Journal of Constructional Steel Research 57 (2001) 1321–1336

www.elsevier.com/locate/jcsr

An efficient and direct method for buckling

analysis of steel frame structures
a,* b
Eduardo Bayo , Alfonso Loureiro
a
Department of Structural Engineering, University of Navarra, 31080 Pamplona, Spain
b
Escuela Politécnica Superior, University of La Coruña, 15403 Ferrol, Spain

Received 12 October 2000; revised 21 May 2001; accepted 17 August 2001

Abstract

The buckling analysis of steel frame structures is customarily carried out in a two step
approach. Firstly, an analysis of the structure, generally linear–elastic, is performed to obtain
the internal forces and moments; and secondly, the buckling analysis and design is done for
each individual element taking into account an initial imperfection. Steel design codes, and
in particular Eurocode-3 (EC3), base the buckling analysis upon the element effective length
factor K. This factor depends on the buckling shape of that particular element within the
structure, and in practical cases it is obtained by means of approximations or even by mere
estimation. This paper presents a direct one-step method for the buckling analysis of steel
frame structures. The method avoids the use of the effective length factor of each individual
element. The key point is to perform a non-linear analysis of the structure starting from an
initial deformation state that includes the initial imperfections of the elements. Such initial
deformation state is obtained from the first buckling mode of the structure by a suitable scaling
procedure based on an energy approach. The resulting initial deformed shape is introduced
along with the external loads in a fully non-linear structural analysis that yields the resulting
element stresses including the buckling effects. The performance of the method is checked
against the procedures established in EC3 by means of a series of numerical examples.
 2001 Published by Elsevier Science Ltd.

Keywords: Buckling; Steel design codes; Linear–elastic

* Corresponding author. Tel.: +34-948-425600; fax: +34-948-425629.

E-mail address: ebeyo@unav.es (E. Bayo).

0143-974X/01/$ - see front matter  2001 Published by Elsevier Science Ltd.

PII: S 0 1 4 3 - 9 7 4 X ( 0 1 ) 0 0 0 4 0 - 2
1322 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336

1. Introduction

The buckling analysis and design of elements under compression is essentially a

non-linear problem, in which second order effects (participation of axial force in
bending deformation) and initial imperfections need to be considered. The correct
solution of the problem requires a non-linear analysis of the structure under the
combined effect of the external loads and an initial imperfection in the compressed
elements. The non-linear analysis takes into account the effects of the equilibrium
in the deformed configuration, and the decrease of flexural stiffness due to the com-
pressive axial force.
The lack of non-linear analysis tools in the past (unavailable to the engineering
profession until a few years ago) has exerted a strong influence in the way the buck-
ling analysis is currently performed. The buckling problem, non-linear in essence,
is currently dealt with by the majority of the steel design codes, including EC3 [1],
by means of a two-stage approach:

1. In the first step, an analysis (usually linear–elastic) of the structure subject to the
specified external loads is carried out, and the values and distribution of element
internal forces are obtained.
2. In the second step, the buckling design as such, is performed using the effective
length factor, K, for each element as the basis to compute the buckling coefficient,
which in the case of the EC3 is called χ and is properly tabulated in clause 5.5.1.2.

The effective length, l=KL, (L is the total length of the element) is the length
between points of inflection of the buckling shape, and is directly related to the
element Euler critical buckling load. The effective length, and consequently the effec-
tive length factor, depend upon the physical characteristics of the element, the distri-
bution of internal forces and moments, and most importantly, on the boundary con-
ditions that the rest of the structure exerts on the element. In the case of a double-
hinged end column of constant section and axial force, K=1; for a cantilever column
K=2; for a hinged and clamped column K=0.5, etc. The reader is referred to classic
textbooks in mechanics of solids [2], for other support conditions. The double-hinged
column of constant section has become a reference point in order to define the buck-
ling behaviour and critical load of other types of elements.
In sway and non-sway steel frames, the K factor is obtained from the joint stiffness
coefficients at the ends of the columns (EC3 [1] Annex E). A rather complicated
and tedious analysis is required to assess the exact degree of rigidity by which the
element is attached to the rest of the structure. Consequently, in most of the codes,
including EC3, the following simplifications are assumed [3]:

1. In order to asses the joint stiffness, only the influence of adjacent columns and
beams is considered.
2. The adjacent columns and beams are assumed to have their non-adjacent ends
clamped.
 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336 1323

3. Both, the upper and lower columns have the same cross section and axial force
as the given column.

Using these simplifications, EC3 supplies the following expressions to compute

K in the case of non-sway frames, and sway frames, respectively:
1 ⫹ 0.145(h1 ⫹ h2)⫺0.265h1h2
K⫽ (1)
2⫺0.364(h1 ⫹ h2)⫺0.247h1h2
1 ⫹ 0.2(h1 ⫹ h2)⫺0.12h1h2
K⫽ 冪 1⫺0.8(h 1 ⫹ h2) ⫹ 0.6h1h2
(2)

where h1 and h2 are the relative joint stiffness defined in Annex E of EC3.
It is well known that structural elements are affected by residual stresses and
geometric imperfections (absence of straightness and homogeneity) produced during
the manufacturing process. These effects have to be accounted for in the buckling
analysis [3]. Dutheil proposed to model these effects by means of an initial imperfec-
tion curve of sinusoidal shape [4]. Such model has been corroborated by experimental
results. When the initial imperfection is introduced in a structural model that includes
second order theory, the coupling between the axial force and bending moment is
automatically accounted for. The solution of the resulting differential equation leads
to the secant formula that provides the resultant buckling stress in the element [5].
According to the model established in EC3, as well as other codes, the displace-
ment due to an initial imperfection in the middle of an element is:
kW
fm ⫽ (3)
A
where W is the resistant module, A the area of the cross section, and k is a weighted
parameter equal to: k ⫽ k0 ⫹ k1l̄ ⫹ k2l̄2, respectively. The imperfection coefficients
k0, k1 and k2 vary for different national codes and standards. The non-dimensional
slenderness l̄ corresponds in this case to the pinned ended bar model.
Table 1 shows the values of imperfection coefficients that are adopted in EC3,
depending on the type of section and corresponding buckling curve.
The way that EC3 accounts for the effects of imperfections is by means of the
parameter l. This parameter reduces the buckling resistance of the element, and is
tabulated as a function of the element slenderness l (that depends on K), and the

Table 1
Imperfection coefficients according to EC3

Curve a Curve b Curve c Curve d

k0 -0.0323 -0.0578 -0.0872 -0.1398

k1 0.1424 0.2686 0.4145 0.6749
k2 0.0966 0.1020 0.1078 0.1216
1324 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336

class of section. It may be seen that the definition of the effective length factor K
is of the outmost importance.
In addition, EC3, adopting recent advances in structural analysis, allows for a
series of possibilities at the time of performing the buckling analysis. The proposed
alternatives are the following:

1. Second order global analysis of the structure followed by member design based
on in-plane buckling lengths corresponding to the non-sway mode.
2. First order global analysis with sway moment amplification and member design
based on non-sway mode buckling lengths.
3. First order global analysis with sway moment at beam-to-column connections
amplified by a factor of 1.2, and member design based on sway mode buckling
lengths.

The flow chart, depicted in Figure 1, illustrates the alternative paths that EC3
permits depending on the type of analysis that the analyst may choose. EC3 even
offers the possibility of performing the buckling analysis by directly including the
initial imperfection in a second order analysis, thus, by-passing the use of buckling

Figure 1. Different types of global analyses contemplated in EC3.

 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336 1325

formula and tables. The initial imperfection may be introduced in either the global
analysis, or in an individual member analysis once the internal element forces have
been obtained from a global analysis. In either case one needs to assume a buckling
mode, therefore a K factor.

2. Description of the proposed method

2.1. Posing the problem

It has been pointed out above in relation to the classic buckling analysis, the
problems concerning the finding of the effective length factor K, and the simplified
hypotheses that are needed to approximate its value in framed structures. On the
other hand, if one chooses the more direct second order analysis and initial imperfec-
tion method offered by modern codes, such as EC3, one still encounters the problem
of how to define the element imperfection in terms of its amplitude and shape.
Such imperfection is specified in EC3 for the only case of a double-hinged mem-
ber. However, for the rest of cases (practically all) one has to assume a factor K in
order to relate the initial deformation state to the double-hinged member, which is
considered as the master case. Therefore one enters into a vicious circle: a more
sophisticated second order analysis is pursued to avoid the calculation of K and
corresponding tables, and yet the K factor is still necessary in order to define the
initial imperfection state that the method requires.
When including the initial imperfection proposed by EC3 in a second order analy-
sis, one may find relatively satisfactory results for the case of statically determinate
structures (see Table 2). However, as the structure becomes statically indeterminate
the results start deviating towards the unsafe side, getting progressively worse as the
degree of indeterminacy increases. The only way to solve the problem correctly is
by knowing exactly the buckling mode of the particular member within the structure,
hence, the K factor, whose calculation one is precisely trying to avoid.
During this research study, another way on introducing the initial imperfection
was studied based upon the use of equivalent distributed loads q. Such loads were
obtained under the condition that they would produce the same bending moment as

Table 2
Stresses obtained by direct second order analysis of different columns

Stress in MPa obtained with buckling load specified in EC3

sy/sm1=250 MPa With initial geometric With equivalent distributed load

imperfection
Double-Hinged Column 250 248
Fixed-Free Column 251 229
Fixed-Hinged Column 245 255
One Storey Sway Frame 231 275
1326 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336

that generated by the axial force and maximum deflection of the imperfection. The
result is:
8fmP
q⫽ (4)
L2
where fm is the amplitude of the initial imperfection, L the length of the element and
P the axial force acting on it.
Table 2 shows the results obtained using second order analysis, in simple structures
of steel S275, that were designed considering the axial force in the columns that
produces its buckling failure according to EC3. The coefficient gm1 was taken equal
to 1.1
Table 2 also illustrates the maximum stress acting on the element for the different
structural systems under the loading that produces its failure. One may see, that when
the initial imperfection proposed by EC3 is included in the direct second order analy-
sis, satisfactory results are obtained for statically determinate structures, however as
the degree of indeterminacy increases the error grows towards the unsafe side. The
second alternative: use of the distributed loads specified in equation (4) leads to
unacceptable results.
Another problem arises at the time of trying to decide which way should the initial
imperfection be introduced when combined with other external actions or loading.
The overall behaviour of the structure, as well as that of a particular member, will
depend on the direction assumed for the initial imperfection.
After all these considerations and previous analysis, one reaches the conclusion
that the method proposed in EC3 is not adequate under the practical point of view.
In what follows a new direct method for buckling analysis is proposed that will
allow us to:

1. Perform a direct one-step second-order analysis of the structural system that will
simultaneously yield the buckling analysis of elements under compression.
2. Eliminate the use of the effective length factor K, thus avoiding its simplifications
and possible inaccuracies.
3. Avoid the use of equivalent forces to model imperfections, which may lead to
unacceptable results.
4. Introduce element imperfections in a direct manner accounting for the most
unfavourable combination with the rest of the actions exerted on the structure.

2.2. Global non-linear analysis with element initial imperfections

The proposed new method is based on including the initial element deformations
as part of a second-order global analysis. In this section the equations leading to the
non-linear second-order analysis with initial imperfections are formulated:
We start from the elementary deformations that result from considering the Green-
Lagrange tensor [6] for the case of an element with axial and bending deformation:
1
e(x) ⫽ u⬘⫺y(v⬙⫺v⬙0) ⫹ (v⬘)2 (5)
2
 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336 1327

where, x is the spatial variable along the longitudinal axis of the element, y is the
flexural axis going through the centroid of the transversal section. The variable u
indicates displacements due to axial deformation, and v displacements due to bend-
ing. The corresponding stress is s=Ee, where E is the Young’s modulus. Using the
virtual displacements method, the first variation of the deformation tensor is calcu-
lated:
de ⫽ du⬘⫺ydv⬙ ⫹ v⬘dv⬘ (6)
and the corresponding virtual work for an element is:

VW ⫽ 冕
Vol
(de)TsdVol (7)

Introducing equations (5) and (6) into (7), and eliminating high order derivatives
and linear terms in y, whose integrals are null, the following expression is obtained:

VW ⫽ 冕
Vol
(u⬘du⬘ ⫹ v⬘u⬘dv⬘ ⫹ y2v⬙dv⬙⫺y2v⬙0dv⬙)dVol (8)

Next, a finite element discretization yields the following interpolation for the dis-
placement field:
u ⫽ N(x)û (9)
and
v ⫽ H(x)v̂
where, N and H are vectors containing the finite element shape functions [7], and
û and v̂ are the displacements and rotations at the joints. Introducing equation (9)
into (8), one obtains the expressions for the elastic, KE, and geometric, KG stiff-
ness [7]:

冕
L

KE ⫽ (EAN⬘TN⬘ ⫹ EIH⬙TH⬙)dx (10)

0

冕
L

KG ⫽ (PH⬘TH⬘)dx
0

The shape functions H used in this work are composed of cubic Hermite polynomials
[8]. Sufficient accuracy is obtained with only one finite element per column as dem-
onstrated below.
1328 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336

Using standard assembly procedures in finite element analysis [7], the following
global matrix equation is formed:
[KE ⫹ KG]ŵ ⫽ F ⫹ KEŵ0 (11)
where, the vector ŵ contains all displacements and rotations of the joints of the
structure. The vector ŵ0 includes all displacements and rotations of the initial element
imperfections, whose values will be defined in the next section, and the vector F
contains the external forces and other actions applied to the structure.
Equation (11) serves as the basis to perform a full non-linear analysis, considering
equilibrium in the deformed configuration. By a proper definition of the initial
element imperfections, the stresses resulting from this global analysis will account
for the buckling effects without the need for additional checks. The method to define
and include these imperfections will be defined in the next section.

2.3. Formulating a method to include initial element deformations

In this section a new method for characterising and introducing imperfections of

the elements in the global analysis of the structure is shown. The method eliminates
the uncertainties originated by the guessing of the buckling shape (K factor) as well
as by equivalent external loads. The new method is based on an energy approach,
in the sense that the initial imperfections are introduced into the structural model by
means of their strain energy.
The initial curvature defined in EC3 (and other codes) is defined for the double-
hinged element in terms of a sinusoid of amplitude fm. The value of this amplitude
has been defined in equation (3). The corresponding element strain energy is easily
obtained through the solution of the integral:

冕
L
1
⌸ ⫽ EI(v⬙0(x))2dx (12)
2
0

where, E is the modulus of elasticity, I the moment of inertia, and v0(x) the initial
imperfection equal to v0(x)= fm sin(px/L). The solution of the integral in equation
(12) yields:
p4EIfm2
⌸⫽ (13)
4L3

Once the element strain energy due to the imperfection is determined, the next
step is to obtain the fundamental buckling mode of the structure under the actuating
loads. To this end an analysis of the structure with the current loads is performed,
thus obtaining the distribution of axial forces in all the elements under compression.
Then KG is formed and the following eigenvalue problem is solved:
[KE⫺lKG]⌽ ⫽ 0 (14)
 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336 1329

where, KE and KG are the structure elastic and geometrical stiffness matrices, respect-
ively.
The fundamental buckling mode f0 will serve as the basis to introduce the initial
imperfection in the elements in the global analysis of the structure along with the
rest of the actions. It is important to emphasise that this deformation state includes
the buckling mode of all the elements in the structural model without the necessity
of having to introduce simplified or approximated hypothesis as those required for
calculating the K factor. All the local and global stiffness characteristics are included
in the solution of equation (14) to the degree of accuracy defined by the element
discretization in the structural model.
However, the buckling mode needs to be scaled so that it will really portray the
initial deformation characteristics specified by the corresponding steel code (EC3 in
this case). For this purpose, this buckling mode is scaled by a factor such that its
strain energy will be equal to the total strain energy due to the initial imperfections
of elements under compression.
The scale factor thus becomes:

冘
n

冪
⌸i
i⫽1
s⫽ (15)
1 T
f K f
2 0 E 0
where, the summation extends to all the elements under compression. The numerator
of this equation encompasses the strain energy due to the initial imperfections of all
the elements under compression (as defined in equation (13)). The denominator
defines the strain energy of the buckling mode f0. It is important to notice that such
mode has been calculated taking into account the distribution of axial forces in the
frame by means of the stiffness matrix KG included in equation (14).
The factor s is used to scale the buckling mode f0 and, afterwards it is introduced
in a global non-linear analysis using equation (11). The result of this analysis will
yield the total stresses, including the buckling effects. The element effective length,
the factor K, and the buckling coefficient c do not need to be considered at all during
the analysis process.
The following algorithm summarises the sequence of analysis:

2.3.1. Algorithm

1. Perform a linear analysis of the structure for the given set of loads to obtain the
force distribution in all the elements.

冘
n

2. Compute the strain energy ⌸i for all the elements under compression.
i⫽1
3. Form KG and solve the eigenvalue problem [KE⫺lKG]⌽ ⫽ 0 to obtain the funda-
mental buckling mode f0.
1330 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336

4. Obtain the scale factor s (equation (15)) and scale the buckling mode f0.
5. Perform a global non-linear analysis including the initial global imperfection,
using the equation: [KE ⫹ KG]ŵ ⫽ F ⫹ KEf0

3. Examples

A series of different structural systems have been analysed using the method
explained above. In all cases the structures are made of steel S275, and were designed
considering the axial force in the columns that produces its buckling failure according
to EC3. The coefficient gm1 was taken equal to 1.1

3.1. Analysis of single columns

First, single columns were analysed under different boundary conditions (see Fig-
ure 2). In all cases the columns are 4 m long with section HEB 200 of steel S275.
Buckling in the major axis is allowed, and is restrained in the minor axis. Table 3
illustrates the maximum stress acting on the element under buckling failure for the
different boundary conditions.

3.1.1. Double-hinged column. (Figure 2(a))

The axial load that produces the failure of this column according to EC3 is 1687
kN, which makes the element reach a maximum stress of 250 Mpa. When this load

Figure 2. Single columns.

 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336 1331

Table 3
Stresses obtained using the proposed method (exact solution 250 MPa)

Largest stress in MPa

sy /γm1 = 250 MPa 1 element per column 2 elements per column

Double-Hinged Column 248.4 250

Fixed-Free Column 249.3 250
Fixed-Hinged Column 248 250
Column with springs 248.6 250

is applied to the column and an analysis is performed using the new method
explained above, a stress of 248.4 Mpa is reached with only one finite element per
column, and 250 Mpa (the exact value) with two elements per column.

3.1.2. Fixed-free column. (Figure 2(b))

The axial load that produces the failure of the fixed-free column according to EC3
is 1056 kN (maximum stress of 250 Mpa). The proposed method yields 249.3 Mpa
with one element, and 250 Mpa (the exact value) with 2 elements.

3.1.3. Fixed-hinged column. (Figure 2(c))

The axial load that produces the failure of the fixed-hinged column according to
EC3 is 1823 kN (maximum stress of 250 Mpa). The proposed method yields 248
Mpa with one element, and 250 Mpa (the exact value) with 2 elements.

3.1.4. Column with rotational springs at the ends. (Figure 2(d))

This example models the behaviour of a column that has its ends partially
restrained in rotation, thus simulating more closely its behaviour within a structural
frame. The stiffness of the springs is equivalent to that of a HEB 200 with 4 m of
length. Therefore, the spring stiffness Km= 4EI/L, where E and I are elasticity modu-
lus of steel and second moment of area of the HEB 200, respectively, and L=4 m.
The buckling load is in this case 1883 kN, and highest stresses obtained with the
proposed method are 248.6 MPa and 250 MPa with 1 and 2 finite elements, respect-
ively.

3.2. Analysis of frames

Out of the three methods that EC3 considers for the frame analysis, that were
explained above, we consider the first one: second order global analysis of the struc-
ture followed by member buckling analysis, based on in-plane buckling lengths cor-
responding to the non-sway mode. This method is the most exact and reliable of the
three, and the one recommended in EC3. This method and the one proposed in this
paper are used to analyse a series of different frames.
1332 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336

Figure 3. Frame 1.

The first frame is shown in Figure 3, and consists of a 2-storey and 2-bay sway
frame with identical columns in both storeys. The second, shown in Figure 4, has
different columns in the lower and upper storey. The other one, shown in Figure 5,
has a non-uniform loading pattern. All beams and columns are modelled with only
one finite element of cubic shape functions. The frame imperfections are modelled
by means of equivalent lateral forces as specified by EC3, and their values are illus-

Figure 4. Frame 2.
 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336 1333

Figure 5. Frame 3.

trated in the corresponding figures. The values of the ratio Vsd/Vcr for the frames are
respectively: 0.03, 0.02 and 0.23. According to EC3 (clause 5.2.5.2(3)) the first two
frames are classified as non-sway frames, and the third one as sway frame. These
laterally and vertically loaded frames have been analysed applying the EC3 method
and the proposed method. The results are summarised in Table 4.
First the EC3 method is applied and after a second order analysis, buckling and
bending resistance are checked according to clauses 5.5.4 (1) and 5.4.8.1 (12),
respectively. The corresponding comparative stresses at the critical cross section are
shown in Table 4. The bending resistance is found to be more restrictive in these
cases. Next, the proposed energy method is applied using the non-sway buckling
mode as the initial imperfection state. Notice that frame imperfections are already
accounted for by means of the equivalent lateral forces. The results are shown in
Table 4 and reveal the fact that the proposed method encompasses the buckling as
well as the resistance checks, thus becoming an envelope to both effects.

Table 4
Stresses obtained with EC3 and the proposed method

s (Mpa) Buckling Resistance Proposed Energy

Method
EC3 5.5.4(1). EC3 5.4.8.1 (12)

Frame 1 226 230 231

Frame 2 225 230 230
Frame 3 187 189 189
1334 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336

A more slender 5 storey frame, shown in Figure 6, is also analysed by both

methods. In this case Vsd/Vcr = 0.35, which according to EC3 classifies as a sway
frame. In addition, the value of l̄ (clause 5.2.4.2 (4)) for the central column of the
first storey is 0.67, which is larger than 0.50 and therefore EC3 specifies that its
initial imperfection be introduced in the global analysis. This imperfection is intro-
duced physically by modelling the column in its deformed state using 2 and 4
elements, respectively. The results of the analyses are shown in Table 5. It is
important to notice, using the EC3, the significant dispersions of the resulting stresses
that are obtained when the imperfection is placed on the left or on the right of the
column. Therefore, when using EC3 one has to perform different analyses to find

Figure 6. Frame 4.
 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336 1335

Table 5
Stresses (Mpa) obtained with EC3 and the proposed method for Frame 4

EC3 EC3 Proposed Energy

2 element / column 4 element / column Method

Imp. Left Imp. Right Imp. Left Imp. Right Non-sway buckling
mode
233 255 231 258 250

the worst possible direction for the imperfection. Less significant is the influence of
the number of elements used to model the initial imperfection
On the other hand, the proposed approach automatically chooses the most
unfavourable direction of the imperfection and yields a unique response. Only 1
element per column is used for this approach. The difference in stresses between
the EC3 (258 Mpa) and the proposed method (250Mpa) lies on the fact that the
latter takes into account the actual buckling length whereas the EC3 considers the
pinned ended bar model (EC3 Fig. 5.5.1).

4. Conclusions

In this paper a direct one-step method for the buckling analysis of steel frame
structures has been presented. The method is based on a non-linear analysis of the
structure starting from an initial deformation state that includes the initial imperfec-
tions of the elements. Such initial deformation state is obtained from the first buckling
mode of the structure by a suitable scaling procedure that takes into account all the
energy of the individual elements under compression. The resulting initial deformed
shape is introduced along with the external loads in a fully non-linear structural
analysis that yields the resulting element stresses including the buckling effects. The
performance of the method has been checked against the procedures established in
EC3, by means of a series of numerical examples.
The following conclusions may be highlighted:
(a) Contrary to the current practice established by steel design codes, including
EC3, the proposed method avoids the use of the effective length factor K, and buck-
ling coefficients χ of each individual element, thus avoiding simplifications and/or
possible inaccuracies.
(b) In addition the proposed method introduces the element imperfections in a
direct manner accounting for the most unfavourable combination (left or right) with
the rest of the actions exerted on the structure, and providing an envelope to the
different resistance checks.
(c) The results obtained for the different examples show that the method is efficient
and accurate.
(d) The method could also be extended to other codes, not just EC3, by considering
the energy associated to the initial deformations established in such codes.
1336 E. Bayo, A. Loureiro / Journal of Constructional Steel Research 57 (2001) 1321–1336

(e) This method can be very useful in case of frames composed of elements with
variable cross section and non-uniform distribution of axial forces.
(f) The solution of the cases presented above lead to the conclusion that only one
finite element (with cubic shape functions) per member is accurate enough. However,
two elements lead to more precise results.
(g) This method is not limited to planar frames, and it may be extended to three-
dimensional structures. This will be part of a future work.

Acknowledgements

The support of this research provided by the CICYT of the Spanish Ministry of
Education under grant TAP98-0377, and the University of Navarra under grant
PIUNA-18098412 is greatly acknowledged.

References

[1] Eurocode No. 3: Design of Steel Structures. Part 1 — General Rules and Rules for Buildings Eurocode
3 Editorial Group. Commission of the European Communities. (1990).
[2] Timoshenko S, Gere J. Theory of elastic stability, 2nd Ed. McGraw-Hill, 1961.
[3] Galambos TV. Guide to Stability Design Criteria for Metal Structures, 5th Ed. John Wiley, 1998.
[4] Marco J. Fundamentos para el Cálculo y Diseño de Estructuras Metálicas de Acero Laminado.
McGraw-Hill, 1997.
[5] Gere J, Timoshenko S. Mechanics of Solids, 4th Ed. PWS-Kent, 1997.
[6] Malvern L. Introduction to the mechanics of a continuous medium. Prentice Hall, 1999.
[7] Cook R, Malkus D, Plesha M. Concepts and Applications of Finite Element Analysis. Wiley, 1989.
[8] Hughes TJ. The Finite Element Method. Prentice-Hall, 1987.