EUROSTEEL 2008, 3-5 September 2008, Graz, Austria

ON THE DESIGN CURVES FOR BUCKLING PROBLEMS

József Szalai a, Ferenc Papp b
a
KÉSZ Ltd., Budapest, Hungary
b
Budapest University of technology and Economics, Department of Structural Engineering, Budapest, Hungary

INTRODUCTION
The paper deals with the design curves for the member buckling problems included in many
modern standards as the basis for the stability design of steel structures. In the Eurocode 3 [1]
member buckling curves are used for the two basic cases: the flexural buckling of columns and the
lateral-torsional buckling of beams (LTB). These two buckling curves are of quite high concern and
necessary for the standard design of general member stability, therefore the accurate theoretical and
experimental verification of them is very important. The column buckling has been researched
comprehensively for many years as the basic and simplest case of stability problems. The research
included a huge number of experimental tests, complete theoretical investigations using both
analytical and numerical models, and deep probabilistic examinations [2]. As a main result the
multiple column curves concept has been developed, and as the basic design model the Ayrton-
Perry formula (APF) was adopted. This formula has the following main advantages: clear
mechanical background, simplicity and flexibility; however it is important to note that it is
connected directly with the flexural buckling phenomena. The LTB is far more complicated to
handle, the experimental research can usually run into modelling difficulties, the analytical
description is limited for the most basic cases; nowadays the most appropriate analysis is the
numerical simulation [3]. Exploiting the flexibility of the APF this form was chosen as the design
curve for LTB adjusting the design parameters carefully so as to be in accordance with the results of
the numerical simulations [3]. Although such a way the LTB curves can be considered adequately
verified, the theoretical background of the direct connection between the APF and LTB is not
clarified so far. The most important drawback of this is connected with the imperfection factor as
the main parameter of the APF: the correct form of the generalized imperfection factor is unknown
and the appropriate equivalent geometric imperfections can not be defined. The paper shows a
possible way to connect directly the LTB and APF by generalizing the APF using a special rule for
the definition of the imperfection factor.

1 COLUMN BUCKLING
The APF was applied originally for geometrically imperfect columns loaded by uniform
compression, where the load carrying capacity is corresponds to the onset of yielding at the most
compressed fibre [4]. It is important to note that this formula neglects the influence of other
imperfections (residual stress, eccentricity) and partial plasticity; however these additional effects
can be efficiently modelled through the generalized imperfection factor. The maximum total second
order lateral displacement of such a simply supported prismatic member assuming sinusoidal
imperfections writes:
v0
v= (1)
1− N N cr . z
where v0 is the midspan amplitude of the half-sine wave, Ncr.z is the elastic critical buckling load
about the minor axis and N is the actual compressive force. At the midspan cross-section the most
compressed fibre should reach the yield stress (first yield criterion):
N N ⋅v
+ = fy (2)
A Wz
where A is the cross-sectional area, Wz is the elastic sectional modulus and fy is the yield stress.
Substituting Eq. (1) into Eq. (2) the original form of the APF can be written:
 (σ cr − σ b )( f y − σ b ) = σ bσ crη (3)

where σcr is the elastic critical compressive stress, σb is the actual compressive stress, and
A
η = v0 is the generalized imperfection factor. Eq. (3) can also be written by the standard
Wz
notations:
 1 1  1
χ 2 + χ  −1− 2
− 2
η+ 2
=0 (4)
 λ λ  λ
Af y N
where λ is the well-known slenderness ( λ = ), χ is the buckling reduction factor ( χ = ).
N cr . z Af y
The solution of Eq. (4) yields the standard form of the column buckling curve.

2 GENERALIZATION OF THE AYRTON-PERRY FORMULA

The derivation of the column buckling curve has distinct steps, which can be generalized in order to
be applicable for other buckling problems, in our case for LTB. The final aim is always the
determination of the generalized form of Eq. (4), which yields the reduction factor (χgen) in terms of
the slenderness and imperfection factor. The general form of the slenderness writes:
αy
λgen = (5)
α cr
where the αy and αcr are the first order load amplifiers corresponding to the first yield of the cross
section and the elastic critical load respectively. This definition is somewhat similar to the one
applied in the general method for stability check in EC3 [1], but is not exactly the same, because the
final solution corresponds still to the first yield criterion. The steps of the derivation of APF for
general stability problems are the following:
1. Determination of the one parameter load and displacement – and corresponding geometrical
imperfection – variables, and defining the governing differential equation(s);
2. Applying suitable initial geometric imperfections;
3. Determination of the second order elastic displacements in terms of the load parameter, elastic
critical loads (amplification factors) and initial imperfections;
4. Definition of the first yield criterion based on second order section forces which depend on the
second order elastic displacements;
5. From the first yield criterion writing the – normally – quadratic equation for the reduction
factor in terms of the generalized slenderness and imperfection, and the solution is the
appropriate buckling curve.
It is very important to understand, that the crucial point of the general process is step 2, since the
proper definition of the shape of imperfections has a great impact on the main equation derived
from the first yield criterion (step 4). The main principle of this paper is that if the shape of the
initial geometric imperfection is taken exactly the same as the buckling shape (or first eigenshape)
of the perfect, elastic member then the equation of step 5 is always has the same form as Eq. (4), the
only deviations between the different buckling problems are the concrete definition of slenderness
(Eq. (4)) and imperfection factor. In the paper this theorem will be proved for LTB of beams
showing that the usual simplifications in the shape of initial imperfection do not lead to the form of
Eq. (4), therefore the APF can not be constructed.

3 LATERAL-TORSIONAL BUCKLING
Consider the basic case for LTB, which is a simply supported prismatic beam loaded by uniform
bending moment about its major axis. In this case the one parameter load is the bending moment M,
 z

y y

z θ(x)

w(x) v(x)

Fig. 1. Displacements of the cross-sections in LTB

the displacements are the two deflections w(x), v(x) and section rotation θ(x) depending from the
axis x through the centreline of the member (Fig. 1), and the well-known governing system of
partial differential equations is the following:
∂ 2 w( x)
EI y = −M
∂x 2
∂ 2v( x)
EI z + M (θ ( x) + θ 0 ( x) ) = 0 (6-8)
∂x 2
∂ 3θ ( x) ∂θ ( x)  ∂v( x) ∂v0 ( x) 
EI ω 2
− GI SV + M + =0
∂x ∂x  ∂x ∂x 
where v0(x) and θ0(x) are the initial imperfections. Since the first equation does not influence the
buckling it will be omitted henceforward. Step 2 is now the selection of the suitable shape for the
initial imperfections. Omitting the initial imperfections from Eqs. (7-8) the solution gives the
buckling shape of the perfect member for which the following relationship holds:
M cr
vb ( x) = θ b ( x) (9)
N crz
where the b subscript refers to the buckling shape, and Mcr is the elastic critical bending moment.
This is the key relationship which should be applied to the shape of initial imperfections further on.
In step 3 solving Eqs. (7-8) with the general imperfection terms yields the total second order elastic
displacements, which takes the following form at the midspan cross-section:
 M 
v  1  1 N cr . z   v0 
θ  = 2
   (10)
  1 − (M / M cr )  N cr . z M 1  θ 0 
 M cr2 
This is a general relationship containing a quadratic amplification factor and showing that both
initial imperfections influence both displacements. At this point it is a usual simplification to
neglect one of the two imperfection terms [5, 6], however following this way the quadratic
amplification factor still remains, and the APF in the form of Eq. (4) can not be achieved, since this
main equation will be cubic instead of quadratic with an untreatably complex solution. So applying
the principle introduced in the second section, and accordingly substituting the relationship Eq. (9)
into Eq. (10) one obtains the following form:
 v  1 1 0  v0 
θ  = 1 − M / M 0 1 θ  (11)
  cr   0 
That is the crucial step described in the second section which is resulted in a far simpler relationship
and has the same form as Eq. (4) obtained for flexural buckling with the same linear amplification
factor. The other main conclusion is that if the direction of the vector of initial imperfections is the
same as the one of the vector of the buckling displacements then the certain displacements depend
on the corresponding imperfection only. Moving on to step 4 the first yield criterion in the midspan
cross-section is the following [7]:
M M ⋅ (v + wθ ) − GI SV ⋅θ M ⋅θ
+ + = fy (12)
Wy Wω Wz
where Wy, Wz and Ww are the elastic major axis, minor axis and warping sectional moduli
respectively, and w is the deflection about the major axis (Fig. 1) which can be expressed at
M ⋅ L2
midspan as w = . On the left side of Eq. (12) the second and third terms are the normal
2 EI y
stresses due to the second order bimoment and minor axis bending respectively. In the second order
bimoment the advantageous effect of the Saint-Venant torsional rigidity, and the additional mixed
term due to the simultaneous deflection (w) and twist (θ) are considered. Substituting Eq. (11) into
Wy f y M
Eq. (12) and applying λLT = and χ LT = the quadratic APF for LTB writes:
M cr Wy f y

 1 1  1
2
χ LT β LT .2 + χ LT  − 1 − 2
− 2
η LT  +
2
β LT .1 = 0 (13)
 λ LT λ LT  λ LT
where the three additional terms are the new generalized imperfection factor, the Saint-Venant
torsional rigidity factor and the deflection factor (applying the relationship Eq. (9)):
Wy Wy  W y N cr . z W y 
η LT = v0 + θ0= v0  + 
Wω Wz  Wω M cr Wz 
GI N GI SV
β LT .1 = 1 + θ 0 SV = 1 + v0 cr . z (14-16)
Ww f y M cr Ww f y
W y M cr π 2 W y N cr . z π 2
β LT .2 = 1−θ0 = 1 − v0
Wω N cr . y 2 Wω N cr . y 2

Using these expressions the buckling curve for LTB can be written as the solution of Eq. (13) in the
well-known form of the EC3 [1]:
β LT .1
χ LT =
2
φ LT + φ LT − β LT .1 β LT .2 λ2LT (17-18)
φ LT = 0.5[1 + η LT + λ 2
LT ]
This is the fundamental form of the APF based LTB curve belonging to the first yield criterion of
Eq. (12) and the specially shaped initial geometric imperfection defined by Eq. (9).

4 DISCUSSION
In order to reveal the most important features and characteristics of the obtained LTB curve –
keeping in mind that it is only the basic form belonging to the first yield criterion – let examine
deeper the main peculiarities of Eqs. (17-18) which are the effects of the special imperfection
factors of Eqs. (14-16). In the recent form of EC3 [1] multiple buckling curves are used for LTB,
 and the distinction is made upon the section type (I-shaped or not), fabrication process (hot-rolled
or welded) and in the case of I profiles the height-width ratio (h/b). In this paper we deal with I-
shaped profiles and since the effect of fabrication – through the residual stresses – has been omitted
only the influence of the geometrical parameters is investigated further. It is found, that it is not
solely the h/b ratio what affects significantly the shape of the LTB curves, therefore the
examinations are carried out for six different hot-rolled sections (HEAA300, HEA300, HEB300,
HEAA900, HEA900, HEB900) in order to show the other types of influential parameters. In Figs.
2-6 the solid lines represent the 300 sections, the dashed lines represent the 900 sections, and the
thickness of the lines show the heaviness of the proper section (thin – HEAA, thick – HEB). All the
results are calculated considering a v0=L/1000 initial out-of-straightness and a S235 material. In
Fig. 2 the buckling curves – and the non-dimensional Euler curve –, in Figs. 3-6 the imperfection
factors of Eqs. (14-16) are plotted for the six cross-sections against the non-dimensional
slenderness.

please improve the

resolution

Fig. 2. The LTB buckling curves

From Fig. 2 it can be first noticed that there is only a little difference between the curves which
correspond to higher h/b ratio (~3, dashed lines), and generally these curves take higher values at
higher slenderness (λLT < 1) and lower values at lower slenderness then the curves correspond to
lower h/b ratio (~1, solid lines). It is important here to note that the EC3 distinguish the curves
belonging to different h/b ratios only by the generalized imperfection factor (here ηLT) and the β
factor remains the same for all the cases, this approach does not yield the above difference between
the curves. On the other hand there is significant difference between the solid curves at medium
slenderness; the curve of the HEB300 section takes higher value with up to 10% then curve of
HEAA300 section. This is mainly because of the fact that more compact and heavier profiles have
considerably smaller ηLT values while the advantageous effect of their Saint-Venant torsional
rigidity (βLT) is more significant as it is seen in Figs. 3-4. This observation suggests an additional
distinction between the heavy and light sections by for example the b/tf ratio. It is also interesting to
remark the effect of the two βLT values in Figs. 5-6. Generally the deflection factor βLT.2 has a
dominant influence in case of wider flange profiles (solid lines); while for the higher sections this
effect is negligible. It is the result of the much greater deflection along with almost the same minor
axis inertia, and that is the reason for the difference between the solid and dashed curves at higher
slenderness. This phenomenon suggests that the β factor in the design LTB curves of the EC3
should also depend on the h/b ratio.
 please improve the
resolution

Fig. 3. The generalized imperfection factors Fig. 4. The Saint-Venant torsional rigidity factors

please improve the

resolution

Fig. 5. The deflection factors Fig. 6. The combined effect of βLT.1 βLT.2

5 CONCLUSION
The main point of the paper was to clarify the mechanical basis of the design curve used for lateral-
torsional buckling in the EC3. This aim has been achieved by generalizing the original form of the
Ayrton-Perry formula – the basis for flexural buckling design curve – introducing a specific
definition of the initial geometric imperfection. By this approach the direct connection between the
generalized Ayrton-Perry formula and the lateral-torsional buckling has been established and the
correct meaning of the special imperfection factors has been deduced. Examining these factors the
specialties of the formula have been revealed and suggestions for improvements have been made.

REFERENCES
[1] European Standard, EuroCode 3, Design of Steel Structures – Part1-1: General rules and rules
for buildings, EN 1993-1-1, 2005
[2] Galambos T. V., Guide to stability design of metal structures, Wiley, 1995
[3] Greiner R., Salzgeber G., Ofner R., New lateral-torsional buckling curves κLT - numerical
simulations and design formulae, ECCS TC8 –Report No. TC8-2000-014, 2000
[4] Ayrton W. E., Perry J., On Struts, The Engineer, 1886
[5] Kaim P., Spatial buckling behaviour of steel members under bending and compression, PhD
Dissertation, TU Graz, 2004
[6] Boissonnade N., Villette M., Muzeau J.P., About amplification factors for lateral-torsional
buckling and torsional buckling, In: Festschrift Richard Greiner, TU Graz, 2001
[7] Papp F., Computer Aided Design of Steel Beam-Column Structures, PhD Dissertation,
Budapest University of Technology and Economics, Edinburgh, 1994