.9 mm into the FEM
model, to represent experiment conditions, results in an extre-
mely close approximation. Fig. 18 illustrates this for EHS-A-R with
30% loading.
An experimental programme at the FireSERT facilities of the
University of Ulster tested 12 unprotected Elliptical Hollow
Section steel columns under hydrocarbon re, six unrestrained
and six with axial restraint, under three different loading levels,
30%, 45% and 60%. The investigation allowed for the successful
collection of experimental data which included, steel temperature
34 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236

Table 6
Restrained experiment and FEM model comparisons.

Load ratio Imp. (mm) Time (min) Temperature (1C) Max. displacement (mm)

(aL) x Exp. FEM Error Exp. FEM Error Exp. FEM Error

EHS-A-R 0.60 30.0 10.13 9.60 0.948 417 434 1.041 3.31 4.44 1.341
0.45 30.0 11.40 10.50 0.921 470 470 1.000 4.35 4.90 1.126
0.30 30.0 12.27 12.50 1.019 501 534 1.066 5.51 5.42 0.984
EHS-B-R 0.60 25.0 10.03 9.08 0.905 430 459 1.067 3.37 3.80 1.128
0.45 25.0 10.88 9.58 0.881 475 480 1.011 4.32 4.34 1.005
0.30 25.0 13.02 12.58 0.966 556 590 1.061 5.56 4.92 0.885

400

EHS-A1-R Experiment

350 EHS-A2-R Experiment

EHS-A3-R Experiment
FEM_A_R_60_30
300
FEM_A_R_45_30
FEM_A_R_30_30
Restraint Force (kN)

250

200

150

100

50

0
0 100 200 300 400 500 600 700
Steel Average Temperature (C)

Fig. 16. EHS-A restraint force development (temperature domain).

proles, axial displacement development, induced restraint force
measurements and ultimate failure times. Using the mechanical
and thermal properties dened for structural carbon steel in the
EC publications, a number of Finite Element models were created
and compared with the collated experimental data.
Using the EC thermal parameters, the FEM model demon-
strated an excellent approximation with the recorded experimen-
tal temperatures for unprotected steel. The EC mechanical
properties for carbon steel at elevated temperatures were then
utilised, for the potential ow structural analysis. Calibrating the
unrestrained EHS FEM model axial displacements, with the
experimental data, highlighted how critical the geometric imper-
fection is in the ultimate failure time of the heated EHS column.
Experiment loading and alignment eccentricities were accounted
for in this geometric imperfection. Imperfection values of 30 mm
for EHS-A and 25 mm for EHS-B produced a close approximation
with the unrestrained ultimate failure times.
The thermal expansion parameter had a signicant impact on
the axial displacement of the EHS columns, substituting the
variable thermal expansion coefcient suggested by EC 3, with a
constant value of 1.2E-05 1C  1 help to aligned the FEM maximum
model displacements with the experiment values. The
introduction of a slack coefcient to the restrained FEM models,
allowed for the recommended EC variable thermal expansion
coefcient to be reintroduced. A combination of non-uniform
heating and inherent test rig slackness, contributed to these
anomalies in the thermal expansion of the EHS columns, between
the FEM model and experimental data.
For the axially restrained FEM models the same geometric
imperfections were used. A very close agreement with the
experiment axial displacement was displayed for EHS-B, the more
slender column type. For the stockier EHS-A column, the FEM
model accuracy deviated as the loading level increased. A number
of reasons may contribute to this divergence between experi-
mental and FEM model results. First, a perfect pin ended
condition is assumed for the FEM model, any increase in the
friction coefcient on the graphite end bearings was not consid-
ered. This friction would increase with thermal expansion
directly inuencing the columns effective length during the
experiment. This increase in friction at the end bearings also
produces additional moments that are not considered in the
FEM model.
Overall, the FEM model calibration process with the experi-
ment data has demonstrated that a very good numerical
 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236 35

400

EHS-B1-R
350 EHS-B2-R
EHS-B3-R
FEM_B_R_60_25
300
FEM_B_R_45_25
FEM_B_R_30_25
Restraint Force (kN)

250

200

150

100

50

0
0 100 200 300 400 500 600 700
Steel Average Temperature (C)

Fig. 17. EHS-B restraint force development (temperature domain).

400

EHS-A3-R Experiment

350 FEM_A_R_30_30
FEM_A_R_30_30_EC -1.9 mm Slack

300
Restraint Force (kN)

250

200

150

100

50

0
0 100 200 300 400 500 600 700
Steel Average Temperature (C)

Fig. 18. EHS-A3-R restraint force development, EC parameters with 1.9 mm slack.

approximate solution can be obtained, on careful consideration to
the thermal expansion coefcient, and geometric imperfection of
the column. However, the investigation has highlighted that
further FEM numerical analysis is needed to successfully model
thermal creep, to scrutinise the impact of the heating-up program
on the EHS columns behaviour. Ultimately, the investigation
recommends that the thermal and mechanical properties, of
structural carbon steel, outlined in the EC publications, are
deemed suitable, for the thermal analysis of unprotected unlled
Elliptical Hollow Section steel columns, in re.
36 T. Scullion et al. / Thin-Walled Structures 55 (2012) 2236
Acknowledgments
The authors would like to take this opportunity to thank the
DELNI Department for Employment and Learning, Northern
Ireland for providing funding for this research, and extend their
thanks also, for the guidance provided by TNO-Diana technical
support team.
References
[1] Eurocode 3 BS EN 1993-1-4:2006 design of steel structures. General
Rules. Supplementary Rules for Stainless Steels.
[2] Feinstein G, Erickson B, Kempner J. Stability of oval cylindrical shells.
Experimental Mechanics 1971;11(11):51420.
[3] Chen YN, Kempner J. Buckling of oval cylindrical shells under compression
and asymmetric bending. AIAA Journal 1976;14(9):123540.
[4] Gardner L, Chan TM. Cross-section classication of elliptical hollow sections.
Steel and Composite Structures 2007;7(3):185200.
[5] Chan TM, Gardner L. Flexural buckling of elliptical hollow section columns.
Journal of Structural Engineering 2009;135(5):54657.
[6] Chan TM, Gardner L. Bending strength of hot-rolled elliptical hollow sections.
Journal of Constructional Steel Research 2008;64(9):97186.
[7] Chan TM, Gardner L. Compressive resistance of hot-rolled elliptical hollow
sections. Engineering Structures 2008;30(2):52232.
[8] Ruiz-Teran AM, Gardner L. Elastic buckling of elliptical tubes. Thin Walled
Structures 2008;46(11):130418.
[9] Silvestre N. Buckling behaviour of elliptical cylindrical shells and tubes under
compression. International Journal of Solids and Structures 2008;45(16):
442747.
[10] Yang H, Lam D, Gardner L. Testing and analysis of concrete-lled elliptical
hollow sections. Engineering Structures 2008;30(12):377181.
[11] Zhao XL, Packer JA. Tests and design of concrete-lled elliptical hollow
section stub columns. Thin Walled Structures 2009;47(67):61728.
[12] Zhu Y, Wilkinson T. Finite element analysis of structural steel elliptical
hollow sections in pure compression. Welding in the World 2006;50(no.
SPEC. ISS.):16370.
[13] Gardner L, Chan TM. Structural design rules for elliptical hollow sections.
presented at the 12th International Symposium on Tubular Structures
2009:3239.
[14] Espinos A, Gardner L, Romero ML, Hospitaler A. Fire behaviour of concrete
lled elliptical steel columns. Thin Walled Structures 2011;49(2):23955.
[15] Scullion T, Ali F, Nadjai A. Experimental study on performance of elliptical
section steel columns, under hydrocarbon re. Journal of Constructional Steel
Research 2011;67(6):98691.
[16] Scullion T, Ali F, Nadjai A. Effect of axial restraint on the performance of
elliptical hollow section steel columns, in hydrocarbon re. Engineering
Structures 2011;33(12):315561.
[17] DIANA Finite Element Analysis Users Manual Element Library Release 9.4.2,
1st edn. 2010.
[18] BS EN 10210-1:2006 Hot nished structural hollow sections of non-alloy and
ne grain steels. Technical Delivery Requirements BSI British Standards.
[19] BS EN 10210-2:2006 Hot nished structural hollow sections of non-alloy
and ne grain steels. Tolerances, Dimensions and Sectional Properties BSI
British Standards.
[20] Eurocode 3 BS EN-1993-1-1:2005 Design of steel structures: General
rules and rules for buildings.
[21] Eurocode 3 BS EN 1993-1-2:2005 design of steel structures. General
Rules. Structural Fire Design.
[22] CT29 Celsius 355s Ovals Sizes and Capacities Eurocode Version. Corus
Tubes, 2004.
[23] Dai X, Lam D. Numerical modelling of the axial compressive behaviour of
short concrete-lled elliptical steel columns. Journal of Constructional Steel
Research 2010;66(7):93142.
[24] Timoshenko S. Theory of Elastic Stability. 2nd edn. New York: McGraw-Hill;
1988.
[25] Eurocode 1 BS EN 1991-1-2:2002 Actions on structures. General actions.
Actions on Structures Exposed to Fire BSI British Standards. BSI, 2002.
[26] H.G. Ransing, How To Undertake Finite Element Based Thermal Analysis.
NAFEMS, 1999.
[27] Twilt L. Strength and deformation properties of steel at elevated tempera-
tures some practical implications. Fire Safety Journal 1986;13(1):915.