You are on page 1of 14

Journal of Constructional Steel Research 62 (2006) 757–770

www.elsevier.com/locate/jcsr

Interaction formulae for members subjected to bending and axial


compression in EUROCODE 3—the Method 2 approach
R. Greiner a,∗ , J. Lindner b
a TU Graz, Institute for Steel Structures and Shell Structures, Lessingstr. 25, 8010 Graz, Austria
b Engineering office, Bismarckallee 4, 14193 Berlin, Germany

Received 9 August 2005; accepted 30 November 2005

Abstract

The final version of EN1993-1-1, EUROCODE 3 [EN1993-1-1. Eurocode 3. Design of steel structures, general rules and rules for buildings.
2005] for Steel Structures provides two alternatives for the buckling check of members subjected to axial compression and bending by interaction
formulae, which are called there Method 1 and Method 2. This paper presents the characteristics, the background and the use of Method 2. The
analogous presentation of Method 1 has already been given in [Boissonnade N, Jaspart J-P, Muzeau J-P, Villette M. New Interaction formulae for
beam-columns in Eurocode 3. The French-Belgian approach. Journal of Constructional Steel Research 2004;60;421–31].
The Method 2 formulae have been derived on the basis of the general format of the interaction concept of existing codes, e.g. the ENV-rules;
however with advanced numerical background and consistent classification of the buckling modes. In this respect new improved interaction factors
were developed from a wide scope of numerical simulations and the concept of the formulae was focussed distinctly on describing the modes
of in-plane and out-of-plane buckling for members susceptible to fail either in flexural buckling or in lateral–torsional buckling. As result two
sets of formulae are provided, which each cover a clear scope of physical member behaviour. Hereby, the specific effects of intermediate lateral
restraints—as often found in steel structures—have also been included.
The Method 2 formulae aim at providing buckling rules with compact simplified interaction factors and transparent application for standard
cases.
c 2006 Elsevier Ltd. All rights reserved.

Keywords: Interaction formulae; Buckling of beam–columns; EUROCODE 3

1. Introduction this basis the interaction factors could be directly recalculated


from the numerical limit load results and the effect of different
In the process of the conversion of the ENV-version of parameters be consistently identified. The final definition of the
the EUROCODE 3 [2] to the final EN-version [1] basic new rules was then based on calibration with the existing test
developments were carried out by Technical Committee 8 results and the numerical simulations.
(TC8) of ECCS, which resulted in new buckling rules for The elaboration of the Method 2-rules therefore presents
members subjected to axial compression and bending. This was the consequent development of the former work on interaction
caused by some physical inconsistencies and overconservatism formulae by use of the opportunities of the new computer
found in the ENV-rules [4], which could lead to considerable techniques. At present they cover double-symmetric cross-
underestimation of the capacity of beam–columns in certain sections in EC3-1-1 [1], however an extension to mono-
cases. The use of modern computer techniques and FE- symmetric sections has already been made in the meantime [6].
programs opened the possibility of simulating the geometrically In this spirit of further development the traditional interaction-
and materially nonlinear buckling behaviour of imperfect
formulae were maintained in format and notation as far as
members in much wider parametric scope than before [4,5]. On
possible in order to facilitate the understanding of the changes
and the practical application for the user. However, in a number
∗ Corresponding author. Tel.: +43 316 873 6200; fax: +43 316 873 6707. of points new formulations were necessary, in particular to
E-mail address: r.greiner@tugraz.at (R. Greiner). provide possible further developments.

0143-974X/$ - see front matter 


c 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2005.11.018
758 R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770

3. Numerical calculations and statistical evaluation

The basis of the new developments of the buck-


ling interaction formulae was the numerical simulation
of the elastic–plastic buckling behaviour of single-span
beam–columns, which accounts for geometric bow imperfec-
tions, residual stresses and a linear-elastic–plastic material law
without hardening effects. The calculations, called geometri-
cally, materially nonlinear analyses of the imperfect structure
[GMNIA], were performed by a computer module for auto-
matic parameter variation and data evaluation, which allowed
us to cover a wide scope of member and load parameters (about
25000 cases) [7]. The analyses were carried out by the FE-
Programme ABAQUS [8] using beam elements. Verifications
were made by comparison of the numerical simulation with
buckling tests and other computer programs [9].
Fig. 1. Characteristics of Method 2 of EC3-1-1. The resulting limit load data were used to recalculate the
interaction factors on the basis of the proposed interaction
equations and afterwards to develop simplified formulae for
The objective of this paper is to present the background of them. These formulae were calibrated with the low number of
the buckling formulae of Method 2 in EC3-1-1, which are in the available buckling tests and with the high number of numerical
tradition of the simplified design rules of many existing codes. results using statistical evaluation [10–12].
A short but not detailed enough explanation to this background
was already given in [16]. 4. Concept of interaction formulae for N + My
EC3-1-1 will provide further methods for the stability check
of members (Method 1 and General Method), which are more As already explained in chapter 2 two sets of buckling
complex and are not dealt with in this paper. formulae are provided, which each describe the buckling
behaviour of either torsionally stiff or torsionally flexible steel
2. Characteristics of the buckling rules of Method 2 of sections (Fig. 2).
EC3-1-1 In the first case they concern flexural buckling, in the second
one lateral–torsional buckling. The first formula of each set
The buckling rules of Method 2 are directed to cover the is related to buckling about the y-axis and the second one to
following three main characteristics of the structural behaviour buckling about the z-axis.
of steel members (Fig. 1): This clear structure may assist the designer to connect the
results of the buckling check with structural provisions. In this
• Cross-section shapes: Depending on the shape of the respect it has already been said, that the buckling formulae
cross-section members subjected to axial compression and are appropriate for both free, single-span members as well
bending may behave susceptible to torsional deformations as those supported laterally by intermediate restraints. In the
or not. Therefore, specific sets of buckling formulae are present state such restraints are expected to be full restraints,
provided for torsionally stiff members (hollow sections or which means that in the case of flexural buckling a pure lateral
torsionally restrained I-sections) and torsionally flexible restraint would be sufficient, while the case of LT-buckling
members (I and H-sections). would require lateral plus torsional restraint. Different kinds of
• Buckling modes: In general spatial buckling deformation restraints are under investigation.
of members subjected to axial compression and bending is The four interaction formulae are given in Fig. 3 for class 1
traditionally split up into the two buckling modes about the and 2 sections (analogous formulae for class 3 and 4 sections
y-axis and about the z-axis. In each set of formulae the first also exist in EC3-1-1). In order to accentuate the systematic
one describes the strong-axis mode and the second one the structure of the two sets of design formulae for practical use the
weak-axis mode. In this way the specific physical behaviour four formulae were numbered consecutively by (1) to (4) and
is transparently connected with the design process (lateral marked by boxes.
restraining etc.). Eqs. (1) and (2) describe flexural buckling of torsionally
• Intermediate lateral restraints: The buckling formulae were stiff members and Eqs. (3) and (4) LT-buckling for torsionally
primarily derived for the free beam–column with end-fork- flexible members. In the case of flexural buckling about the
conditions. However, at the same time the application to weak axis a simplified approach (2a) is allowed, which is based
the practically frequent cases of members with intermediate just on the axial compression term alone This traditional rule—
lateral restraints has also been provided. This requires us used in many existing codes—has been maintained in EC3 for
to differentiate between the buckling lengths and moment uniaxial bending and compression.
diagrams along the span and along a segmental part between The new interaction formulae of Method 2 use the following
the lateral restraints. factors:
R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770 759

Fig. 2. Concept of interaction formulae of Method 2.

Fig. 3. Interaction formulae of Method 2, EC3-1-1.

• Interaction factors k y , k z , kLT developed for the EN-version. The latter concerns mainly the
• Equivalent uniform moment factors Cmy , Cmz , CmLT . classification of the LT-buckling curves connected with the
torsional capacity of the cross-section (h/b ≥ or < 2) and
In addition the buckling reduction factors χ y , χz , χLT for the
the inclusion of the beneficial effect of non-uniform moment
basic buckling modes under pure compression or pure bending
diagrams by the modified coefficient χLT,mod .
are necessary. They cover the cases of separate flexural buckling
It has to be pointed out, that the stability check by the
caused by pure compression or LT-buckling by pure bending.
interaction formulae (1) to (4) has to be supplemented by
In the following the derivation of these terms is presented.
the cross-section check at the member ends, if the bending
The interaction factors k y and k z apply to the uniform
moments are non-uniform along the span and the moment
distribution of the bending moments M y and Mz , while the kLT -
maximum appears at the member end. This means, that the
factors include the non-uniformity of the moment diagrams of
cross-section capacity at the end section may govern, if the
M y . This non-uniformity is accounted for by Cm -factors, which
stability effect in the span is low.
are based on the widely used concept of Austin [13], i.e. to take
the constant moment (not the sinusoidal moment distribution) 5. Members not susceptible to torsional deformations—
as reference, so that Cm = 1.0 holds for uniform moments and Class 1 and 2-sections
any other moment diagram leads to Cm -values lower than 1.0.
In this respect this definition of Cm is different from the β M - 5.1. General buckling behaviour
factors of the ENV-version.
It may also be noted, that the coefficients χ y and χz If torsionally stiff members are present the buckling
(European column buckling curves) are the same as in the behaviour leads to flexural buckling. This is illustrated by the
ENV-version, but that the coefficients χLT have been further example of an RHS 200/100/10 of a beam–column of 4 m
760 R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770

Fig. 4. Flexural buckling of torsionally stiff members.

length under axial compression and uniform bending moment


M y in the form of an interaction diagram (Fig. 4). In the case of Fig. 5. Design diagrams for in-plane buckling under M y .
sufficient intermediate lateral restraint in-plane buckling is the
governing buckling mode, which is described by the interaction The form of the k y -curves illustrates the physical behaviour
curve y–y and is approximated by Eq. (1). For an unrestrained of members with plastic cross-section:
member, which is free to deflect spatially along its span, out- At λ̄ y = 0 they express the plastic cross-section interaction
of-plane buckling will occur if N/N pl is larger than 0.4. This of N + M y .
is described by the branch z–z of the interaction curve and is
With increasing λ̄ y the second-order effect causes the growth
approximated by Eq. (2). The approximation by Eq. (2a) is also
of the bending moment and by that the rise of the k y -factor.
presented. The key-parameter of the interaction behaviour is
Approximately at λ̄ y ∼ = 1.0 the point is reached, where
the interaction factor k, whose development is explained in the
a further increase of k y is not needed, since further stability
following for the different buckling modes.
effects are covered by the χ y in the term of axial compression.
As already said before, the interaction factors are the
For λ̄ y > 1.0 the k y -values remain at approximately constant
result of recalculations from a large scope of limit load-data
level, which may be explained by the fact, that the member
resulting from GMNIA-calculations. The parameters used for
this recalculation were taken from the general derivation of the in this slenderness range behaves increasingly elastically, and
buckling interaction formula on basis of the second order theory therefore, the effect of M pl,y used in Eq. (1) needs to be
(see e.g. [4]), which shows that the three parameters λ̄ y , χ y counter-compensated by k y (this effect is obviously more
pronounced in RHS, than in IPE-section, as indicated by the
and NNpl determine the interaction behaviour. The interaction
diagrams in Fig. 6).
factors were finally expressed by λ̄ y and n y , the latter of which The k y -curves have been transformed into a formula, which
connects NNpl and χ y as in the term for buckling under pure axial is given by Eq. (5). The bi-linear form of this formula, with
compression. a kink at λ̄ y = 1.0 (see Fig. 5) has been chosen with respect
to user-friendliness. This means that related to λ̄ y ≥ or < 1.0
5.2. In-plane buckling under N + M y just one part of Eq. (5) needs to be determined. It may also
be noted that the differences of the k y -curves in the range
The design buckling formulae are as follows (Fig. 5):
of λ̄ y > 1.0 are of just minor effect on the design, because
N Ed Cmy · M y,Ed the differences mainly occur at high values of n y , where the
+ ky ≤1 (1)
χ y · N pl,Rd M pl,y,Rd moment terms are accordingly small, so that k y has very little
  influence.
k y = 1 + λ̄ y − 0.2 · n y ≤ 1 + 0.8 · n y (5)
The k y -factors of the ENV-version are also given in Fig. 6.
N Ed
ny = (6) Compared with the new k y -factors of Fig. 5 the economical
χ y N pl,Rd improvement is obvious.
Cmy = 0.6 + 0.4ψ ≥ 0.4 (see Fig. 5 and also Table 1). (7) Results for examples shown as interaction diagrams for
The recalculations of the k y -factor from GMNIA-results different forms of moment distributions are illustrated in Fig. 7.
were made on basis of (Eq. (1) for different cross section The results of the new formulae are compared with the
shapes. Examples are shown in Fig. 6 for an IPE and an RHS. GMNIA-results and those of the ENV-version. The improved
They indicate a certain influence of the cross-section shape, approximation in the cases (a) and (b) is obvious, the
however it turned out to be of minor effect on the end result, differences in case (c) are due to the conservatism of the Austin-
so that it could be ignored as a design parameter further on. formula for bilinear moment diagrams (see chapter 5.6).
R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770 761

Fig. 6. Interaction factor k y .

Fig. 7. Comparison of in-plane flexural buckling.

5.3. Out-of-plane buckling under N + M y sections under uniaxial bending, since the range of unconser-
vatism (about 7%–9% maximum) has been considered tolerable
The design buckling formula is as follows: with respect to the safety concept. It should be noted, that just
N Ed Cmy · M y,Ed torsionally stiff sections are concerned here, since torsionally
+ 0.6 k y ≤1 (2) flexible I-sections are treated by LT-buckling.
χz · N pl,Rd M pl,y,Rd
N Ed
≤ 1. (2a) 5.4. In-plane buckling under N + Mz
χz · N pl,Rd
As discussed in par. Section 5.1 (Fig. 4) the spatial buckling The buckling behaviour of members in weak-axis bending is
tendency of a laterally free member may be described by very similar to that of par. 5.2. The k z -factors recalculated from
two branches of the interaction curve, which are related to GMNIA-results of IPE- and HEB-sections on the basis of the
the in-plane or to the out-of-plane buckling mode. Formula design Eq. (8) are given in Fig. 10:
(1) describes the first one and formula (2) the second one.
The interaction factor of Eq. (2) has been defined as 0.6 · k y , N Ed Cmz · Mz,Ed
+ kz ≤1 (8)
which means that 60% of the in-plane bending term affects χz · N pl,Rd M pl,z,Rd
 
the out-of-plane buckling. As Fig. 4 illustrates, this definition k z = 1 + 2λ̄z − 0.6 · n z ≤ 1 + 1.4 · n z . . . I-profiles
is a conservative linear approximation of the branch z–z, which (see Fig. 9) (9)
has been chosen firstly with respect to simple calculation and  
secondly because the factor 0.6 will be used later for describing k z = 1 + λ̄z − 0.2 · n z ≤ 1 + 0.8 · n z . . . RHS-profiles
the biaxial bending behaviour under M y and Mz . (analogous to Fig. 5) (10)
Fig. 8 presents a comparison of Eq. (2) with the GMNIA- N Ed
nz = (11)
results and those of the ENV-version. It shows the overconser- χz · N pl,Rd
vative results of the ENV-version due to the use of χmin (in- Cmz = 0.6 + 0.4ψ ≥ 0.4 (see Fig. 9). (12)
stead of differentiating between χ y and χz ). It further shows
the result of Eq. (2a), which is the widely used out-of-plane The k z -curves (see Fig. 10) show similar shapes as the k y -
buckling check of many existing codes. In contrast to Eq. (2) curves of Fig. 6 in principle, however—understandably—the
it may locally be a bit unconservative, however TC8 supported effect of the cross-section interaction at small λ̄z is much larger
keeping this traditional check for use with double-symmetric and the level of the curves at λ̄z > 1.0 is considerably higher.
762 R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770

Fig. 8. In- and out-of-plane flexural buckling.

The interaction formulae Eqs. (13) and (14) for biaxial


bending have been developed by extending the given formulae
(1) and (2) by the terms of Mz . For the buckling mode about
z–z the full k z -factor is applied and for the buckling mode y–y
the reduced value 0.6 · k z .
By this, the two equations show the following properties:
• their format is ‘symmetrical’ with respect to the axes y
and z; this is significant for describing the transition from
rectangular to square hollow sections consistently,
• they allow by use of the factors 0.6 a qualitatively good
approximation of the plastic cross-section interaction for
biaxial bending M y + Mz . For N → 0 two linear equations
appear, which approximate the convex interaction curve, see
Fig. 12, compared to the linear interaction used before.
The interaction formulae (13), (14) have been compared
graphically and statistically with the GMNIA-results, which
Fig. 9. Design diagrams for in-plane buckling under Mz . now lead to a three-dimensional graph. Fig. 13 illustrates the
results of the new formulae compared with the GMNIA-
Both may be explained by the much higher plastic reserve of results and those of the ENV-version. Although still an
I-sections in bending about the weak axis than about the strong approximation, the simplified formulae may lead to a
axis. The k z -factor for I- and H-sections needs, therefore, to be considerable improvement. In particular, the effect of the factor
defined separately (Fig. 9). 0.6 is an amendment for biaxial bending.
In contrast to the I-sections the k z -factor for RHS may
be kept as for buckling about the strong axis, which is 5.6. Equivalent uniform moment factor Cm
understandable from the similarity of their behaviour about
both axes. The Cm -factor has the purpose to take account of bending
The Cmz -factor is now related to the moment diagram of Mz , moment diagrams, which are non-uniform along the span or
which—in the case of intermediate lateral restraints—should be along a segment between lateral restraints, if relevant.
taken as the segmental diagram between the lateral restraints; It is assumed that the fictitious uniform bending moment
see Figs. 9 and 11. Cm · M has the same effect on the buckling behaviour as
the actual moment diagram. Non-uniform bending moment
5.5. Buckling caused by biaxial bending and axial compression diagrams have generally more favourable effects than uniform
N + M y + Mz ones—therefore, the Cm -factors are always smaller than or
equal to 1.0.
The design formulae are as follows (see also Fig. 11): As already mentioned in chapter 4 the “equivalent uniform
N Ed Cmy · M y,Ed Cmz · Mz,Ed moment factor” Cm differs from the analogous β M -factor in the
+ ky + 0.6 · k z ≤ 1 (13) previous ENV-version and it differs also from the Cm -factor
χ y · N pl,Rd M pl,y,Rd M pl,z,Rd
of Method 1, which takes the sinusoidal shape of the moment
N Ed Cmy · M y,Ed Cmz · Mz,Ed
+ 0.6 · k y + kz ≤ 1. diagram as its basic form. In Method 2 the approach of the
χz · N pl,Rd M pl,y,Rd M pl,z,Rd widely used “Austin formula” [13] has been applied for reasons
(14) of simplicity in practical use.
R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770 763

Fig. 10. Interaction factor k z .

– Secondly the Cm -factor is related to the shape of the moment


diagram of the relevant span between lateral restraints.
While Cmy is always related to the overall span of the
member, CmLT and Cmz may be determined by the segmental
moment diagram between intermediate restraints.

6. Members susceptible to torsional deformations—Class 1


and Class 2 sections

Fig. 11. Definition of Cm -factors. 6.1. General buckling behaviour


The Austin-formula (Eqs. (7) and (12)) is valid for linear
moment diagrams (Fig. 14(a)). The buckling behaviour of torsionally flexible members is
Although very simple in format it is able to describe the given by lateral–torsional buckling. It is illustrated by the
buckling effects very closely. Recent investigations [14] showed example of an HEB 300 under axial compression and uniform
that this concerns not only elastic behaviour, but also plastic bending moment M y (Fig. 15). The beam–column is free to
behaviour with or without imperfections. It further showed, that deflect about the y-axis along its overall span, however it
the cut-off by Cm = 0.4 at ψ ≤ −0.5 is conservative since is laterally restrained at midspan against buckling about the
Cm -values down to about 0.2 may be reached by numerical z-axis. At N = 0 pure LT-buckling between the lateral
simulation. This margin, however, has not been exploited in the restraints occurs. As the axial force grows, buckling about the
design rules of Method 2, since firstly the cross-section check y-axis—superposed by the local LT-buckling effect—becomes
at the end of the member becomes frequently governing and dominating. This behaviour is described by the interaction
secondly an additional reserve was considered appropriate for curve y–y and is approximated by Eq. (3). The difference
the cases of the bi-linear moment diagram. from pure flexural buckling acc. to Eq. (1) is obvious. The
In more general cases of non-uniform moment diagrams, LT-buckling behaviour about the z-axis (along the segment
e.g. under end-moments and transverse loading no formula between the lateral restraints) is described by the interaction
existed, so that the Cm -factor had to be recalculated from curve z–z and is approximated by Eq. (4); it is not relevant in
GMNIA-results of different moment distributions. Fig. 14(b) the given case.
shows such recalculated values for the case of uniform An illustration of the different typical buckling modes is
transverse loading. Herein, the horizontal parts of the curves given in Fig. 16. While for members, which are laterally free
indicate that the cross-section resistance is governing. The solid along the span, LT-buckling about the z-axis always governs,
line may be regarded as upper bound of the different curves, members with intermediate lateral restraints may also fail in
which are—analogously to the Austin formula—cut off by a buckling mode, where the deformation about the y-axis
the conservative limit of 0.4. Formulae for the Cm -factors, dominates and the LT-buckling effect between the restraints
which were derived from the calculated curves, are presented is just a superposed effect. The first case is described by the
in Table 1. interaction Eq. (4), the second one by Eq. (3). If the lateral
For the use of the Cm -factors the following two points should restraints are arranged at small distances, the buckling mode
be noted (see Table 1): y–y tends to the case of in-plane flexural buckling. So it may
– Firstly, the given Cm -factors may only be used if the two be resumed, that in the case of free members just Eq. (4) for
ends of the member can be regarded as fixed. In cases of buckling about z–z has to be considered, while for members
buckling in a sway mode the Cm -factor should be taken due with intermediate lateral restraints both Eqs. (3) and (4) have to
to simplification as 0.9. This applies in particular to columns be checked, since depending on the specific in- and out-of-plane
with a deformable upper end. slenderness one of the two buckling modes may be decisive.
764 R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770

Fig. 12. Cross-section interaction at N Ed → 0.

Table 1
Equivalent uniform moment factor Cm

6.2. In-plane buckling under N + M y An example of a very slender column (e.g. to be understood
as a column with the equivalent buckling length 2 · l) with IPE-
The design formula is given by Eq. (3): section illustrates again this behaviour (Fig. 17). This formula
(3) is a new one compared to the ENV-version and to many
N Ed Cmy · M y,Ed
+ ky = 1. (3) existing codes, where it was meant to be physically represented
χ y · N pl,Rd χLT · M pl,y,Rd by the response of Eqs. (1) and (4). It has been incorporated in
The interaction formula is equivalent to Eq. (1) apart from the EN-version mainly for reasons of consistency of the design
replacing the bending resistance M pl,y,Rd by the buckling concept.
resistance χLT · M pl,y,Rd . The buckling reduction factor
6.3. Out-of-plane buckling under N + M y
χLT accounts for the LT-buckling effect between the lateral
restraints. If the slenderness λ̄LT decreases, χLT will approach The design formulae are as follows:
1.0 and Eq. (3) becomes identical with Eq. (1), which means
N Ed M y,Ed
that the basically torsionally susceptible section will fail in + kLT =1 (4)
the flexural, in-plane buckling mode. Therefore, no separate χz · N pl,Rd χLT · M pl,y,Rd
flexural buckling check is needed for I- or H-sections, because 0.1 λ̄z n z 0.1 n z
it is always included in Eq. (3). kLT = 1 − ≥1− (15)
CmLT − 0.25 CmLT − 0.25
R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770 765

Fig. 13. Comparisons for flexural buckling under bi-axial bending.

Fig. 14. Equivalent uniform moment factor Cm .

λ̄z < 0.4 : kLT = 0.6 + λ̄z (16)


N Ed
nz = (17)
χz . N pl,Rd
CmLT = 0.6 + 0.4ψ ≥ 0.4 (see also Table 1). (18)
The interaction formula Eq. (4) is the traditional LT-buckling
formula. The interaction factor kLT has been recalculated
from GMNIA-results in analogous form as explained for the
k y -factor above. Fig. 18 illustrates the recalculated kLT -factor
for an IPE-section and three different moment diagrams.
The systematic behaviour is obvious. All the kLT -values
are smaller than 1.0. At λ̄z = 0 they start with the cross-
section interaction, then they increase moderately up to about
λ̄z ∼= 1 before they decline again to an approximately constant
level for large slenderness. The non-uniformity of the moment
diagram has a beneficial effect, which, however, reduces the Fig. 15. LT-buckling of torsionally flexible members.
kLT -factor significantly just for high n z -values, where—as
766 R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770

(a) LT—buckling mode y–y. (b) LT—buckling mode z–z.

Fig. 16. General buckling modes of torsionally flexible members.

already noted above—the influence is limited because of the


reduced magnitude of the bending moment. For most practical
cases the kLT -values will be closely below 1.0.
The formula for the interaction factor kLT has been
approximated in bilinear form—analogously to those of k y and
k z . The kink was chosen again at λ̄z = 1.0 and for higher λ̄z
the kLT -values remain constant (Fig. 19). A difference exists by
the inclusion of the effect of non-uniform moments implicitly
by CmLT in the kLT -factor. The reason for this is, that using the
Cm -factor of the Austin-formula again, an explicit application
of CmLT in the bending term is not appropriate.
As Fig. 19 shows, the simplified bi-linear form of the
formula leads to a linear approach of the kLT -factor towards 1,0
with λ̄z → 0. Since this does not represent the moderate drop of
the actual kLT -curves to the cross-section resistance at λ̄z = 0,
Fig. 17. LT-buckling of member with intermediate lateral restraints.
a cut-off formula Eq. (16) has been used, which allows us to
exploit the full resistance in this range of transition for small The formulae—in the case of vanishing bending M y —
slenderness below λ̄z = 0.4. This will be reached by applying provide a consistent transition to flexural buckling about the
Eq. (3), which accounts for the cross-section interaction in this z-axis.
slenderness range consistently. It may further be noted, that the Fig. 20 illustrates the results of the design formulae in
cut-off formula by approaching kLT = 0.6 for small slenderness comparison with those of the ENV-version and with GMNIA-
allows us to exploit approximately the convex plastic cross- results.
section interaction for biaxial bending M y + Mz (as already
shown in chapter 5). 7. Member buckling for Class 3 and Class 4-sections
For practical use it should be indicated, that the λ̄LT factor
The fully analogous derivation of interaction formulae by
and the CmLT -factor should be related to the specific moment
numerical simulation for elastic cross-sections as shown for
diagram between the relevant lateral restraints. This is the full
plastic sections above causes a number of basic difficulties.
span for a laterally free member or the segmental part of the
These are connected with the abstract definition of “elastic”
member in the case of intermediate lateral restraints. It need sections in EC3 on the one hand and with the limited knowledge
not be mentioned that this relation to lateral restraints should on the partial-plastic buckling behaviour of such sections
also be used for the determination of λ̄z and n z respectively. on the other. In this sense, the definition of fully elastic
behaviour excludes the consideration of residual stresses and
6.4. LT-buckling under biaxial bending and axial compression also does not allow us to recalculate buckling reduction factors
N + M y + Mz of such sections. Recent numerical simulations of Class 3-
sections have, however, confirmed partial-plastic resistance of
The design formulae are as follows: such sections [17,18]. They indicate that plastic interaction
behaviour—although related to the resistance Mel,Rd —exists.
N Ed Cmy · M y,Ed Cmz · Mz,Ed On this basis the derivation of the design formulae for Class 3
+ ky + 0.6 k z ≤ 1 (19)
χ y N pl,Rd χLT · M pl,y,Rd M pl,z,Rd and Class 4 sections takes a moderate account of partial-plastic
N Ed M y,Ed Cmz · Mz,Ed effects, in order to use some of this beneficial behaviour for
+ kLT + kz ≤ 1. (20) design.
χz N pl,Rd χLT · M pl,y,Ed M pl,z,Rd
The interaction formulae for sections of Class 3 and 4 are
The interaction formulae Eqs. (19) and (20) have been fully analogous with Eqs. (1) to (4), however the moment
developed by extending the given formulae (3) and (4) by the resistances M pl have been replaced by Mel or Me f f and the
terms of Mz . This is in full analogy to the development of the shift of the neutral axis of Class 4-sections has been accounted
procedure in chapter 5.4 for flexural buckling. for by M = e N · N.
R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770 767

Fig. 18. Interaction factor kLT .

Fig. 19. Design values of interaction factor kLT .

The interaction factors k y and k z have been derived by elastic the accuracy or safety of the results and in the practical use of
second order theory and have been developed in simplified bi- the two methods.
linear format as above. The interaction factor for the out-of- In the following explanations are given to the above
plane buckling mode under M y has been defined by 0.8 k y , mentioned topics, concerning Method 1 and Method 2 although
in which the factor 0.8 accounts for the partial-plastic effect. being aware that full answering of all aspects would require
Similarly such an effect has been used for the factor kLT by a paper at least so long as the existing one. The author’s
applying a reduced beneficial plastic effect of about 50%. objective for this paper has originally been focussed just on the
Especially for the calculation of the reduction factor χLT for background of Method 2, so that together with [3] both methods
lateral torsional buckling an enhanced method is given in [17], are presented similarly by their authors.
which is also compared to test results. A first explanation refers to differences in the meaning of
These design formulae, presently take just moderate account the Cm -factors, which are used in the case of flexural buckling
of the plastic effects existing in these sections. A more extensive to account for non-uniform moment distribution. While the
use of this load carrying behaviour may be expected after Cm factors of Method 2 are based on equivalent “uniform”
further studies and test confirmation, which however are still moments, the Cm factors of Method 1 are based on equivalent
under way at present. “sinusoidal” moments. These Cm factors of Method 1—being
derived from the elastic buckling theory—depend on the critical
8. Differences between Method 1 and Method 2
flexural buckling load, while those of Method 2 follow the
As said in the introduction of this paper, EN1993-1-1 Austin-formula. Because of these different basic assumptions
(EUROCODE 3) [1] also provides besides Method 2 the the Cm -factors are generally different in magnitude.
alternative of Method 1 for member buckling. However the It is important to mention that the CmLT -factors of both
code does not give any indication on the differences between methods to account for non-uniform moment distribution in
the two methods, which had been formally poured into the same the case of lateral torsional buckling are also quite different in
type of basic interaction formulae Eqs. (6.61) and (6.62) in principle. While CmLT in Method 1 is a factor connected with
EN1993-1-1 by the authors of Eurocode 3. In this respect it Cmy , the CmLT in Method 2 is an independent factor covering
should also be noted that in addition to Method 1 and Method 2 the effect of the relevant moment diagram between intermediate
the code EN1993-1-1 provides two more methods, i.e. the lateral restraints. Again the CmLT -factors of the two methods are
so-called General Method and the Stable-Length Method. generally different.
Therefore, in total there are four alternative methods open to The application range of the two methods is in both cases
the designer, which concern member buckling under bending basically a single span member of doubly symmetrical cross-
and axial compression. section subjected to axial compression, end-moments and
Therefore some comments might be useful for designers, if transverse loading. In the case of Method 1 the transition
there are differences in the application range of the formulae, in from flexural buckling to lateral torsional buckling has been
768 R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770

Fig. 21. System and internal forces.

user to choose, which format would be more beneficial in his or


her specific design situation.

9. Worked examples

9.1. General

The buckling check is based on the design values of actions


and resistances as defined in [1]. The safety factors γ M0 and
γ M1 have been taken as 1.0 as recommended in EC3 for
Fig. 20. Comparison for LT-buckling under biaxial bending. buildings.

9.2. Flexural buckling of member under bending and axial


approximated according to the torsional rigidity of the member, compression
so that, therefore, a smooth transition between the two cases
is given. Method 2 distinguishes the two standard cases of Forces and buckling lengths, Fig. 21
torsionally stiff and torsionally flexible members separately,
which seems to be in line with the traditional thinking of N Ed = 300 kN
design engineers. It should be mentioned that Method 2 has M y,Ed = 40 kN m (parabolic moment diagram)
been developed principally in such a way that the application L = L cr,y = L cr,z = 5.6 m
range also covers members with intermediate lateral restraints,
which may not be effective for in plane flexural buckling. The cross-section: RHS 200/100/10, S235, hot-finished
explicitly given guidelines for the use of specific CmLT -factors
N Rd = 54.9 · 23.5/1.0 = 1290 kN
makes the practical application for these cases easy.
The accuracy of the two methods has been checked for M y,Rd = 341 · 23.5/1.0 = 8010 kN cm
each of them by statistical evaluations of the small number of slenderness ratio:
available test results and the big number of theoretical FEM
560
calculations, reported in several TC8-documents, e.g. [12]. So λy = = 0.857
from this point of view the methods are equivalent in general. 6.96 · 93.9
560
However, this does not exclude differences of the results in λz = = 1.50
specific examples. Since Method 1 has been based on the 3.98 · 93.9
theoretical derivation of the spatial flexural buckling modes buckling reduction factors
and uses several specific factors it covers both flexural modes
χ y = 0.762 buckling curve a
most closely, in particular out-of-plane buckling. Method 2 is
based on the theoretical derivation of the in-plane buckling χz = 0.372 buckling curve b.
formula with just one compact factor, which was calibrated Equivalent uniform moment factor and interaction factor:
on the elastic–plastic member-capacity. Therefore, it covers
in-plane buckling most closely and approximates out-of-plane Cmy = 0.95 (see Table 1)
buckling. In the case of lateral–torsional buckling both methods
are extensions from flexural buckling calibrated mainly at the 300
ny = = 0.305 (6)
results of elastic–plastic numerical simulations. 0.762 · 1290
With respect to the practical use of the formulae the two
for λ y ≤ 1.0 follows
methods have different objectives. Method 2 was definitely
aimed at the use by hand-calculation and, therefore, used the k y = 1 + (0.857 − 0.2) · 0.305 = 1.20. (5)
same simple structure of the formulae as traditional interaction
Verification:
formulae of many existing codes. Method 1 aimed at a structure
of the formulae, which is as far as possible based on elastic, 0.95 · 4000
0.305 + 1.20 · = 0.874 ≤ 1 (1)
theoretical derivation described by a larger number of factors. 8010
Accordingly, the use of computer aids seems to be useful and 300
= 0.625 ≤ 1. (2b)
necessary for practical application. In this view it is up to the 0.372 · 1290
R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770 769

Fig. 23. System and internal forces.

Fig. 22. System and internal forces. non-uniform moment diagram to χLT,mod = χLT / f , where f
has approximatively been based on
9.3. Lateral–torsional buckling of a column in a two-storey
kc = 0.752 for ψ = 0.
frame  
f = 1 − 0.5(1 − 0.752) 1 − 2(0.784 − 0.8)2 = 0.876
The lower column of a two-storey frame with a canopy
0.774
(see Fig. 22) should be checked on the basis of the equivalent χLT,mod = = 0.884.
column method. First order internal forces and buckling length 0.876
have been taken from example 8.17 in [15]. Equivalent uniform moment factor and interaction factors:
Forces and buckling length:
Cmy = 0.9 for sway mode, see Table 1
N Ed = 664 kN
M y,Ed = 668 kN m CmLT = 0.6 (18)
(triangular moment diagram with internal step) 664
ny = = 0.154 (6)
L cr,y = 2.25 · 4.2 = 9.45 m 0.906 · 4760
L = L cr,z = 4.2 m for λ y ≤ 1.0 follows:
cross-section: IPE 550, S355 k y = 1 + (0.555 − 0.2) · 0.154 = 1.05 (5)
664
N Rd = 134 · 35.5/1.0 = 4760 kN nz = = 0.305 (17)
0.457 · 4760
M y,Rd = 2787 · 35.5/1.0 = 98900 kN cm.
for λz ≥ 1.0 follows:
Critical buckling moment Mcr :
0.1
π 2 · 21000 · 2670 kLT = 1 − · 0.305 = 0.913. (15)
Ncr,z = = 3140 kN 0.6 − 0.25
4202
 Verification:
1884000 420 · 124 2 0.9 · 66800
Mcr = 1.6 · 3140 · + 2 0.154 + 1.05 · = 0.876 ≤ 1 (3)
2670 π · 2.6 · 2670 0.884 · 98900
= 161000 kN cm 66800
0.305 + 0.913 · = 1.00 ≤ 1. (4)
0.884 · 98900
hereby, the factor C1 for the given moment diagram has been
taken as 1.6 acc. to [15]. 9.4. Lateral–torsional buckling of a column under bi-axial
Slenderness ratio: bending and compression
945
λy = = 0.555 The column of a single-storey building should be checked
22.3 · 76.4
420 on basis of the equivalent column method (see Fig. 23). First-
λz = = 1.24 order internal forces and buckling lengths have been taken from
4.45 · 76.4
 example 8.7 in [15].
98900
λLT = = 0.784 Forces and buckling length:
161000
buckling reduction factors N Ed = 620 kN
M y,Ed = 202 kN m (triangular moment diagram)
χ y = 0.906 buckling curve a
Mz = Mz,Ed = 7.16 kN m (parabolic moment diagram)
χz = 0.457 buckling curve b
L cr,y = 2.3 · 4.6 = 10.6 m
χLT = 0.774 LT-buckling curve c.
L = L cr,z = 4.6 m
The reduction factor χLT has been determined according to
cross-section: IPE 450, S355
EC3-1-1, 6.3.2.3, based on λ̄LT,0 = 0.4 and β = 0.75. It has
further been modified by use of the f -factor accounting for the N Rd = 98.8 · 35.5/1.0 = 3510 kN
770 R. Greiner, J. Lindner / Journal of Constructional Steel Research 62 (2006) 757–770

M y,Rd = 1702 · 35.5/1.0 = 60400 kN cm 0.214 + 0.406 + 0.07 = 0.69 ≤ 1


Mz,Rd = 276 · 35.5/1.0 = 9800 kN cm
20200 0.95 · 716
critical buckling moment Mcr : 0.495 + 0.859 · + 1.69 · ≤1 (20)
0.831 · 60400 9800
π 2 · 21000 · 1680
Ncr,z = = 1650 kN 0.495 + 0.346 + 0.117 = 0.96 ≤ 1.
4602

791000 4602 · 67.1 Acknowledgements
Mcr = 1.77 · 1650 · + 2
1680 π · 2.6 · 1680
= 82600 kN cm The developments of the new interaction formulae were
based on a wide scope of numerical investigations, comparisons
hereby, the factor C1 for the given moment diagram has been and statistical evaluations and could only be carried out through
taken as 1.77. the efforts of the collaborators of the two authors. Special
slenderness ratio: thanks are given to Robert Ofner, Günther Salzgeber and
1060 Peter Kaim of TU Graz and to Andreas Rusch, Junping
λy = = 0.750 Wang/Kunming and Stefan Heyde of TU Berlin.
18.5 · 76.4
460
λz = = 1.46 References
4.12 · 76.4

60400
λLT = = 0.855 [1] EN1993-1-1. Eurocode 3. Design of steel structures, general rules and
82600 rules for buildings. 2005.
[2] ENV1993-1-1. Eurocode 3. Design of steel structures, general rules and
buckling reduction factors:
rules for buildings. 1993.
χ y = 0.824 buckling curve a [3] Boissonnade N, Jaspart J-P, Muzeau J-P, Villette M. New Interaction
formulae for beam-columns in Eurocode 3. The French-Belgian approach.
χz = 0.357 buckling curve b Journal of Constructional Steel Research 2004;60:421–31.
χLT = 0.729 LT-buckling curve c. [4] Greiner R. Background information on the beam–column interaction
Formulae at Level 1. ECCS Report No. TC8-2001-021, 20; Sept. 2001.
The reduction factor χLT has been determined according to [5] Greiner R, Lindner J. Die neuen Regelungen in der europäischen Norm
EC3-1-1, 6.3.2.3 based on λ̄LT,0 = 0.4 and β = 0.75. It has EN1993-1-1 für Stäbe unter Druck und Biegung. Stahlbau 2003, Heft 3,
p. 157–72.
further been modified by use of the f -factor accounting for the
[6] Greiner R, Kaim P. Erweiterung der Traglastuntersuchungen an Stäben
triangular moment diagram with ψ = 0. unter Druck und Biegung auf einfach—symmetrische Querschnitte.
  Stahlbau 2003;72(Heft 3):173–80.
f = 1 − 0.5(1 − 0.752) 1 − 2(0.855 − 0.8)2 = 0.877 [7] Ofner R. Traglast von Stäben aus Stahl bei Druck und Biegung.
Dissertation, Institut für Stahlbau, Holzbau und Flächentragwerke der TU
0.729
χLT,mod = = 0.831. Graz, Heft 9; 1997.
0.877 [8] ABAQUS Software. Hibbitt, Karlsson & Sorensen Inc., Version 5.7.
Equivalent uniform moment factor and interaction factors: [9] Greiner R, Ofner R, Salzgeber G. Verification of GMNIA-results. ECCS-
validation group. Report 2; July 1998.
Cmy = 0.9 for sway mode, see Table 1 [10] Lindner J, Rusch A, Heyde S. Evaluation of different design concepts for
flexural buckling with regard to test results and ultimate load calculations.
Cmz = 0.95 for parabolic moment Report 2131E, TU Berlin; November 1998.
CmLT = 0.6 for triangular moment [11] Lindner J. Interaktionsgleichungen für das Biegeknicken bei Druck und
zweiachsiger Biegung. Schlussbericht zum DIBt-Forschungsvorhaben
620 IV 1-5-866/98, Bericht 2135 des Instituts für Baukonstruktionen und
ny = = 0.214 (6) Festigkeit der TU Berlin, 10.6.19999.
0.824 · 3510 [12] Lindner J, Heyde S. Evaluation of interaction formulae at Level 1
approach with regard to ultimate load calculations and test results—
for λ y ≤ 1.0 follows
flexural buckling and lateral torsional buckling. Report 2144E, TU Berlin.
k y = 1 + (0.750 − 0.2) · 0.214 = 1.12 (5) ECCS Report No. TC8-2001-017, 30.7.2001.
[13] Austin WJ. Strength and design of metal beam-columns. Journal of
620
nz = = 0.495 (11) Structural Division 1961;ASCE 87(ST 4):1–32.
0.357 · 3510 [14] Kaim P. Spatial buckling behaviour of steel members under bending
and axial compression. Ph.D. Institute for Steel Structures and Shell
for λz ≥ 1.0 follows Structures, TU Graz; Heft 12-2004.
[15] Lindner J, Schmidt H, Scheer J. Stahlbauten, Erläuterungen zur DIN
k z = 1 + 1.4 · 0.495 = 1.69 (9)
18800 Teil 1 bis 4. Berlin: Ernst & Sohn, Beuth, 3. Auflage; 1998.
0.1 [16] Lindner J. Design of beams and beam columns. Progress in Structural
kLT = 1 − · 0.495 = 0.859. (15)
0.6 − 0.25 Engineerings and Materials. 2003;5:38–47.
[17] Rusch A, Lindner J. :Application of Level1 interaction formulae to class
Verification: 4 sections. Thin-walled Structures 2004;42:279–93.
0.9 · 20200 0.95 · 716 [18] Lechner A. Plastic cross-section capacity of semi-compact steel sections.
0.214 + 1.12 · + 0.6 · 1.69 · ≤ 1(19) Ph.D. Institute for Steel Structures and Shell Structures, TU Graz; 2005.
0.831 · 60400 9800

You might also like