Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Engineering Structures 283 (2023) 115884

Contents lists available at ScienceDirect

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

The overall imperfection method for fire design situation


Samer Nemer a, *, József A. Szalai b, Ferenc Papp a
a
Széchenyi István University, Hungary
b
ConSteel Solutions Ltd, Hungary

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

Keywords: In this paper the Overall Imperfection Method, which is a well validated method for the assessment of the global
Overall imperfection method stability resistance of thin-walled steel structural members with any load and supporting conditions, is used for
Fire design fire design situation. The method uses equivalent initial geometrical imperfection in the shape of relevant elastic
GMNIA
global buckling mode. The assessment of the global stability resistance is performed by geometrically nonlinear
Beam-columns
analysis and by checking of cross-sectional resistance using reduced elastic moduli in the analysis and reduced
design strength for the cross-section checking. The validation of the method is carried out through the analysis of
more thousands different structural members with hot-rolled I cross-sections. The reference values for the safety
study are calculated by geometrically and materially nonlinear analysis with imperfections.

The most detailed description of the generalized OIM was published


1. Introduction by Szalai and Papp [7]. The OIM was used by Papp et al. for design of
irregular structural members and simple portal frames [8]. All previous
1.1. Overall imperfection method studies have been limited to the use of the OIM for the stability design of
steel members at normal temperatures. The current European standard
At normal temperature, the current European standard EN1993-1-1 for fire design of steel members EN1993-1-2 [9] gives a simplified
method for the design of steel members at elevated temperatures. The
[1] gives formulae for calculating the buckling resistance of uniform
steel members under compression, bending, and combined axial method adopts only one buckling curve and limited to uniform members
design. Therefore, there is a need to find more accurate broadly appli­
compression and bending in its paragraphs (6.3.1), (6.3.2) and (6.3.3),
respectively. There are two alternative methods available to check the cable methods.
stability resistance of steel members subjected to either uniform or non-
uniform load conditions. These methods are: the General Method (GM) 1.2. Aims and basic assumptions
which can be found in EN 1993-1-1: paragraph 6.3.4 and the Overall
Imperfection Method (OIM) in EN 1993-1-1: paragraph 5.3.2(11). In this paper the OIM will be adapted to design structural members
The Overall Imperfection Method (OIM) is the generalization of the subjected to elevated temperatures. The method utilizes the result of the
Unique Global and Local Imperfection (UGLI) method introduced by the Linear Buckling Analysis (LBA) executed on the structure at elevated
EN1993-1-1 [1] and published by Chladný et al. [2]. The original temperature, namely the elastic critical load factor, the buckling mode
method was valid for structural members subjected to flexural buckling. shape and the second order internal moments induced by the modal
Later, Agüero et al. generalized the UGLI method for lateral-torsional geometric imperfection.
buckling [3], while Papp extended it to assess structural members sen­ The methodology assumes that also in case of elevated temperature
sitive the interaction of flexural and lateral-torsional buckling [4]. In [5] any complex global buckling mode can be classified into one of the finite
the application of the OIM was introduced for beam-columns with more number of fundamental buckling modes (e.g. flexural buckling, lateral-
different load conditions, at normal member temperature. Moreover, the torsional buckling or interaction of them, etc.).
accuracy of the method was examined for beam-columns with different Furthermore, it is also assumed that the Ayrton-Perry formula based
load conditions in [6]. It was found that the OIM at normal member standard buckling curves calibrated for elevated temperatures are
temperature is safe and has a similar safety level to the available design available for the fundamental global buckling modes, and by using them
methods. the required reliability level can be ensured.

* Corresponding author.
E-mail addresses: samer.nemer.91@gmail.com (S. Nemer), jozsef.szalai@consteelsoftware.com (J.A. Szalai), pappfe@sze.hu (F. Papp).

https://doi.org/10.1016/j.engstruct.2023.115884
Received 23 October 2022; Received in revised form 28 January 2023; Accepted 20 February 2023
Available online 28 February 2023
0141-0296/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
S. Nemer et al. Engineering Structures 283 (2023) 115884

Nomenclature moment Ma;


MIIcr,θ,ep modal second order internal bending moment generated
Acronyms by buckling mode shape at elevated temperature, taken in
OIM Overall Imperfection Method – a method for the equivalent point;
assessment of the global buckling resistance of structural MIIfi,θ,Ed internal bending moment about axis y or z calculated by
members and structures using equivalent geometrical GNIA at elevated temperature;
imperfection where the checking the cross-section ΔMfi,θ,Ed secondary internal bending moment in case of class 4 cross-
resistance using geometrically nonlinear analysis involves sections at elevated temperature, due to shift of centroid;
the checking the global stability resistance. Mfi,θ,Rd design cross-sectional resistance for pure bending at
GNIA Geometrically Nonlinear Analysis with Imperfection – an elevated temperature in fire design situation (see EN1993-
elastic analysis using initial geometrical imperfection and 1-2);
assuming geometrically nonlinear behavior. Msec,θ bending cross-sectional resistance about axis y or z at
GMNIA Geometrically and Materially Nonlinear Analysis with elevated temperature;
Imperfections – a materially and geometrically nonlinear Nfi,Ed design axial load at fire design situation;
analysis using relevant initial imperfections. Sometimes it NI internal axial force calculated by linear elastic first-order
is referred as ‘simulation’ since it is assumed that using analysis, in the paper it is always equal to the buckling-
characteristic model parameters the result of the analysis active internal axial force Na;
leads to characteristic value of the ultimate load. Ncr critical axial force for flexural buckling mode;
LBA Linear Bifurcation Analysis – a buckling analysis using Ncr,z,θ critical axial force for flexural buckling mode around weak
linearly elastic material law and assuming linear axis at elevated temperature;
relationship between the load factor and the internal forces Ncr,x,θ critical axial force for torsional buckling mode at elevated
and moments. temperature;
UGLI Unique Global and Local Imperfection – an initial NIIfi,θ,Ed internal axial force calculated by GNIA at elevated
geometrical imperfection in the shape of a relevant temperature;
buckling mode which shape involves both the bow and the Nfi,θ,Rd design cross-sectional resistance for pure compression at
sway type buckling modes. elevated temperature in fire design situation (see EN1993-
ERM Equivalent Reference Member – a simply supported, 1-2);
uniform and straight structural member with constant Nu ultimate value of compressive load of structural test;
internal forces and moments, which member has the same Usec utilization of cross-sectional resistance
critical load factor than the examined member has; the Umax,θ maximum utilization of the cross-sectional resistance along
properties of the equivalent reference member are derived the examined member at elevated temperature;
from the properties of the examined member at the UIIsec,cr,θ second order cross-section utilization from the model
equivalent point; in the equivalent point the utilization of internal moments, which are generated by shape of
the cross-sectional resistance calculated from the modal buckling mode with arbitrary amplitude, at elevated
shape generated internal forces and moments has temperature;
maximum. W cross-sectional modulus about axis y or z, or for warping,
APF Ayrton-Perry Formula – an equation for buckling reduction considering class of section;
factor related to the first yielding design state of a column Weff effective sectional modulus about axis y or z, or for
with initial bow imperfection, assuming geometrically warping, in case of class 4 sections.
nonlinear behavior; the equation involves a parameter
called imperfection factor which ensures the calibration of Latin lower-case letters
the formula. ep location of the equivalent point along the examined
BMC Buckling Mode Class – a possible fundamental buckling structure;
mode (e.g. flexural buckling or lateral-torsional buckling) vcr,θ flexural component of the shape of the elastic buckling
of the examined structural member. mode at elevated temperature;
ECCS European Convention for Constructional Steelworks. y strong axis of the cross-section;
z weak axis of the cross-section;
Latin upper case letters fy design yield strength;
A area of the cross-section; ky,θ reduction factor from Section 3 for the yield strength of
Aeff effective area for class 4 cross-sections; steel at elevated temperature θ;
BIIfi,θ,Ed internal bimoment calculated by GNIA at elevated kp0.2,θ reduction factor for the yield strength at elevated
temperature; temperature in case of class 4 cross-sections;
Bsec,θ warping cross-sectional resistance at elevated temperature; n number of test cases to be considered;
BIIcr,θ,ep modal second order internal bimoment generated by the re reduced ultimate load parameter calculated
buckling mode shape at elevated temperature, taken in experimentally;
equivalent point; rt reduced ultimate load parameter calculated numerically.
E modulus of elasticity;
G shear modulus of elasticity; Greek lower case letters
Leq equivalent length of ERM; α imperfection factor in the AP formula of the flexural
Mcr critical bending moment for lateral-torsional buckling buckling curve at fire design situation;
mode; αult minimum load amplifier of the design load to reach the
Mfi,Ed design bending moment load at fire design situation; characteristic resistance of the most critical cross-section;
MIy internal bending moment around the strong axis (y) αLT imperfection factor in the AP formula of the lateral
calculated by linear elastic first-order analysis, in the paper torsional buckling curve at fire design situation;
it is always equal to the buckling-active internal bending δeq equivalent amplitude for the initial geometrical

2
S. Nemer et al. Engineering Structures 283 (2023) 115884

imperfection; member (compression and bending).


αcr,θ minimum force amplifier to reach the elastic critical
buckling load at elevated temperature; Subscripts which refers to
αsec,θ cross-sectional resistance multiplication factors taking the sec cross-sectional parameter;
buckling active load (N, M or N + M) into account at cr elastic buckling state;
elevated temperature; θ elevated temperature;
eq equivalent value;
λ non dimensional slenderness for the flexural buckling;
Ed design action calculated by elastic theory;
λθ the reduced non-dimensional slenderness for the flexural
fi fire design situation;
buckling mode for the temperature θ;
y strong axis of cross-section;
λLT non dimensional slenderness for the lateral torsional weak axis of cross-section;
z
buckling; warping;
ω
λLT,θ the reduced non-dimensional slenderness for the lateral- effect by axial force;
N
torsional buckling mode for the temperature θ; effect by bending moment;
M
θ the temperature [0C] coupled axial force and bending moment effect;
NM
φcr torsional component of the shape of the elastic buckling ultimate load carrying capacity;
u
mode; elastic cross-sectional property;
el
μcr,θ shape of elastic buckling mode with arbitrary amplitude; plastic cross-sectional property;
pl
μcr,eq equivalent geometrical imperfection in the shape of the effective cross-sectional property of class 4 sections;
eff
elastic buckling mode with arbitrary amplitude; lateral-torsional buckling mode.
LT
μθ interaction factor for interpolating the beam-column
imperfection factor at elevated temperature; Superscripts which refers to
a
ηN,θ shape of clastic critical buckling mode of the column buckling-active effect (which directly causes the buckling);
I
member (pure compression); calculated by linear elastic (1st order) theory;
II
ηM,θ shape of clastic critical buckling mode of the beam member calculated by geometrically non-linear elastic (2st order)
(pure bending); theory;
ERM
ηNM,θ shape of clastic critical buckling mode of the beam-column calculated on the equivalent reference member.

1.3. Scheme of the generalized OIM called equivalent point of the member is determined (equivalent point is
in the cross-section where the utilization is maximum due to the second
Fig. 1 shows the main steps of the OIM procedure. The steps can be order modal internal forces). The properties of the equivalent reference
grouped into two theoretical blocks: member (ERM), such as cross-section, internal forces and buckling mode
type, are taken from the equivalent point. The length of the ERM is
• Universal Transformation (BS-1) determined by the equality of the critical load factors of the examined
• Analytical Solution (BS-2) structural member and the corresponding ERM. For the practical details
see the Section 2 of the article.
The theoretical contents of the blocks are explained shortly in the The validity of the Universal Transformation was checked by Hajdú
following sections, the detailed practical contents in the Section 2. [11] for normal temperature: his results confirmed the applicability of
the method. In this paper it is assumed that this conclusion is valid also
1.3.1. Universal transformation (BS-1) for elevated temperatures.
By the steps of Universal Transformation block (BS-1) the examined
structural member subjected to a complex buckling mode is converted 1.3.2. Analytical solution (BS-2)
into a properly defined Equivalent Reference Member (ERM) which is The steps of the Analytical Solution block (BS-2) employ the closed-
the prototype member of the corresponding fundamental buckling mode form equations provided by the theory of elastic stability [12] and the
– this is the complex generalization of the Effective Length Method [10], standard buckling curves corresponding to the equivalent global buck­
which method is often used in structural engineering. ling mode [9] – this is the generalization of the interaction equations,
First, numerical analyses should be performed on the perfect model which type of equations are often used in the design standard like EC3.
of the structural member with the imposed loads, namely: Linear Elastic The OIM uses the imperfection factors of the Ayrton-Perry formula based
Analysis (LA) and Linear Buckling Analysis (LBA). In the next step the so standard design curves to calculate the initial equivalent amplitude for

Fig. 1. The scheme of the generalized OIM.

3
S. Nemer et al. Engineering Structures 283 (2023) 115884

the geometrical imperfection given in the shape of the buckling mode. the buckling mode shape is the highest (in other words, the second order
Thus, in this block we need first to calculate the amplitude scale factor flexural curvature of the compressed flange from the buckling mode is
for the buckling shape to get the equivalent geometrical imperfection. If the highest, as it is defined in the paragraph 5.3.1 (11) of the EN 1993-1-
the design model of the examined member is loaded with this initial 1). This specific highest modal utilization is assumed to be equal to the
equivalent geometrical imperfection, and internal forces and moments same one calculated on the ERM forming the equivalency relationship
calculated by geometrically nonlinear analysis (GNIA), the checking of between the examined model of the structural member and the ERM:
the cross-section resistance assuming conservative interaction is equiv­ II
Usec,cr,θ II,ERM
(ep) = Usec,cr,θ (1)
alent to the checking of the global buckling resistance.
In Eq. (1) the left hand side symbolizes the modal utilization calcu­
1.4. Content of current research lated in the equivalent point (ep) of the examined structural member,
while the right hand side symbolizes the modal utilization of the ERM
In this paper first the main steps of the OIM used for elevated tem­ (evidently in the midspan point), where the II superscript refers to sec­
perature are explained, then all the steps of the method are described. ond order theory and the θ subscript to the elevated temperature. The
The validation of the method is carried out through the examination of position of the equivalent point of the examined structural model can be
1100 simple regular structural members composed of hot-rolled I cross- determined according to the following condition:
sections. The global buckling strengths calculated by the method are II
Usec,cr,θ II
(ep) = maxUsec,cr,θ (x) → x = ep (2)
compared to reference values calculated by geometrically and materi­
ally nonlinear analysis with imperfections (GMNIA) as well as by the Generally, the ERM is a straight, simply supported member with
corresponding design formulas of the EN 1993-1-2 standard. The uniform cross-section subjected to constant compressive force and/or
GMNIA is nowadays considered the most accurate and advanced bending moment. To determine the cross-section, the length, the loading
approach for testing and validating structural analysis and design and the buckling mode of the actual ERM, the information from the
methods. According to this comparative study, conclusions are drawn equivalent point of the examined structural model is utilized:
regarding to the reliability of the OIM and the accuracy of the standard
design formulas used for fire design of structural members. • the cross-section of the ERM will be the same as the cross-section of
It should be noted that the proposed method may be extended to the structural member at the equivalent point;
irregular members with other types of cross-sections (mono-symmetric • the constant axial force and moments of the ERM will be same as the
and asymmetric), loading and boundary conditions, but the detailed internal axial force and moments calculated at the equivalent point
explanation of this is out of the scope of this paper. of the examined structural member.

2. Steps of OIM at elevated temperatures To determine the Leq length of the ERM firstly the proper funda­
mental buckling mode should be selected from the Table 1. Since the
This chapter covers the description and explanation of the steps of cross-section and the loading at the equivalent point of the examined
the OIM as well as presents the step-by-step procedure of the complete structural member are identical to the ones of the ERM, the elastic
method at elevated temperatures. The application of the method is critical load factors should be equal in order to satisfy Eq. (1). For the
illustrated by a simple example where the result is compared to the calculation of the elastic critical load of the ERM there are well-known
result of the geometrically and materially nonlinear analysis with im­ analytical formulas for each fundamental buckling modes (e.g. see
perfections (GMNIA). [12]) from which the equivalent length of the ERM can be calculated
from the equivalency of the elastic critical load factors. By this, the ERM
2.1. ‘Structural Member → ERM’ transformation is fully defined.

The ‘Structural Member → ERM’ transformation is illustrated in Fig. 2.


The details of the transformation in case of members with normal
temperature are described in the paper published by Szalai and Papp
Table 1
[7]. This transformation technique is based on the following assump­ The fundamental Buckling Mode Classes (BMC) in case of double symmetric
tion: the most intense information about the mechanical nature of the cross-sections.
actual complex buckling problem can be received from a certain point
BMC Cross- Active load Buckling mode Buckling
on the structural model where the utilization of the cross-section resis­ section at components at shape mode
tance calculated from modal second order internal forces generated by the ep the ep displacement(s)*
at the ep

BMC_01 doubly NI wcr flexural


symmetric buckling
about strong
axis (FBy)
BMC_02 NI vcr flexural
buckling
about weak
axis (FBz)
BMC_03 NI φcr torsional
buckling (TB)
BMC_04 MIy vcr;φcr lateral-
torsional
buckling
(LTB)
BMC_05 NI; MIy vcr;φcr coupled
buckling (FBz
+ LTB)

* v and w denote the displacements in directions of the principle axes of the


Fig. 2. The illustration of the ‘Structural Member → ERM’ transformation. cross-section, while φ denotes the rotation around the member axis.

4
S. Nemer et al. Engineering Structures 283 (2023) 115884

Table 2 Table 3
Imperfection factor for flexural buckling at elevated temperature. Imperfection factor for lateral-torsional buckling at elevated temperature.
Parameter Class of cross-section (EN1993-1-2) Parameter Class of cross-section (EN1993-1-2)

Class 1; class 2; class 3 Class 4 Class 1; class 2; class 3 Class 4


√̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅
Reduced slenderness ky,θ kp0.2,θ Reduced slenderness ky,θ kp0.2,θ
λθ = λ where λ = λθ = λ where λ = λLT,θ = λLT ; λ = λLT,θ = λLT ; λLT =
kE,θ kE,θ kE,θ kE,θ
√̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅̅̅̅̅̅̅
Afy Aeff fy Wfy Weff fy
Ncr Ncr Mcr M
√̅̅̅̅̅̅̅̅ ̅ √cr ̅̅̅̅̅̅̅̅̅
Imperfection 235 Imperfection 235
coefficient αLT = 0.65 coefficient αLT = 0.65
fy fy
Imperfection factor ηN,θ = αλθ Imperfection factor ηM,θ = αLT λLT,θ
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Note: Aeff must be calculated using ε= 235/fy. Notes: (i) W should be calculated according to the class of cross-section.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
(ii) Weff should be calculated using ε= 235/fy.

2.2. The solution of the ERM


amplitude. The right-hand side of Eq. (1) can be calculated from the
Once the ERM is completely defined, the analytical solution gives the imperfection coefficient of Eq. (3) using the generalized APF defined in
necessary information which is needed for the solution of the buckling [7] and [13]:
problem of the examined structural member. The analytical solution of 1 1
the ERM is based on the generalized Ayrton-Perry Formula (APF). The
II,ERM
Usec,cr,θ = ηNM,θ 1
(6)
αsec,NM,θ 1 −
key parameter of the APF is the imperfection factor which is responsible αcr,θ

for the correct consideration of the second order effect and which en­ Using the equivalency relationship of Eqs. (1) with Eqs. (5)–(6) the
sures the standard safety calibration. Following Szalai [13], the APF standard conform equivalent imperfection amplitude can be calulated:
based imperfection coefficient for coupled flexural and lateral-torsional
buckling can be written in case of elevated temperature as follows:
αcr,θ 1
δeq = ηNM,θ (7)
αsec,NM,θ Usec,cr,θ (ep)
αsec,NM,θ α
ηNM,θ = η + μ sec,NM,θ ηM,θ (3)
αsec,N,θ N,θ αsec,M,θ
In Eq. (3) ηN,θ and ηM,θ are the calibrated standard imperfection 2.4. Checking utilization of global buckling resistance
factors corresponding to the buckling modes for flexural buckling and
lateral-torsional buckling at elevated temperature respectively, αsec,N,θ The final step of the OIM is checking the utilization of cross-sectional
αsec,M,θ and αsec,NM,θ are the cross-sectional resistance multiplication resistances of the examined structural member. For this the internal
factors taking the buckling active loads into account, and μθ is an forces and moments should be calculated by GNIA using the initial
interaction factor dependent on the pure elastic critical loads of the equivalent geometrical imperfection, as follows:
equivalent reference member [13]: ηcr,eq (x) = δeq μcr,θ (x) (8)
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
/
1 − αcr,θ N I Ncr,z,θ The examined structural member is adequate for global buckling
μθ = / (4)
1 − αcr,θ N I Ncr,x,θ mode at elevated temperature if the Umax,θ maximum utilization of cross-
sectional resistances along the examined member is equal or <1.0.
In Eq. (4) αcr,θ is the critical load factor of the structural member, Ncr, Consequently to the basic condition of the APF the Umax,θ must be
z,θand Ncr,x,θ are the flexural and the torsional critical load of the ERM at calculated with the following cross-section class dependent linear
elevated temperature. The imperfection factors are determined in the EN interaction formulas:
1993-1-2 for all the classes of cross-sections [9], see Table 2 and Table 3.
• class 1 and class 2 cross-sections:
2.3. The ‘ERM → Structural Member’ transformation [
II II II
]
Nfi,θ,Ed My,fi,θ,Ed Mz,fi,θ,Ed BIIfi,θ,Ed
Umax,θ = max + + + (9)
The δeq equivalent amplitude of the equivalent initial geometrical ky,θ Afy ky,θ Wpl,y fy ky,θ Wpl,z fy ky,θ Wpl,B fy
imperfection must be determined for the ηcr,θ shape of the elastic
buckling mode of the examined structural member. It is done using
• class 3 cross-sections:
again Eq. (1), where the left-hand side can be calculated as follows:
[ ]
II II II
Nfi,θ,Ed My,fi,θ,Ed Mz,fi,θ,Ed BIIfi,θ,Ed
II
Usec,cr,θ (ep) = δeq Usec,cr,θ (ep)
1
(5) Umax,θ = max + + + (10)
αcr,θ − 1 ky,θ Afy ky,θ Wel,y fy ky,θ Wel,z fy ky,θ Wel,B fy

In Eq. (5) Usec,cr,θ(ep) is the modal cross-sectional utilization at the


equivalent point computed from the buckling mode shape with arbitrary • class 4 cross-sections:

[ ]
II II II
Nfi,θ,Ed My,fi,θ,Ed + ΔMy,fi,θ,Ed Mz,fi,θ,Ed + ΔMz,fi,θ,Ed BIIfi,θ,Ed
Umax,θ = max + + + (11)
kp0.2,θ Aeff fy kp0.2,θ Weff ,y fy kp0.2,θ Weff ,z fy kp0.2,θ Weff ,w fy

5
S. Nemer et al. Engineering Structures 283 (2023) 115884

Table 4 3.2. Benchmark example


The step-by-step procedure of the OIM at design situation of elevated
temperature. For easier understanding, the application of the method is illustrated
BS-1 in this chapter by an example. Fig. 3 shows a simply supported beam-
Step 1: Structural analysis of perfect model of structural member or structure with material
column member whose global stability resistance will be determined
properties at elevated temperature by the OIM.
1.1 Linear Elastic Analysis (LA) → NIθ; MIy,θ The cross-section of the member is doubly-symmetric hot-rolled
1.2 Linear Buckling Analysis (LBA) → αcr,θ; ηcr,θ IPE160 shape. The grade of the steel is S235. According to the EN1993-
Step 2: Determination of the equivalent point
1-1 and EN1993-1-2 at both normal and elevated temperature, respec­
2.1 Calculation of internal second order moments generated by buckling mode
deformation → MIIz,cr,θ,ep;BIIcr,θ,ep tively, this cross-section belongs to the class 1 category (plastic cross-
2.2 Calculation of cross-section resistances → Mz,sec,θ; Bsec,θ section), in case of any type of internal force. The member is affected
2.3 Calculation of cross-section utilizations → UIIMz,sec.cr,θ; UIIB,sec,cr,θ;UIIsec,cr,θ by elevated temperature θ = 500 ◦ C. According to the EN1993-1-2, at
2.4 Determination of location of equivalent point → ep 500 ◦ C the steel material has the following mechanical properties: fy,θ =
Step 3: Buckling mode classification through the equivalent point
3.1 Determination of buckling active force and moment → Na,ep,θ May,ep,θ3.2 Classify
183.3 N/mm2 yield strength, Eθ = 126000 N/mm2 elastic moduli and Gθ
the actual buckling mode into appropriate fundamental case = 48462 N/mm2 shear moduli. It is important to realize that the relevant
(Table 1) → BMC ERM of this beam-column member is the examined member itself.
Step 4: Equivalent length of the equivalent reference member For design load the ultimate load dcomputed by GMNIA is used.
Calculation of length of ERM from equality of critical load factors of ERM and
Fig. 4 shows the result of the GMNIA, where the ultimate load consists of
examined structure → Leq
BS-2 Nu,GMNIA,θ = 60.35 kN axial compressive force and Mu,GMNIA,θ = 3.725
Step 5: Buckling-active cross-sectional load multiplication factor kNm bending moment acting at the ends of the member.
Calculation of linear cross-sectional load multiplication factor from buckling-active Regarding the numerical model used to obtain the GMNIA buckling
loads capacity, the beam-columns investigated here are pin-supported. In
→ αsec,NM,θ
order to simulate this condition, two reference points (one at each side)
Step 6: Equivalent imperfection factor
Calculation of the equivalent standard imperfection factor of the ERM considering were coupled with the nodes of the corresponding end surface of the
the relevant BMC and Eq. (3) → ηNM,θ member using kinematic coupling restraints. Boundary conditions were
Step 7: Equivalent geometrical imperfection applied through these reference points at the member ends as follows:
7.1 Calculation of the equivalent scale factor → δeq
The reference points of the member were restrained against all degrees
7.2 Calculation of the equivalent geometrical imperfection as the scaled buckling
mode shape → ηcr,eq of freedom except for the displacement in the direction of the applied
Final load at the loaded end, and the rotations about the axes of buckling at
Step 8: Checking global buckling resistance both ends.
8.1 Geometrically Nonlinear Analysis with Imperfection (GNIA) executed on the The loads were modeled by applying distributed forces (by means of
examined structural member with the equivalent geometrical imperfection
nodal forces) on the flanges and on the web of the two loaded ends using
computed in Step 7.2 → NIIfi,θ,Ed; MIIy,fi,θ,Ed; MIIz,fi,θ,Ed; BIIfi,θ,Ed8.2 Checking cross-sectional
utilizations using conservative interaction formula according to the modified RIKS tool (also known as Arc length method) available in
Eq. (9), Eq. (10) or Eq. (11), which is the relevant → Umax,θ the ABAQUS library. More detailed description of the numerical model
used for the GMNIA is given later in Section 4.3.
In this example this ultimate load is assumed as design load in fire
situation. Basically, the OIM is a CAD-oriented procedure, consequently
In Eq. (9)–(11) the effect of the B bimoment related to warping the detailed and hand-calculated presentation of the example would
torsion can be generally quite relevant, so it is considered as it is pro­ extend beyond the size of this paper. Therefore, helping the easier
posed by the background paper for the second generation of the Euro­ reproduction of the steps of the OIM, the contents and the results of the
code 3 [14], and the cross-section properties denoted by ‘eff’ may be OIM steps are summarized below (in case of example in Fig. 3).
calculated according to the EN1993-1-2 E.2 (2). Step 1.1: Linear Elastic Analysis (LA) of the structural member
The internal forces will be selected after the equivalent point is
3. OIM steps at elevated temperature through a benchmark determined in Step 2.4:
example

3.1. Step-by-step procedure

The step-by-step procedure of the OIM at design situation of elevated


temperature is summarized in the Table 4. For the easier understanding
the application of the OIM is illustrated by an example.

Nu,GMNIA,
Mu,GMNIA,

Fig. 3. Structural model of the examined beam-column member. (load in fire Fig. 4. Ultimate load of the example model calculated by GMNIA using a model
design situation: Nfi,Ed = 60.35kN; My,fi,Ed = 3.725kNm). with characteristic properties.

6
S. Nemer et al. Engineering Structures 283 (2023) 115884

- internal axial force: NIep,θ = 60.35 kN Table 5


- internal bending moment about strong axis: MIy,ep,θ = 3.725 kNm The Comparison of the results calculated by different methods.
method resistance utilization Umax,θ ultimate load parameter αult
Note: internal forces will be selected after the equivalent point is deter­
GMNIA by Abaqus – 1
mined in Step 2.4; in the case of this example this step would be neglected, but EN1993-1-2 4.2.3.5 1.172 0.853
it is needed in general cases. OIM 1.028 0.987
Step 1.2: Linear Buckling Analysis (LBA)
Applying material properties determined for elevated temperature,
using 14DOF thin-walled beam-column finite element method: Step 4: Equivalent length of ERM
Generally, the theory of elastic stability of structures provides rele­
- elastic critical load amplifier: αcr,θ = 1.550 vant buckling formulas to express the equivalent member length. In case
- shape and amplitude of the global buckling mode: ηcr(x) where of coupled buckling of FB and LTB the relevant formula is as follows:
vcr.max,θ =65.27 mm √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )( )̅
αcr My = r Ncr,z − αcr N Ncr,x − αcr N → Leq
Note: Displacements relate to the centroid of the cross-section; the value of where N = αcr N a and My = αcr Mya
the amplitude can be arbitrary (some software normalize it as unit.
In the present example the equivalent length is equal to the member
Step 2.1: Modal second order internal moments generated by the
length since the examined structural member itself is the relevant ERM.
buckling mode shape
Leq = 2770 mm.
The bending moment about weak axis and the bimoment should be
Step 5: Buckling-active load multiplication factor of ERM
computed from the modal deformation (modal strains) of the examined
The calculation of the buckling-active load multiplication factor of
structure assuming second order linear elastic theory:
ERM is based on the EN1993-1-2. For the buckling-active force and
moment see Step 3.1:
- Bending moment around weak axis: MIIz,cr,θ,ep = 7.249 kNm
- reduced slenderness for flexural buckling: λz = 1.6000; λz,θ = 1.824.
- Bimoment: BIIcr,θ,ep = 0.123 kNm2
- reduced slenderness for lateral-torsional buckling: λLT = 1.036;
λLT,θ = 1.182.
Note: These internal forces will be selected after the equivalent point is
- imperfection coefficients: αz = 0.650; αLT = 0.650.
determined in Step 2.4.
- imperfection factors: ηN,θ = 1.186; ηM,θ = 0.768.
Step 2.2: Modal cross-section resistances
- load multiplication factors: αsec,N,θ = 6.102; αsec,M,θ = 6.087; αsec,
The modal cross-sectional resistances should be calculated using
NM,θ = 3.047.
modal internal bending moment and the bimoment calculated in Step
Step 6: Equivalent imperfection factor of ERM
2.1 and using cross-sectional properties according to the cross-sectional
The calculation of the equivalent imperfection factor is based on the
class:
standard imperfection factors of the pure load cases calculated in Step 5:
- pure bending moment effect around weak axis: Mz,sec,θ = 4.758
- modifying factor: μθ = 0.435.
kNm.
- equivalent imperfection factor: ηNM,θ = 0.759.
- pure warping moment effect: Bsec,θ = 0.348 kN/m2.
Note: for the formulas see Eq. (3).
Step 2.3: Modal cross-section utilization
Step 7.1: Equivalent scale factor
The modal cross-sectional utilizations should be calculated from the
For the equivalent scale factor see Eq. (7).
internal moments calculated in Step 2.1 and cross-sectional resistances
calculated in Step 2.2:
- δeq = 0.206
- Component from bending moment around weak axis UIIMz,sec.cr,θ,ep =
1.523.
Step 7.2: Equivalent geometrical imperfection
- Component from bimoment: UIIB,sec,cr,θ,ep = 0.354.
For the calculation of the equivalent geometrical imperfection see
- Full utilization UIIsec,cr,θ,ep = 1.877.
the Eq. (8), where the equivalent scale factor is calculated in Step 7.1,
Note: These values will be selected after the equivalent point is determined
the buckling shape of the examined structure is calculated in Step 1.2:
in Step 2.4.
Step 2.4: Location of the equivalent point (from the reference end of
- ηcr,eq = 13.43 mm
the member)
Generally, the equivalent point can be determined by analyzing the
Step 8.1: Internal force and moments by geometrical nonlinear
distribution of the modal cross-sectional utilization along the member.
analysis with equivalent geometric imperfection.
In the present example, the location of the equivalent point was known
Geometrically Nonlinear (materially linear) Analysis with equivalent
in advance: the midspan of the examined structural member (ep = 1385
geometric imperfection (GNIA) is used to compute the second order
mm).
internal forces and moments of examined member subjected to elevated
Step 3.1: Buckling-active internal force and moment
temperature:
The buckling active force and moment are those first order force and
- axial compressive force NIIfi,θ,Ed = 60.35 kN.
moment in the equivalent point that directly cause the buckling, usually
- bending moments MIIy,fi,θ,Ed = 3.800 kNm; MIIz,fi,θ,Ed = 2.700 kNm.
the first order N axial force and My bending moment. All the second
- bimoment BIIfi,θ,Ed = 0.045 kNm2.
order forces and moments are passive:
Step 8.2: Cross-section utilizations
- Axial force: Naep,θ = 60.35 kN.
First the cross-section utilization are calculated along the examined
- Bending moment around the strong axis: May,θ,ep = 3.725 kNm.
structure, than the maximum utilization (utilization of the critical cross-
Note: for the values see Step 1.1.
section) is considered as the global buckling resistance. For this Eq. (9),
Step 3.2: Classification of buckling mode.
Eq. (10) or Eq. (11), is used, which is the relevant:
The Buckling Mode Class (BMC) can be determined by the analysis of
the buckling shape and the buckling active internal forces of the
- Umax,θ = 1.028
examined structure (see the Table 1):
The ultimate load factor (where the Umax,θ = 1.0):
- BMC = Coupled FB & LTB

7
S. Nemer et al. Engineering Structures 283 (2023) 115884

Table 6 modeled using shell elements in ABAQUS is not totally equal to the
Properties of the cross-sections used in the numerical investigation. actual cross section of hot-rolled I-shaped members because the fillet
Cross- h b tw tf A class of cross- radius at the web-flange intersections cannot be modeled directly in
section (mm) (mm) (mm) (mm) (mm2 ) section exact way. Besides, there is a small area at the modeled cross-section at
IPE 100 100 55 4.1 5.7 1032 1
the web-flange intersection that is taken into account twice. For this,
IPE 160 160 82 5.0 7.4 2009 1 beam elements with SHS cross-section (B31) were added at the web-
IPE 180 180 91 5.3 8 2395 1 flange intersections. The width and thickness of the added SHS profile
IPE 240 240 120 6.2 9.8 3912 1 (bSHS , tSHS ) were calculated to compensate for these lacking moments
IPE 300 300 150 7.1 10.7 5381 1
around the y-axis (Iy,lac ) and torsional moment of inertia (It,lac ) between
HE300A 290 300 8.5 14 11,253 2
HEB 300 300 300 11 19 14,908 1 the actual rolled cross-section and the modeled cross-section, as follows:
HEB 340 340 300 12 21.5 17,090 1 √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
HEB 500 500 300 14.5 28 23,864 1 It,lac .h0 2
bSHS = (12)
Iy,lac − 23It,lac
- αult = 0.987
It,lac
tSHS = (13)
b3SHS
The global buckling resistance of the examined structural member
computed by GMNIA, by EN1993-1-2 and by OIM is summarized in the where h0 is the distance between the center lines of flanges of the
Table 5. One can see that the OIM’s result is very close to the reference cross-section.
result computed by GMNIA, while the result given by the EN1993-1-2 is The mesh size is determined as follows: 16 elements in the flange, 16
very conservative. elements in the web depth.
While along the member length, the size of the elements is 20 mm, as
4. Numerical investigation shown in Fig. 5.

In order to assess the accuracy of the proposed OIM a reference 4.3.2. Load and boundary conditions
database is needed, to which the OIM’s results can be compared. In the All members investigated in this study are pin-supported. For simu­
present investigation it is assumed that the appropriate database can be lating these boundary conditions, two reference points (one at each
created using GMNIA, and using a structural model with characteristic member end) were coupled with the nodes of both end surfaces of the
parameters (especially the characteristic value of the yield strength), the members using kinematic coupling restraints. Boundary conditions were
resulted ultimate load will be characteristic value. The database created applied through these reference points, as follows: the reference points
by this way may be considered as reference for comparing. It is noted of the members were restrained against all degrees of freedom except for
that nowadays this concept is generally accepted in the steel structural the displacement in the direction of the applied load at the loaded end,
engineering research [15]. and the rotations about the axes of buckling at both ends.
The load was modeled by applying distributed forces (by means of
4.1. Test program nodal forces) on the flanges and on the web of the loaded end using the
modified RIKS tool (Arc length method) which is available in the ABA­
The evaluation of the accuracy of proposed OIM is based on a nu­ QUS library [20].
merical test program. The test program is consisted of eight different
cross-sections (see Table 6), seven different non-dimensional slender­ 4.3.3. Geometrical and material imperfections
ness ranging from 0.5 to 2.3 and three different temperatures (400 ◦ C, Regarding the imperfections, the initial geometrical imperfections of
500 ◦ C, 600 ◦ C). Steel grade S235 is used for all members. It is worthy to beam-columns are introduced in the numerical analysis by first per­
mention that the focus of this study is to propose the OIM for global forming a linear buckling analysis (LBA) on the perfect prismatic
buckling capacity of steel beam-columns at elevated temperatures. member with given boundary conditions, then the relevant normalized
Therefore, only stocky (class 1 or class 2) cross-sections are considered global buckling mode is extracted. In this way, the first global buckling
in order to avoid any local effects and imperfections. mode shape (corresponding to the lowest elastic critical load) derived by
the linear buckling analysis is introduced into non-linear finite element
model (GMNIA), multiplying this by the amplitude of initial geometrical
4.2. Program implementing OIM
(bow) imperfection and updating the nodal coordinates of the model by
adding the established nodal imperfections. The amplitude of initial
In order to perform the numerical test program, a proper code for the
geometrical imperfection of the members is taken equals to L/1000,
implementation of the OIM has been developed in GNU Octave software
which is used in most studies in the literature [21] and corresponds to
[16]. The validation of this own developed program was published in
75 % of the recommended tolerance value of L/750 for steel column in
[17] and [18]. The program uses 14 degrees of freedom thin-walled
Annex D of EN1090-2:2008 [22], where L is the member length. Only
beam-column finite element method according to [19]. The program
global buckling analysis is considered for this study. Therefore, local
runs the steps of the OIM defined in Table 4.
imperfections are ignored.
The basic load for any examined structural member is generated by
The ECCS type residual stress model for hot-rolled cross-sections is
GMNIA (see Section 5.3).
taken into account [23] in this investigation. This model is the basis for
The ultimate load factor within the OIM is found with an iteration
the European buckling curves, and this is the most commonly used re­
procedure. This factor indicates the deviation from the reference ulti­
sidual stress pattern in the literature [6 21]. The magnitude of the initial
mate load computed by GMNIA.
stress depends on height to width ratio of the section analyzed, as
illustrated in Fig. 6.
4.3. Numerical model for GMNIA
4.3.4. Material properties
4.3.1. General
S235 steel grade (fy = 235 N/mm2) is considered in this study. The
The reference ultimate load for each member case is calculated by
temperature introduced to the numerical model is considered to be
GMNIA, for which the Abaqus software is used. For the member models,
uniformly distributed along the member. Thus, it is possible to compare
the general purpose S4R shell element is used. The cross-section

8
S. Nemer et al. Engineering Structures 283 (2023) 115884

Fig. 5. Meshing of the GMNIA model.

the GMNIA results to the result of the proposed OIM and of the valid
EN1993-1-2 formula. For linear buckling analysis, linear elastic material
law with Young’s modulus E = 2.1⋅105 N/mm2 is used. The Poisson’s
ratio is set to 0.3. For the non-linear analysis at elevated temperatures,
the reduction factors for carbon steel are used, according to the Table 3.1
of EN1993-1-2 [12], and the shape of the stress–strain curves given in
the fire part of Eurocode EN1993-1-2 at different temperatures as shown
in Fig. 7 is adopted in the numerical model. The influence of thermal
expansion has not been considered.

4.3.5. Numerical model validation


The ultimate loads obtained from the GMNIA model described above
are compared with the ultimate loads measured in the test performed by
Pauli et al. [24]. The test was implemented on both stub (short) and
slender (long) columns made of HEA100 cross-section. All structural
members were made from hot finished. The experiments were per­
formed at temperatures of 400, 550 and 700 ◦ C. It can be seen from
Table 7 that the predicted buckling loads are generally in good agree­
ment with the test results with maximum difference of 7.6 % and
average difference of 5.7 %. Fig. 7. Stress–strain relationship for S235 carbon steel at elevated
Moreover, The GMNIA model underestimates the ultimate load of temperatures.
steel columns at elevated temperatures for all cases, and thus it provides
a safe prediction for the fire resistance of steel columns. of the load–displacement curves of three different beams subjected to:
It should be mentioned that the difference between the numerical simple bending (section resistance) at 450 ◦ C (Test 3), simple bending at
and the experimental study may be interpreted due to many factors such 650 ◦ C (Test 4), lateral torsional buckling at 450 ◦ C (Test 6). the GMNIA
as human errors, the scale of the member, accurate specifications of curve (solid line) is given by the ABAQUS model described above. The
materials, software errors, mesh dimensions and differences between shown displacement is the vertical displacement of the bottom flange at
the material model and the real material properties can all lead to errors mid-span. Fig. 9 shows the buckling load factors calculated on pin-ended
in the results between analysis and testing. Therefore, additional vali­ column models with IPE 220 cross-sections and S235 steel materials
dations against other experimental and numerical results found in the using SAFIR software by Vila Real et al. [26] and using the GMNIA
literature are presented hereafter. model using Abaqus by the authors. Fig. 10 shows the lateral-torsional
The model is validated against the elevated temperature tests on buckling load factors calculated on the same members like in Fig. 9
beams reported by Prachar et al. in [25]. Fig. 8 shows three comparisons but subjected to constant bending moment.
As it can be observed from Fig. 9, and Fig. 10, there is a good
agreement between the current numerical model and the one used in
[26]. These good predictions from the model indicate that the model is
capable of predicting buckling capacity of steel members at elevated
temperatures, and thus can be used for the comparative investigation
presented in this paper.

Table 7
Comparison of FE and experimental results from Pauli et al. [24].
Column ID Temperature End conditions Nu,test Nu,GMNIA Difference

C y z (kN) (kN)

S19 400 tie tie 996 964 3.3 %


S13 550 tie tie 511 472 7.6 %
M02 400 tie pin 646 615 5.1 %
Fig. 6. Considered residual stress patterns for h/b ≤ 1.2 (left) and h/b > 1.2 M03 550 tie pin 405 375 7.4 %
(right) [23].

9
S. Nemer et al. Engineering Structures 283 (2023) 115884

Fig. 8. Comparisons of numerical and experimental load–vertical displacement curves reported by Prachar et al. and the curve using our developed ABAQUS model
for three different tests given in [25].

Fig. 9. Comparison of results obtained by Vila Real et al. [26] using SAFIR, and Fig. 10. Comparison of results obtained by Vila Real et al. [26] using SAFIR,
the results of the developed GMNIA model for the IPE220 cross-section and the results of the developed ABAQUS model for the IPE220 cross-section
steel column. steel beam.

10
S. Nemer et al. Engineering Structures 283 (2023) 115884

Fig. 11. Interaction curves given by: (i) proposed OIM and EN1993-1-2 with calibrated imperfection factors, (ii) GMNIA.

5. Comparative study conservative.


In the second step, the OIM is applied for all the members of the
In this section, the results of the proposed Overall Imperfection investigation program. Fig. 12 shows scatter-plots where the theoretical
Method are compared with the results of the numerical tests given by value rt,EN (which correspond to the EN1993-1-2 interaction formula)
GMNIA executed on characteristic structural models. and the numerical value re (GMNIA) are compared (the figure on the
left), while the figure on the right shows a comparison between the
theoretical value rt,OIM that corresponds to the OIM results and the nu­
5.1. Accuracy assessment with exact imperfection factors merical value re (GMNIA), where these values are defined as follows:
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
)2̅
In the first step, the OIM and the EN1993-1-2 interaction formula are (
Mu,GMNIA
)2 (
Nu,GMNIA
applied with imperfection factors calibrated from the reference results re = + (14)
Mfi,θ,Rd Nfi,θ,Rd
given by GMNIA in order to get back the same numerical results for the
pure cases (pure compression and pure bending) than those given by √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )2 ( )2
GMNIA and the EN1993-1-2 results. After this, the N-M interaction rt,EN =
Mu,EN
+
Nu,EN
(15)
curves given by the proposed OIM, the EN1993-1-2 interaction formula Mfi,θ,Rd Nfi,θ,Rdd
and the reference GMNIA are plotted in Fig. 11 for two cross-sections √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )2 ( )2
(IPE160 and HEA300) and two different temperatures (500 ◦ C and Mu,OIM Nu,OIM
600 ◦ C). One can see the very good agreement between the OIM and the rt,OIM = + (16)
Mfi,θ,Rd Nfi,θ,Rdd
GMNIA curves, while the EN1993-1-2 curves sometimes seems to be

11
S. Nemer et al. Engineering Structures 283 (2023) 115884

Fig. 12. Statistical evaluation of the results given by (i) the EN1993-1-2 interaction equation (Left) (ii) the proposed OIM method (Right).

√̅̅̅̅̅̅̅̅
Table 8 235
Statistical parameters of the normalized OIM method and EN1993-1-2 interac­ α = 0.65 (17)
fy
tion curve.
Mean C.O.V Min. Max. n n n< In this subsection, the accuracy and reliability of the proposed OIM
value value value <1 0.97 with the imperfection coefficient given in Eq. (17) for assessment of
GMNIA/ 1.017 2.00 % 0.959 1.075 1010 46 3 buckling resistance of steel beam-columns at elevated temperatures are
OIM investigated. The results from the proposed OIM are compared with both
GMNIA/ 1.057 4.79 % 0.979 1.190 1010 40 0 the numerical results (GMNIA) and EN1993-1-2 results, see Fig. 13. First
EN1993-
of all, it can be seen from Fig. 13 that the proposed OIM curve and the
1-2
EN1993-1-2 buckling curve are almost identical for the two fundamental
cases of pure compression and pure bending moment.
where (Nu, Mu)GMNIA is the global buckling resistance given by In addition to that, Fig. 14 shows scatter-plots where the corre­
GMNIA, (Nu, Mu)EN by EN1993-1-2 interaction formula, while sponding theoretical values rt and numerical values re are compared. It
(Nu, Mu)OIM by the proposed OIM method. Each value is normalized by can be seen that there is a relatively good agreement between the pro­
the corresponding cross-sectional resistances Nfi,θ,Rd , Mfi,θ,Rd defined by posed OIM, the numerical results, with the data being inside the 10 %
the EN1993-1-2. unsafe and 10 % safe range. However, it can be noticed from Fig. 13 that
Based on the results presented in Fig. 12 and the statistical evalua­ EN1993-1-2 curve results in unsafe capacities mainly for short member.
tion presented in Table 8, it can be noticed that the mean value of re /rt of Thus, the most unsafe estimations are produced for short members of
the proposed OIM is 1.017 on the safe side. The minimum value is 0.959, low slenderness submitted to axial forces. Therefore, the flexural
and the maximum value is 1.075. Moreover, it is important to see that in buckling curve may require a calibration in order to be able to safely
the examined cases, altogether <5.0 % of the results lied under the perform the stability check of relatively short steel columns at elevated
unsafe side (most of these values for HEA cross-sections), and the co­ temperatures.
efficient of variation (C.O.V) is very low. It should be mentioned that these unsafe estimations are bigger in the
On the other side, the data given by EN1993-1-2 interaction method members made of HE300A cross-section while other investigated cross-
is much more conservative with mean value of 1.057, and maximum sections showed similar behavior to that of IPE160. This discrepancy is
value of 1.19 and also the C.O.V is more than twice larger than it is for probably due to the single imperfection value in fire design situation, as
the OIM. The largest deviations from the GMNIA results can be detected it was shown for the case of lateral-torsional buckling by Vila Real et al.
in the higher slenderness region where the OIM results are still very in [27], while at room temperature different curves are used for different
accurate. cross-section types.
Considering these results, the conclusion may be drawn: if the cali­
bration of the imperfection coefficients of the pure cases (column
buckling and beam buckling curves) is exact, the accuracy of the OIM in 5.3. Accuracy assessment with modified imperfection factor
case of beam-columns is excellent and significantly better than that of
EN1993-1-2 interaction formula. The method is capable to depict the The key parameter for the calibration process is the imperfection
actual behavior given by the numerical model and predict much closer factor α which is the base for the standard safety calibration. The
capacities to the numerical results at elevated temperatures, due to the improvement could be achieved adopting a more conservative value for
deep and accurate mechanical background of the method. the imperfection factor in the AP formula at fire design situation, as
proposed in [21]:
√̅̅̅̅̅̅̅̅
5.2. Accuracy assessment with imperfection factors specified by EN1993- 235
1-2 α = 0.85 (18)
fy

The current standard of EN1993-1-2 [12] adopts only one buckling To verify the correctness of this proposal for beam-column cases, it is
curve with unique imperfection factor at fire design situation: worthy to examine the fundamental cases of flexural buckling of steel

12
S. Nemer et al. Engineering Structures 283 (2023) 115884

Fig. 13. Interaction curves computed by different methods: (i) proposed OIM with EN1993-1-2′ s calibrations; (ii) GMNIA; (iii) EN1993-1-2.

Fig. 14. Validation of the proposed OIM (statistical evaluation).

13
S. Nemer et al. Engineering Structures 283 (2023) 115884

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Fig. 15. Comparison of the flexural buckling curves (left) and lateral-torsional buckling curves (right) of EN1993-1-2 and the modified one with (α = 0.85 235/f y )
against numerical results.

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Fig. 16. Statistical evaluation of the proposed Overall Imperfection Method with (α = 0.85 235/f y ).

columns under pure compression, and lateral torsional buckling of steel By introducing the modified value for the imperfection factor α in the
beams under pure bending. AP formula, the results of the proposed OIM are compared with the
Fig. 15 shows comparisons between the EN1993-1-2 buckling curve, numerical results, as shown in Fig. 16 and Fig. 17. Based on the statis­
the proposed OIM curve (α = 0.65), the modified EN1993-1-2 curve (α tical evaluation parameters presented in Table 9, it can be seen that:
= 0.85), and the numerical results for (i) members under pure axial
compression (left) and for (ii) members under pure bending moment • the OIM with imperfection coefficient given by EN1993-1-2 is
(right). It can be seen clearly that the modified EN1993-1-2 curve with generally unsafe approach with mean value <1, and minimum value
the modified imperfection coefficient of Eq. (18) produces generally equals to 0.889;
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
good results with two exceptions, as follows: • the OIM with the modified imperfection factor α = 0.85 235/fy
leads to much better data with mean value of 1.07 and a better C.O.V.
(1) unsafe predictions of buckling capacity (with maximum differ­ However, for members of high slenderness under lateral torsional
ence of 3 %) for short columns especially that made of HEA cross- buckling the method produces conservative data with maximum
sections; value of 1.205;
(2) for high values of slenderness, relatively conservative results for
members under lateral torsional buckling are produced.

14
S. Nemer et al. Engineering Structures 283 (2023) 115884

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Fig. 17. Comparison between the results of the proposed Overall Imperfection Method with (α = 0.85 235/f y ) and the numerical results (GMNIA).

Table 9
One can conclude that the OIM method leads to safe, accurate, and
Statistical parameters of the proposed OIM method and EN1993-1-2 interaction
stable results. Moreover, the study showed the need to adopt two
curve.
different proper values for the pure cases (FB and LTB), and not only one
GMNIA/OIM GMNIA/OIM GMNIA/ GMNIA/
single value as the case in the current EN1993-1-2.
(α = 0.65) (α = 0.85) EN1993-1-2 (α EN1993-1-2 (α
= 0.65) = 0.85)
6. Summary and discussion
Mean 0.995 1.070 1.029 1.119
value
C.O.V 5.07 % 4.75 % 8.15 % 7.81 % In this paper, the Overall Imperfection Method (OIM) for global
Min. 0.889 0.969 0.882 0.926 stability design of steel beam-column members at elevated temperatures
value was proposed and applied. In order to investigate the accuracy and the
Max. 1.111 1.205 1.226 1.305
value
consistency of the proposed OIM at elevated temperatures, a large
n 1010 1010 1010 1010 number of simply-supported beam-column members were examined by
n<1 462 26 363 57 geometrically and materially nonlinear analysis with imperfections
n< 302 1 229 24 (GMNIA), incorporating hot-rolled I-shape cross-sections at three
0.97
different temperatures. This GMNIA model was validated through a
number of numerical studies found in the literature.
• the EN1993-1-2 interaction formula results in similar data to that of First, the OIM was applied using exact Ayrton-Perry formula type
OIM for pure cases, but lead to more conservative results for inter­ imperfection factors of the fundamental buckling modes (flexural and
mediate slenderness and yields considerably higher C.O.V. the lateral-torsional buckling modes). According to the semi-
probabilistic safety level assessment of the results, it can be seen that
the proposed OIM leads to almost the same results as the GMNIA’s ones.

15
S. Nemer et al. Engineering Structures 283 (2023) 115884

Then, the proposed OIM with an imperfection coefficient of α = Data availability


√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
0.65 235/f y was investigated (this value is adopted in EN1993-1-2 as
No data was used for the research described in the article.
imperfection coefficient for both buckling modes (FB and LTB)). Both
the proposed OIM and EN1993-1-2 resulted in unsafe design mainly for Acknowledgments
short columns.
Thus, a more conservative value of the imperfection coefficient at This research did not receive any specific grant funding from
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
α = 0.85 235/f y was considered. Based on the statistical evaluation, it agencies in the public, commercial or not-for-profit sectors.
was seen that the proposed OIM results in much better results compared
to the numerical results than that of the proposed OIM with α = References
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
0.65 235/f y , but conservative predictions were reported for long [1] EN 1993-1-1:2005. Eurocode 3. Design of steel structures – Part 1–1: General rules
and rules for buildings. Brussels, Belgium: European Committee for
members under bending.
Standardization; 2005.
This shows clearly the need to consider two different imperfection [2] Chladný E, Štujberová M. Frames with unique global and local imperfection in the
coefficients, one for each buckling mode (flexural buckling and lateral- shape of the elastic buckling mode (Part1). Stahlbau 2013;8:609–17. https://doi.
org/10.1002/stab.201310080.
torsional buckling).
[3] Agüero A, Pallarés FJ, Pallares L. Equivalent geometric imperfection definition in
steel structures sensitive to lateral torsional buckling due to bending moment. Eng
7. Conclusion Struct 2015;96(1):41–55. https://doi.org/10.1016/j.engstruct.2015.03.066.
[4] Papp F. Buckling assessment of steel members trough overall imperfection method.
Eng Struct 2016;106:124–36. https://doi.org/10.1016/j.engstruct.2015.10.021.
According to the results of the paper, the following conclusion can be [5] Hajdú G, Papp F, Rubert A. Vollständige äquivalente Imperfectionsmethode für
drawn: biege- und drucbeanspruchte Stahlträger. Stahlbau 2017;86:483–96. https://doi.
org/10.1002/stab.201710471.
[6] Hajdú G, Papp F. Safety assessment of different stability design rules for beam-
• the OIM leads to accurate and consistent results if the imperfection columns. Structures 2018;14. https://doi.org/10.1016/j.istruc.2018.05.002.
factors applied for the pure buckling modes have exact values [7] Szalai JA, Papp F. New stability design methodology through overall linear
(computed with GMNIA); buckling analysis. In: The 14th Nordic steel construction conference, September
18–20, 2019, Copenhagen, Denmark. Berlin: Ernst & Sohn Verlag für Architektur
• the OIM was executed with imperfection factor specified in the valid und technische Wissenschaften GmbH & Co. KG.• ce/papers; 2019. doi: 10.1002/
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
EN1993-1-2 (α=αLT = 0.65 235/fy ); the semi-probabilistic safety cepa.1145.
[8] Papp F, Szalai JA, Movahedi RM. Out-of-plane buckling assessment of frames
level assessment of the results shows that the proposed OIM leads to
through overall stability design method. The 14th Nordic Steel Construct
the same results as those produced by the EN1993-1-2 design for­ Conference, September 18–20, 2019, Copenhagen, Denmark. Berlin: Ernst & Sohn
mula, but both approaches may lead to unsafe design, mainly in case Verlag für Architektur und technische Wissenschaften GmbH & Co. KG. ce/papers;
of short members and under compression. 2019. doi: 10.1002/cepa.1146.
[9] EN 1993-1-2:2005. Eurocode 3. Design of steel structures – Part 1–2: General rules
• the OIM was investigated with the new value of the imperfection Structural fire design. Brussels, Belgium: European Committee for Standardization;
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
factors, namely using the α=αLT = 0.85 235/fy proposal; by this 2005.
[10] Szalai J, Papp F, Hajdú G. Validation of the overall stability design method (OSDM)
proposal, the results show good agreement with the numerical re­ for tapered members, stability and ductility of steel structures. In: Proceedings of
sults given by GMNIA. the international colloquia stability and ductility of steel structures, Prague; 2019.
[11] Hajdú G. The Validity of the universal transformation method in global buckling
design. In: Jeppe J (szerk.), The 14th Nordic steel construction conference. Berlin,
The final conclusion of this study is that the proposed Overall
Germany: Ernst und Sohn; 2019. p. 1–6. doi: 10.1002/cepa.1149.
Imperfection Method is capable of predicting the fire resistance of steel [12] Trahair NS. Flexural-torsional buckling of structures. 1st edition. CRC Press (July
beam-columns with hot-rolled I and H cross-sections, and may be used as 20, 1993); 1993. p. 388, URL: http://worldcat.org/isbn/0849377633.
[13] Szalai JA. Complete generalization of the Ayrton-Perry formula for beam-column
an alternative design method. It is also concluded that the research must
buckling problems. Eng Struct 2017;153:205–23. https://doi.org/10.1016/j.
be concentrated to improve the imperfection coefficients (may be fac­ engstruct.2017.10.031.
tors) of the pure fundamental buckling modes. [14] Knobloch M, Bureau A, Kuhlmann U, Simões L, da Silva HH, Snijder AT, et al.
The authors are sure that the further investigations on irregular Structural member stability verification in the new Part 1–1 of the second
generation of Eurocode 3. Part 2: member buckling design rules and further
members with irregular load and support conditions (those are not the innovations. Steel Construct 2020;13:1–14. https://doi.org/10.1002/
focus of this paper) will justify the above conclusions drawn for simple stco.20200027.
prismatic members with hot-rolled I and H cross-sections. [15] Zhao B, Sanzel A, Wald F, Vila Real P, Hricak J, Jandera M, et al. Development of
simple fire design method for I shape thin wall steel members under simple
bending. Revue Construction Métallique 2014;2.
CRediT authorship contribution statement [16] Eaton JW, Bateman D, Hauberg S, Wehbring R. GNU Octave version 3.8.1 manual:
a high-level interactive language for numerical computations. ISBN 1441413006;
2014. URL http://www.gnu.org/software/octave/doc/interpreter/.
Samer Nemer: Methodology, Visualization, Software, Validation, [17] Hajdú G, Papp F. Safety assessment of different stability design rules for beam-
Formal analysis, Investigation, Writing – original draft, Writing – review columns. Structures 2018;14:376–88. https://doi.org/10.1016/j.
& editing. József A. Szalai: Writing – original draft, Conceptualization, istruc.2018.05.002.
[18] Hajdú G. Buckling resistance of steel beam-columns - the verification of Overall
Methodology, Software, Formal analysis, Investigation, Writing – review
Imperfection Method with numerical method (In Hungarian), dissertation; 2021.
& editing. Ferenc Papp: Visualization, Conceptualization, Methodol­ [19] Kindmann R, Kraus M. Steel structures design using FEM. Ernst & Sohn; 2011.
ogy, Software, Validation, Formal analysis, Investigation, Writing – Online ISBN: 9783433600771, doi: 10.1002/9783433600771.
[20] ABAQUS 3DEXPERINECE r2018x.
original draft, Writing – review & editing, Supervision, Project
[21] Maia É, Vila Real P, Lopes N, Couto C. The General Method for the fire design of
administration. slender I-section web-tapered columns. Thin-Walled Struct 2020;155:106920. ISSN
0263-8231. doi: 10.1016/j.tws.2020.106920.
Declaration of Competing Interest [22] BS EN 1090 2. Execution of steel structures and aluminium structures Part 2:
Technical requirements for steel structures. London, UK: BSI; 2008.
[23] ECCS European Convention for Constructional Steelwork. Manual on stability of
The authors declare that they have no known competing financial steel structures. 2nd ed. Technical Committee 8, Structural stability; 1976. www.
interests or personal relationships that could have appeared to influence eccspublications.eu.
[24] Pauli J, Somaini D, Knobloch M, Fontana M. Experiments on steel columns under
the work reported in this paper. fire conditions. Zurich, Switzerland: ETH Zurich, Institute of Structural
Engineering. IBK test report No. 340; 2012. doi: 10.3929/ethz-a-007600651.

16
S. Nemer et al. Engineering Structures 283 (2023) 115884

[25] Prachar M, Hricak J, Jandera M, Wald F, Zhao B. Experiments of class 4 open [27] Vila Real PMM, Lopes N, Simões Da Silva L, Franssen J-M. Parametric analysis of
section beams at elevated temperature. Thin-Walled Struct 2016;98:2–18. ISSN the Lateral-torsional buckling resistance of Steel beams in case of fire. Fire Saf J
0263-8231. doi: 10.1016/j.tws.2015.04.025. 2007;42(6–7):416–24. https://doi.org/10.1016/j.firesaf.2006.11.010. ISSN 0379-
[26] Vila Real PMM, Lopes N, Simões da Silva L, Piloto P, Franssen J-M. Numerical 7112.
modelling of steel beam-columns in case of fire - comparisons with Eurocode 3. Fire
Saf J 2004;39(1). doi: 10.1016/j.firesaf.2003.07.002.

17

You might also like