Journal of Constructional Steel Research 72 (2012) 61–74

Contents lists available at SciVerse ScienceDirect

Journal of Constructional Steel Research

Development of a consistent buckling design procedure for tapered columns

Liliana Marques a,⁎, Andreas Taras b, Luís Simões da Silva a, Richard Greiner b, Carlos Rebelo a
a
ISISE, Department of Civil Engineering, University of Coimbra, Coimbra, Portugal
b
Graz University of Technology, Institute for Steel Structures, Graz, Austria

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

Article history: EC3–EN 1993-1-1 provides several methodologies for the stability verification of members and frames.
Received 15 June 2011 When dealing with the verification of non-uniform members in general, with tapered cross-section, irregular
Accepted 6 October 2011 distribution of restraints, non-linear axis, castellated, etc., the code mentions the possibility of carrying out a
Available online 17 November 2011
verification based on 2nd order theory; however, several difficulties are noted when doing so, in particular
when the benefit of plasticity should be taken into consideration.
Keywords:
Stability
Other than this, there are yet no guidelines on how to apply standardized, easily reproducible rules as those
Eurocode 3 contained in Section 6.3.1 to 6.3.3 of the code to non-uniform members. As a result, practical safety verifica-
Non-uniform members tions for these members are often carried out using conservative assumptions, not accounting for the advan-
Tapered columns tages non-uniform members provide. In this paper, firstly, available approaches for the stability verification
FEM of non-uniform members are discussed. An Ayrton–Perry formulation is then derived for the case of non-
Steel structures uniform columns. Finally, and followed by a numerical parametric study covering a range of slenderness,
cross-sections and fabrication process, a design proposal is made for the relevant case of in-plane flexural
buckling of linearly tapered columns subject to constant axial force. The proposal is consistent with current
rules for uniform columns provided in EC3-1-1, i.e., clause 6.3.1.
© 2011 Elsevier Ltd. All rights reserved.

1. Introduction advanced numerical analysis (GMNIA), besides the lack of guidance

The stability of uniform members in EC3-1-1 [1] is checked by the
application of clauses 6.3.1 — stability of columns; clause 6.3.2 — stability
of beams and clause 6.3.3 — interaction formulae for beam-columns.
Regarding the stability of a tapered member, clauses 6.3.1 to 6.3.3
do not apply and verification should be performed either by a cross-
sectional verification based on second-order internal forces or, with
some difficulties, according to clause 6.3.4 (the so-called “General
Method”, see [2,3]). Alternatively, as the most advanced but often
practically not feasible variant, the resistance may also be checked
by a numerical analysis that accounts for (geometrical and/or material)
imperfections and (material and/or geometrical) nonlinearities, hence-
forth denoted as GMNIA.
However, for the “General Method”, several difficulties are noted
for the verification of a non-uniform member [4]. These are:
(i) shape and magnitude of imperfections (geometrical and materi-
al); (ii) choice of an appropriate buckling curve and, as a result, of
an adequate imperfection factor; (iii) definition of cross-section
class; (iv) determination of cross-section properties for verification
(or critical design location) — this problem also exists with respect
to the application of clauses 6.3.1 to 6.3.3. For the application of
⁎ Corresponding author at: Department of Civil Engineering, University of Coimbra,
Polo II, Pinhal de Marrocos, 3030-290 Coimbra, Portugal. Tel.: + 351 239 797260;
fax: + 351 239 797217.
E-mail address: lmarques@dec.uc.pt (L. Marques).
concerning the shape and magnitude of imperfections, the volume
of work is still incompatible with practical application and a highly
experienced engineer is required [5]. Moreover, there are no guide-
lines yet to overcome any of these issues.
In this paper, the case of columns subject to in-plane buckling
with varying cross-section and with constant axial force is studied.
It is the purpose of this paper to: (i) discuss the current difficulties
in performing stability verification of non-uniform members; (ii) pre-
sent the theoretical background for non-uniform columns; (iii) carry
out a parametric study of FEM numerical simulations of non-uniform
columns; and (iv) develop a proposal for the stability verification of
in-plane buckling of tapered columns with constant axial force.
2. Available approaches for the stability verification of
non-uniform members
2.1. General
Fig. 1(a) and (b) illustrates recent examples of the use of curved
and tapered members or members with polygonal centroidal axis.
The evaluation of the buckling resistance of such members lies outside
the range of application of the interaction formulae of EC3-1-1 and
raises some new problems to be solved.
Firstly, taking as an example the case of beam-columns (uniform
or not) with varying ratios of MEd to NEd over the member length,
the cross-sectional classification changes from cross-section to cross-

0143-974X/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2011.10.008
62 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74

Notations
ðxÞ
λ Non-dimensional slenderness at a given position

y
λ Non-dimensional slenderness for flexural buckling, y-y
A Cross-section area axis
E Modulus of elasticity ξ. η Rectangular coordinates, longitudinal and transversal
FEM Finite Element Method χ(x) Reduction factor at a given position
GMNIA Geometrical and Material Non-linear Analysis with χnum Reduction factor (numerical)
Imperfections
Imin Minimum 2nd moment of area
Iy,eq Equivalent 2nd moment of area, y-y axis
L Member length section, see the example of Fig. 2. On the safe side, an elastic verifica-
Leq Equivalent Member length tion considering class 3 cross-section is likely to be performed al-
LEA Linear Eigenvalue Analysis though a qualitative analysis of the example shows that the stresses
MEd Design bending moment in the interval corresponding to class 3 cross-section are not critical
MR Resistant bending moment compared to the stresses in the remainder of the member.
My,Ed Design bending moment, y-y axis Secondly, the determination of an adequate buckling curve is also
NEd Design normal force necessary and leads to inconsistencies, such as:
Nconc Concentrated axial force
Ncr,tapered Elastic critical force of the tapered column (i) The buckling curves in the code are geared towards specific
Npl Plastic resistance to normal force at a given cross-section buckling cases. That is why the interaction formulae and coef-
NR Resistant normal force ficients for uniform members have to take into account the
Q Shear force transitions from one failure mode to the other (flexural buck-
a, b Auxiliary terms for application of proposed methodology ling to lateral–torsional buckling, etc.) The “general method”
a,b,c,d Class indexes for buckling curves can only treat these transitions in a very superficial way, by in-
b Cross section width terpolation (not recommended by [7]) or, on the other hand,
bmax Maximum cross section width by a time-consuming specific calibration, not practical;
bmin Minimum cross section width (ii) If the method is applied to a tapered member, the question also
e0 Maximum amplitude of a member imperfection arises of how to categorize the member in terms of buckling
fy Yield strength
h Cross section height
hmax Maximum cross section height
hmin Minimum cross section height (a) Curved and tapered elements – Barajas
n(x) Distributed axial force; Airport, Madrid, Spain
nEd(x) Design distributed axial force;
tf Flange thickness
tw Web thickness
x-x Axis along the member
x0 First order failure cross-section (with h = hmin)
xc Location of the critical cross-section
y-y Cross-section axis parallel to the flanges
y (x) Displacement at a given position
y0 (x) Initial imperfection
ycr(x) Critical displacement at a given position
z-z Cross-section axis perpendicular to the flanges
α, αEC3 Imperfection factor according to EC3-1-1
αb Load multiplier which leads to the flexural buckling
resistance of the column
αcr Load multiplier which leads to the elastic critical
resistance
β Generalized imperfection factor accounting for taper-
(b) Members with polygonal centroidal axis (stairs) –
Italy pavilion, World Expo 2010 – Shanghai
ing of the member
γ Taper ratio
ε Utilization ratio at a given cross-section
εM Utilization ratio regarding the bending moment M the
at a given cross-section
εN Utilization ratio regarding the axial force N at a given
cross-section
η Generalized imperfection
η″ Curvature
ηEC3, ηuniform Generalized imperfection for the prismatic mem-
ber (considering cross-section properties at the criti-
cal position)
ηnon-uniform
Generalized imperfection for the tapered member
ηnum Generalized imperfection (numerical)
Fig. 1. Non-uniform elements.
Pictures obtained from [6)
 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74 63

Class 3 Class 2 Class 1 Table 1

Expressions from literature for calculation of critical axial force of tapered members.

Source Description

Hirt and Crisinel,Expression for equivalent inertia for the tapered column, Ieq,
(2001) [9] depending on the type of web variation. Suitable for I-shaped
cross-sections.
π 2 EI
N cr ¼ L2y;eq
Lee et. al (1972) Expression for a modification factor of the tapered member
M y,Ed [10] length, g, i.e., calculation of the equivalent length of a prismatic
column with the smallest cross-section which leads to the
same critical load. Suitable for I-shaped cross-sections.
π 2 EI
Galambos (1998) N cr ¼ Leqy;2min ; Leq ¼ g⋅L
NEd <<< Afy [11]
Petersen (1980) Design charts for extraction of a factor β to be applied to the
[12] critical load of a column with the same length and the smallest
Fig. 2. Uniform beam-column with non-uniform loading.
cross-section. Suitable for different boundary conditions and
cross-section shapes.
N cr ¼ β12 π EIL2min
2

curves as the main parameter h/b (height/width) changes con-

tinuously, see Fig. 3, in which hmin (and bmin) corresponds to
the properties of the column in the smaller extreme, and hmax
(and bmax) corresponds to the properties of the column in the K, calibrated numerically and presented in the form of an abacus, is
other extreme. Because of this, the more restrictive buckling applied to the reduction factor of a column with the smallest cross-
curve is most likely to be chosen, leading to over conservative section (see Eq. 1):
results [8].

Nb;rd;Tapered ¼ K  Nb;rd;Min ð1Þ

Finally, on one hand the General Method requires sophisticated
global FEM models but on the other hand it contains so many sim-
plifications that one must wonder if it's worth to apply it when
compared to a “full” non-linear second-order analysis of the system.
The latter is not really more complicated but more precise and
“readable” for the designer (a designer understands such things as
imperfections and internal second-order forces much better than a
choice of buckling curves and terms involving y″, see term η″cr,max
from eq. (5.9) of EC3-1-1).
Therefore it nowadays makes sense to develop simple rules for the
basic cases and to include as much knowledge as possible of the “real”
behavior of members in these rules, as it will be carried out in this
paper regarding in-plane flexural buckling of tapered columns.
Finally, some analytical formulations are available: in [14] the
equilibrium equation of a tapered column subject to flexural buckling
is derived, considering a parabolic shape for the imperfection; in [15],
the equilibrium equation is also derived, considering the eigenmode
shape. However, these expressions are not applicable for practical
verification, as adequate factors for a design rule were not calibrated
for this purpose.
3. Theoretical background for non-uniform columns
3.1. Differential equation — elastic solution

2.2. Stability verification procedures for tapered columns
As mentioned in Section 2.1, it is mainly formulae for the calcula-
tion of tapered member elastic critical forces that are available in the
Fig. 4: illustrates the equilibrium of a column segment for arbitrary
boundary conditions in its deformed configuration:
Considering the axial force as N ðxÞ ¼ Nconc þ ∫Lx nðξÞdξ, neglecting
2nd order terms and considering the internal moment given by
2

literature. Some of these are summarized in Table 1. MðxÞ ¼ −EIðxÞ ddxy2 , the differential equation is given in Eq. (2):
Nevertheless, the consideration of a critical position is still unde-


″ ″ ′ ′
fined, which, on the safe side, requires the consideration of the smal- E I ðxÞ⋅y þ N ðxÞ⋅y ¼ 0 ð2Þ
lest cross-section and as a result leads to over-conservative design.
Regarding design rules for the verification, there is not much avail- The solution of this equation leads to the elastic critical load, see
able in the literature. A design proposal for verification of tapered Eq. (3). As it is not the purpose of this paper to solve Eq. (2) analyti-
columns can be found in [13], in which an additional coefficient cally, numerical Linear Eigenvalue Analysis (LEA) will be carried out
and used to obtain the shape of the eigenmode as well as the critical
load multiplier, αcr.

Curve d Curve c 8
(h/b>2) (h/b≤2) < N ðxÞ ¼ α cr NEd ðxÞ
nðxÞ ¼ α cr nEd ðxÞ ð3Þ
:
yðxÞ ¼ ycr ðxÞ

NEd(x) is the applied axial force and αcr is the critical load multiplier,
and ycr(x) is the critical eigenmode.

hmax hmin≤hmax 3.2. Imperfect column

bmin≤bmax
bmax 3.2.1. Differential equation
Consider now an initial imperfection proportional to the shape of
Fig. 3. Change of buckling curve in a tapered member. the eigenmode (ycr(x)). Considering a similar approach to paragraph
64 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74

(a) Non-uniform column (b) Equilibrium of forces (c) Detail regarding distributed force
Nconc dy
dM
M+ dx
dx
dN dy
dQ N+ dx
Q+ dx dx
dx
B B

ξ y
n(x) y η
dx dx
n(x)
n(x)
x Q
A
N A
x
M

Fig. 4. Equilibrium of a column segment.

3.1 and assuming that the internal forces are independent of the In the above, e0 denotes the maximum amplitude of a member
imperfection, the differential equation, Eq. (2), becomes imperfection.



″ ″ ′ ′ ′
EIðxÞy þ NðxÞy þ NðxÞy0 ¼ 0 ð4Þ 3.2.2.1. Imperfection consistent with European column buckling curves
formulation. Following a similar approach as for the derivation of the
European Column Buckling Curves, the imperfection is given by
Defining N(x) = αbNEd(x), where αb is the load multiplier which
leads to the flexural buckling resistance of the column, the solution
to Eq. (4) is given by y0 ðxÞ ¼ ycr ðxÞe0 ð8Þ

αb
yðxÞ ¼ y ðxÞ ð5Þ The utilization ratio ε considering this imperfection can now be
α cr −α b 0
derived

This leads to a second order bending moment of 2 3

αb

  EI ðxÞ4 ″ ″5
⋅ ð−1Þ ycr ðxÞe0 −y0
″ αb ″ α cr −α b
MðxÞ ¼ −EIðxÞy ðxÞ ¼ −EI ðxÞ y0 ðxÞ ð6Þ α b NEd ðxÞ M ðxÞ α N ðxÞ
α cr −α b εðxÞ ¼ þ ¼ b Ed þ
N R ðxÞ M R ðxÞ NR ðxÞ MR ðxÞ
ð9Þ
Defining the utilization ratio ε as the ratio between the applied
forces and the corresponding, and considering a linear interaction be-
tween moment and axial force, the utilization ratio at each section of Considering
the column is given by
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h
i N R ðxÞ=N Ed ðxÞ αb
EIðxÞ α αb
−y0 ″ ðxÞ
ðxÞ ¼
λ ; χ ðxÞ ¼ ð10Þ
α b NEd ðxÞ M ðxÞ α N ðxÞ cr −α b α cr NR ðxÞ=NEd ðxÞ
εðxÞ ¼ þ ¼ b Ed þ ð7Þ
N R ðxÞ M R ðxÞ NR ðxÞ M R ðxÞ

As a result, considering a first yield criterion, for a certain load mul- After some manipulations and reorganizing terms, the utilization
tiplier αb, the utilization ratio attains a maximum of ε = 1 at the critical ratio ε becomes (Eq. (11)):
position of the column, xc. As only one equation is given (Eq. (7)), but
3
2
two variables are unknown (αb and xc), an iterative procedure is needed   ″  
1 NR ðxc Þ 4EIðxÞ −ycr ðxÞ 5 NR ðxÞ MR ðxc Þ
to obtain the solution. ε ðxÞ ¼ χ ðxÞ þ χ ðxÞ⋅ e
1− ααb
0
MR ðxc Þ NEd ðxÞα cr NR ðxc Þ MR ðxÞ
cr

3.2.2. Assumptions for the magnitude of the imperfection ð11Þ

As already mentioned, a similar derivation was carried out in [15]
applicable to flexural buckling in general, in which, for the magnitude
At the position x = xc, ε(xc) = 1,
of the initial imperfection, equation (5.9) of EC3-1-1 was considered.
It will be shown in this section that this assumption leads to an ex- 2
3
  ″
pression matching clause 6.3.1 of EC3-1-1 for uniform columns at χ ðx c Þ NR ðxc Þ 4EIðxc Þ: −ycr ðxc Þ 5
ε ðxc Þ ¼ 1 ¼ χ ðxc Þ þ
2 ðx Þχ ðx Þ e0 M R ðxc Þ ð12Þ
the critical position. This topic will be further discussed in this paper. 1−λ c c α cr :NEd ðxc Þ
Two cases are then considered for the proportionality factor of the
eigenmode deformed shape:
Considering
a) Imperfection consistent with the derivation of the column buck-
ling curves — the amplitude of this deflection is given by e0;  
N ðx Þ  
b) Imperfection according to equation (5.9) of EC3-1-1 (equivalent e0 R c ¼ α EC3 λ
ðx Þ−0:2 ð13Þ
c
geometric imperfection) — the amplitude of the critical mode is M R ðxc Þ
given multiplied by e0 and an additional factor. This derivation
may be found in [15]; where αEC3 = αEC3(xd).
 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74 65

Eq. (12) becomes

2
3
″
EI ðxc Þ: −y cr ðxc Þ L/1000
1  4 5
1 ¼ χ ðxc Þ þ χðxc Þ α EC3 ðxc Þ λ
ðx Þ−0:2 α cr :N Ed ðxc Þ ð14Þ

2 ðx Þχ ðx Þ
1−λ
c
c c
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
ηuniform ðxc Þ β ðxc Þ
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl
ffl} Fig. 6. Shape and magnitude of the imperfection.
ηnon−unif orm ðxc Þ
for columns if the cross-section properties at the position x = xc are
Eq. (14) is identical to the Ayrton–Perry formulation for uniform considered.
columns. It can be shown that, for prismatic columns with constant
axial force, the factor β(xc) is unity. This factor takes into account 3.2.3. Interpretation of the utilization ratio ε
the non-uniformity of the column and leads to a modification of the xc and αb are obtained as follows:
current column buckling curves, i.e. clause 6.3.1 of EC3-1-1.
• xi (0 ≤ xi ≤ L) is assumed as xc and Eq. (14) (or (18)) is solved for αb;
3.2.2.2. Imperfection according to equation (5.9) of EC3-1-1. The equiv- • After this, ε(x) in Eq. (11) (or (17)) is obtained for all values of x;
alent imperfection of equation (5.9) of EC3-1-1 is given by • If ε(xi) ≥ε(x), then xc = xi. If not, the procedure is repeated for x = xi + 1.

Ncr N :α The variation of the utilization ratio ε of a non-uniform column

y0 ðxÞ ¼ e0 ycr ðxÞ⋅ ¼ e0 ycr ðxÞ Ed″ cr ð15Þ
EIy″ cr; max E:Iy cr; max (L = 10 m) is illustrated in Fig. 5, concerning Eq. (14). For this, αcr
and y″cr are obtained numerically from a LEA analysis. The utilization
For a uniform column, the critical position is at mid-span and, there- ratio ε is divided into 2 terms — εN concerns first order forces, i.e.,
fore, y″cr,max =y″cr(L/2)=y″cr(xc). Analogously, NEd =constant=NEd(xc). axial force, and εM concerns 2nd order forces, i.e., bending moment
Eq. (15) becomes due to curvature of the member. The sum of these terms leads
" # to the total utilization ratio ε. For short members, εN is much
N Ed ðxc Þ:α cr higher compared to εM as cross-sectional resistance is more signifi-
y0 ðxÞ ¼ ycr ðxÞe0   ð16Þ
E:Iðxc Þ: −y″ cr ðxc Þ cant, and vice-versa. It can be seen that the critical position xc is lo-
cated at about 10% of the member length, close to the smallest
cross-section.
Note: The sign (−) in Eq. (16) leads to a positive value of the
Note that in Fig. 5a discontinuity can be noticed at about x = 3 m. At
imperfection.
this position, the class of the flanges in compression and bending about
Analogous to 3.2.2.1, the utilization ratio ε becomes
zz (out-of-plane buckling for this example) changes from 2 (plastic ver-
  ification) to 3 (elastic verification), which leads to a modification of the
αb ″
IIEIðxÞ y0 ðxÞ resistant moment Mr and, therefore, a discontinuity in the utilization
α N ðxÞ M ðxÞ α b NEd ðxÞ α cr −α b
ε ðxÞ ¼ b Ed þ ¼ þ ratio due to 2nd order forces. Cross-sectional class was calculated
NR ðxÞ M R ðxÞ N R ðxÞ MR ðxÞ
ð…Þ according to Clause 5.6 of EC3-1-1, in which there is a jump between
 "  #
χ ðxÞ
″
NR ðxc Þ EI ðxÞ ycr ðxÞ NEd ðxc Þ NR ðxÞ MR ðxc Þ class 2 and class 3. Recent research project SEMI-COMP overcome this
ε ðxÞ ¼ χ ðxÞ þ e

2 ðxÞχ ðxÞ 0 MR ðxc Þ EIðxc Þ y″ ðx Þ NEd ðxÞ NR ðxc Þ MR ðxÞ issue by developing new evaluation procedures for the design resis-
1−λ cr c
tance of class 3 steel cross-sections [22]. In this project, a smooth trans-
ð17Þ
action between class 2 and class 4 cross-sections is provided.

At the position x = xd, ε(xc) = 1,

4. Numerical model
 
1 N ðx Þ
εðxc Þ ¼ 1 ¼ χ ðxc Þ þ χ ðxc Þ
2 ðx Þχ ðx Þ e0 R c ð18Þ 4.1. General
1−λ c c M R ðxc Þ
|fflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl}
α ðλ

ðx Þ−0:2Þ≡η
c EC3
A finite element model was implemented using the commercial fi-
nite element package Abaqus, version 6.10 [16]. Four-node linear
shell elements (S4) with six degrees of freedom per node and finite
In this case, Eq. (18) coincides exactly with equation (6.49)
strain formulation were used.
(clause 6.3.1 of EC3-1-1) and leads to the application of clause 6.3.1
S235 steel grade was considered in the reference examples, with a
yield stress of 235 MPa (perfect elastic–plastic), a modulus of elastic-
1.2 ity of 210 GPa, and a Poisson's ratio of 0.3.
εN+εM [-] Loading is applied with reference to the plastic resistance values of
1 εN [-] the smaller cross-section.
εM [-]
0.8 y'' [m-1]
ε (x)

0.6
0.4
0.2
0 Critical
0 2 4 6 8 10 Bow
x [m]
Fig. 5. Curvature y″ and utilization ratio ε (total, due to axial force only; due to 2nd
order forces only). Fig. 7. Critical load imperfection vs. Bow imperfection.
66 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74

Table 2
Analysis of the shape of the imperfection.

Taper ratio Axial force αb,GMNIA Diff

hmax/hmin (≡bmax/bmin) (%)
Critical Bow Fig. 9. Fabrication procedure for hot-rolled tapered elements.
1 Concentrated 0.0505 0
Distributed 0.0935 0.0938 − 0.32
5. Parametric study
3 Concentrated 0.2496 0.2522 − 1.04
Distributed 0.3495 0.3635 − 4.01
5 Concentrated 0.5211 0.5462 − 4.82 5.1. Definition
Distributed 0.6454 0.7050 − 9.23
Table 3 summarizes the sub-set of cases to be compared with the
advanced numerical simulations. The case of in-plane flexural buck-
4.2. Geometrical imperfections ling of linearly web-tapered columns subject to uniform axial force
is considered. More than 350 numerical simulations with shell ele-
Regarding global imperfections, a geometrical imperfection pro- ments were carried out. Both GMNIA (Geometrical and Material
portional to the eigenmode deflection is considered with a maximum Non-linear Analysis with Imperfections) numerical simulations con-
value of L/1000 (Fig. 6). This is consistent with the values considered strained in-plane and LEA (Linear Eigenvalue Analysis) are carried
during the development of the European column buckling curves [17]: out to provide data for application of the analytical formulations
and for calibration of necessary parameters. Table 3 summarizes the
parametric study, where the Taper Ratio is defined as γ = hmax/hmin.
L
y0 ðxÞ ¼ ycr ðxÞe0 ¼ ycr ðxÞ ð19Þ
1000
5.2. Methodology

The difference between considering either bow or eigenmode im- Table 4 summarizes the alternative procedures to obtain the resis-
perfections (see Fig. 7) is analyzed in Table 2. It can be observed that tance of the tapered column:
the consideration of bow imperfections leads to an over-evaluation of The first two cases (a) were already described in Section 3.2.3. Re-
resistance with the increase of the level of taper and/or the shape of garding the other cases (b), no iteration procedure is needed because
the normal force diagram relatively to a concentrated axial force. the critical location xc is assumed to be known from the numerical
The Taper Ratio γ is defined as the ratio between the maximum model. The procedure is implemented as follows:
height and the minimum height (hmax/hmin), or the maximum width
and the minimum width (bmax/bmin). 1. Extraction of xc from GMNIA model and of the critical load multi-
As for local imperfections, these were not considered, neither in plier αcr from LEA model;
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ðxc Þ ¼ NR ðxc Þ=NEd ðxc Þ, see Eq. (10);
2. Calculation of λ
the numerical models, nor in the analytical models (effective cross- α cr
section properties). However, this will not influence these results as, 3. Calculation of the generalized imperfection ηnon-uniform(xc) (when
for the analyzed cases, the critical position is always (at the most) applicable) defined in Eq. (14) as ηnon−unif orm ðxc Þ ¼ ηuniform ðxc Þ 
 
 

ðxc Þ−0:2  EIðxc Þ:ð−y cr ðxc ÞÞ .
″
in a class 3 zone of the column. βðxc Þ ¼ α EC3 ðxc Þ λ α cr :NEd ðxc Þ

4.3. Material imperfections 4. Calculation of the reduction factor χ(xc) and finally of αb, given by
αb = χ(xc). NR(xc)/NEd(xc), see see Eq. (10).
The material imperfections, residual stress patterns corresponding
both to stocky hot-rolled (i.e. with a magnitude of 0.5fy) and welded
Table 3
cross-sections were considered. Fig. 8 shows the adopted residual
Parametric study.
stress pattern.
In Fig. 9, a possible fabrication procedure for the rolled case is il- Taper Reference cross-section Reference column qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Fabrication
ratio γ (i.e. with hmin, at x=x0)
ðx0 Þ ¼ NR ðx0 Þ=NEd procedure
slenderness λ
lustrated (cutting of the web along the length of the column). This α cr

choice allowed the direct observation of the influence of the taper 1…8 IPE 200 (h=200 mm; 0…3 Welded
by comparing buckling curves for tapered members with curves for b = 100 mm; tf = 19 mm; Hot-rolled
tw = 11 mm) (0.5 fy)
members without taper, but with otherwise the same residual stress HEB 300
distribution (Fig. 8(a)). (h=b=300 mm;
tf =19 mm; tw =11 mm)
100×100×10×10
(h=b=100 mm;
tf =tw =10 mm)
(a) Hot Rolled (h/b ≤ 1.2) (b) Welded
0.5 fy fy Table 4
Considered procedures for stability verification.
0.5 fy 0.25 fy
0.2b Method Description
(a)
Eq. (11) Solution of the equation by an iterative procedure
(a)
Eq. (17) Solution of the equation by an iterative procedure
(b)
Eq. (14) Direct application — xc is extracted numerically
0.8h

0.5 fy Eq. (18) (b)

Direct application — xc is extracted numerically
≡Eq. (14) considering β(xc) = 1
EC3-1-1 ≡Eq. (14) considering xc at the smaller cross-section and β(xc) = 1
or
≡Eq. (18) considering xc at the smaller cross-section
GMNIA –
Fig. 8. Residual stresses: + Tension and − Compression.
 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74 67

5.3. Results

5.3.1. Accuracy of the analytical model

Fig. 11 illustrates the numerical results from GMNIA analyses
against results from Eq. (11) for different Taper Ratios γ, regarding
xc the maximum load factor αb and the relative critical position xc/L. A
column consisting of the hot-rolled cross-section 100 × 100 × 10 × 10
Fig. 10. Critical position according to GMNIA analysis.
defined in Table 3 is chosen for illustration. Although differences
of − 5% (unsafe) to 7% (safe) are noticed, it can be seen that the ana-
Finally, concerning nonlinear numerical calculations, the maxi- lytical model characterizes the behavior of the tapered column
mum load factor of GMNIA analysis corresponds to αb load multiplier. well when compared to the numerical model. This was also found
The critical position xc is also extracted from the numerical model for other cases of Table 3 (not shown in Fig. 11). It is also noticeable
corresponding to the element with the maximum strain at the maxi- an increase of up to 20% in terms of resistance with the increase of ta-
mum load factor, αb, see Fig. 10. pering at a slenderness range of λ
ðx0 Þ ¼ 0:5 to λ

ðx0 Þ ¼ 1, which shows
Results are represented relatively to the location of the smallest the relevance of Eq. (11). Eq. (11) was solved considering an imper-
cross-section, x0. Because NR(x0) = NEd, Eq. (10) becomes: fection factor αEC3 = 0.34 (curve b of EC3-1-1), in agreement with
the adopted residual stresses of 0.5fy. Finally, in Fig. 11(a), the Taper
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Ratio of γ = 1, i.e., prismatic column, is shown for comparison. It is

ðx Þ ¼ NR ðx0 Þ=NEd ðx0 Þ 1 expected that the relative critical location is located at mid-span
λ 0 ¼ pffiffiffiffiffiffiffi ; χ ðx0 Þ
α cr α cr where the curvature is maximum and therefore xc/L = 0.5 (see
αb Fig. 11(a.2)). Moreover, Eq. (11) should give the same results as the
¼ ð20Þ
N R ðx0 Þ=N Ed ðx0 Þ ¼ 1 →χ ðx0 Þ≡α b EC3 curve b. This is visible in Fig. 11(a.1).

(a.1) γ=1 (a.2) γ=1

1.1 0.6
EC3 Curve b
Euler 0.5
0.9 γ=1 | GMNIA
γ=1 | Model 0.4
0.7
αb xc/L 0.3
0.5
0.2
0.3 γ=1 | GMNIA
0.1
γ=1 | Model
0.1 0.0
0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5
λy (x0) λy (x0)

(b.1) γ=2.5 (b.2) γ=2.5

1.1 0.6
EC3 Curve b
Euler 0.5 γ=2.5 | GMNIA
0.9 γ=2.5 | GMNIA γ=2.5 | Model
γ=2.5 | Model 0.4
0.7
αb xc/L 0.3
0.5
0.2
0.3 0.1
0.1 0.0
0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
λy (x0) λy (x0)

(c.1) γ=5 (c.2) γ=5

1.1
EC3 Curve b
Euler 0.5 γ=5.0 | GMNIA
0.9 γ=5.0 | Model
γ=5.0 | GMNIA
γ=5.0 | Model 0.4
0.7
αb xc/L 0.3
0.5
0.2
0.3 0.1
0.1 0.0
0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
λy (x0) λy (x0)

ðx0 Þ; (.2) Critical position xc/L against slenderness λ
Fig. 11. Analytical derivation Eq. (11) against numerical calculations GMNIA. (.1) Resistance αb ≡ χ(x0) against slenderness λ
ðx0 Þ.
68 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74

(a) 100x100x10x10 Hot-Rolled (b) 100x100x10x10 Welded

1.0 1.0

0.8 0.8

αb αb
0.6 0.6

0.4 γ=1 γ=1.5 0.4 γ=1 γ=2

γ=2 γ=3 γ=3 γ=5
γ=6 EC3-b γ=8 EC3-b
Euler Euler
0.2 0.2
0 0.5 1 1.5 0 0.5 1 1.5
λy (x0) λy (x0)

(c)IPE200 Welded (d) HEB300 Welded

1.0 1.0

0.8 0.8

αb αb
0.6 0.6

γ=1 γ=1.25 γ =1 γ =1.33

0.4 γ=1.5 γ=2 0.4 γ =2 γ =3
γ=3 γ=5 γ =5 EC3-b
EC3-b Euler Euler
0.2 0.2
0 0.5 1 1.5 0 0.5 1 1.5
λy (x0) λy (x0)

ðx0 Þ.
Fig. 12. Numerical calculations GMNIA organized by Taper Ratio. Resistance αb ≡ χ(x0) against slenderness λ

5.3.2. Influence of the taper ratio, γ = hmax/hmin lower and, as a consequence the resistance of the tapered member
Fig. 12 illustrates GMNIA results of resistance against slender- becomes higher.
ness (based on Npl of the smallest cross-section) organized by


Taper Ratio. Curve b of EC3-1-1 is shown for comparison.
EIðxc Þ: −y″ cr ðxc Þ
Note that, for the welded cross-section cases, the numerical βðxc Þ ¼ ð21Þ
curve corresponding to the uniform element (γ = 1) shows devia- α cr :NEd ðxc Þ
tions that fall below the code curve results for the relevant slen-
derness range up to 1. This will be discussed in Section 6 — Fig. 13 illustrates the influence of these two parameters for a member
Verification procedure. with an initial cross-section of 100×100×10×10 (hot-rolled) and γ=4
(hmax =400 mm). Table 5 shows results for the case of λ
y ðx0 Þ ¼ 0:74. In
A smooth increase in the resistance with the increase of Taper
Ratio γ along all slenderness ranges can be observed in all cases of
Fig. 12. It also shows to be less significant for higher levels of Taper
Ratio. 1.0

5.3.3. Analysis of the critical position xc and of an additional factor β(xc)

to the generalized imperfection ηuniform
The importance of identification of the critical location has already 0.8
been discussed. Nowadays, there is no straight-forward procedure to
αb
obtain this location. Therefore, most designers in practice will use the No xc | No β
smallest cross-section properties for verification according to clause xc | No β
0.6 xc | β
6.3.1 of EC3-1-1.
EC3-b
Moreover, an additional factor β(xc) derived in Section 3.2.2.1 GMNIA
and given in Eq. (17) characterizes the increase of resistance of Euler
the tapered member relatively to the uniform member. This factor 0.4
0 0.5 1 1.5
attains a limit for the case of uniform members, reaching unity
λy(x0)
for those cases. When associated to the generalized imperfection
of the uniform member ηuniform to give a generalized imperfection Fig. 13. Influence of the critical position and of the imperfection in the resistance of the
of the tapered member ηnon-uniform, see Eq. (14), the latter becomes tapered column.
 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74 69

Table 5 1
Influence of the critical position and of the imperfection in the resistance of the tapered

y ðx0 Þ ¼ 0:74; αcr = 1.85).
column (λ 0.9

Case xc/L β(xc)
y ðxc Þ

λ χ(xc) αb = χ(xc)NRk/NEd Diff (%)
0.8

No xc | No β 0 1 0.735 0.764 0.764 17.5

0.7
xc | No β 0.10 (GMNIA) 1 0.773 0.741 0.820 11.5 0.6 γ= 1.00
xc | β 0.10 (GMNIA) 0.48 0.773 0.842 0.932 − 0.6
GMNIA 0.10 – – – 0.926 –
β(xc) 0.5 1.50
2.00
0.4
2.50
0.3 3.00
0.2 4.00
order to obtain resistance for the cases considering xc, the numerical po- 5.00
0.1
sition was considered. Moreover, to calculate β(xc), ycr is extracted 6.00
from LEA analysis. It can be seen that factor β has a great influence 0
0 0.5 1 1.5 2 2.5 3 3.5
in the resistance of the column. Regarding the case in which xc is con- λy(xc)
sidered with current EC3 imperfection (no β), note that the first 3
cases coincide with current EC3 buckling curve b. This happens because
y ðxc Þ, all cases.
Fig. 15. Additional imperfection factor β(xc) against the relative slenderness λ
at this slenderness range and regarding this taper ratio, the 2nd order
failure cross-section (or xc) is the same as the 1st order failure cross- imperfection should only be dependent on the member length [18].
section (smaller end). Table 5 shows an imperfection decrease of Moreover, for the calibration of EC3 imperfection factors for columns,
more than 50% (β(xc) =0.48) for the analyzed case of λ
y ðx0 Þ ¼ 0:74. this magnitude was given by e0 = L/1000 (and additional residual stres-
Finally, it can also be observed that the relative critical location xc/L ses for the material imperfections). The same approach is considered
and the additional imperfection factor β(xc) are independent of the in this study. Both Fig. 16(a) and (b) show a better agreement with
fabrication process or of the initial cross-section proportions, see the EC3 consistent approach regarding Eq. (11).
Figs. 14 and 15, computed for all the analyzed cases. Note that concerning Fig. 16(b), GMNIA is also illustrated in terms of
the reduction factor χ(xc), in which xc is obtained from the numerical
model. Finally, Fig. 17(a) and (b) respectively illustrate resistance αb
5.3.4. Influence of the function for the magnitude of the imperfection and relative critical location xc/L regarding all Taper Ratios of the analyzed
In Section 3.2.2 Assumptions for the magnitude of the imperfection, cross-section 100×100×10×10 (hot-rolled). A higher spread is noticed
two cases are considered. Results have been shown regarding the am- for Eq. (17).
plitude of the imperfection given by e0 (Section 3.2.2.1), i.e., consistent
with the derivation of the column buckling curves of EC3-1-1. Fig. 16 6. Design methodology
compares the solution of Eqs. (11) and (17), in which for the latter,
Eq. (16) (equation (5.9) of EC3-1-1) is considered for the imperfection. 6.1. Introduction
This derivation is also given in [15]. Two representations of resistance
are considered and illustrated concerning a Taper Ratio of γ = 4 and Considering the developed analytical formulation and the numer-
the reference cross-section 100× 100 × 10× 10 (hot-rolled): Fig. 16(a) ical calculations, a verification procedure for the stability of tapered
illustrates the reduction factor λ
ðx0 Þas a function of the reduction factor
columns subject to in-plane bending is now proposed.
χ(x0) ≡ αb, and, therefore, resistance can be directly compared; In a first step, regarding the imperfection factor for uniform
Fig. 16(b) illustrates the reduction factor λ
ðxc Þas a function of the reduc-
welded cross-sections it was noticed that, for I-sections, the imperfec-
tion factor χ(xc) — it is stated in [15] that when equation (5.9) of EC3-1- tion factor α = 0.34 provides unsafe results for slenderness up to ap-
1 (Eq. (16)) is considered for the imperfection, results of the reduction proximately 1 (differences of 8% were observed, see Fig. 12(b), (c)
factor χ(xc) coincide with current buckling curves for columns (see also and (d)). Because this proposal has, as a reference limit, the case of
Fig. 16(b)) and that good agreement is achieved with numerical uniform members (γ = 1), it was decided to calibrate new imperfec-
models. This is to be expected if the imperfections considered in the nu- tion factors for welded cross-sections for that purpose.
merical models are also obtained from Eq. (16). However, and as al- In a second step, the development of a verification procedure for
ready mentioned in Section 2.2, the magnitude of the geometrical tapered columns is done. Here, expressions for the critical location
xc and the additional imperfection factor β(xc) are calibrated against
numerical results shown in Figs. 14 and 15.
0.5
0.45 γ= 6.2. Imperfection factors for flexural buckling of uniform welded columns
0.4 1.00
Ayrton–Perry formulation for uniform columns is given by
0.35 1.17
 
1.33 1 N  
0.3 1¼χþχ e0 R α λ
ðx Þ−0:2 ≡η →ð…Þ→η ð22Þ
1.50

1−λ χ
2 MR c EC3 EC3
xc/L 0.25
 

1

2.00 ¼α λ
ðx Þ−0:2 ¼ 1−λ
χ 1−
2
0.2 c
χ
3.00
0.15
5.00
0.1 The generalized imperfection ηEC3 for in-plane buckling of uniform
8.00
0.05 columns is given by Eq. (23), for a flange thickness tf ≤ 100 mm
   
0 ηEC3 ¼ α λ
ðx Þ−0:2 ¼ 0:34 λ
ðx Þ−0:2 ð23Þ
0 0.5 1 1.5 2 2.5 3 3.5 c c

λy(x0)
Fig. 18(a) illustrates the generalized imperfection of EC3-1-1 ηEC3

y ðx0 Þ, all cases.
Fig. 14. Relative critical position xc/L against the relative slenderness λ compared to the generalized imperfection ηnum of about 100
70 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74

(a) Resistance αb≡χ(x0) against slenderness λ (x0) (b) Resistance χ(xc) against slenderness λ (xc)
1.1 1.1
EC3 -b EC3 -b
Euler Euler
0.9 0.9
GMNIA GMNIA
Eq. (11) Eq. (11)
0.7 0.7
χ(x0) Eq. (17)
χ(xc) Eq. (17)
≡αb
0.5 0.5

0.3 0.3

0.1 0.1
0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5
λy (x0) λy (xc)

Fig. 16. Influence of imperfection magnitude — buckling curve representation.

(a) Resistance αb (b) Critical location xc/L

1.2 0.6
Eq. (18) Eq. (17)
1 0.5
Eq. (12) Eq. (11)

xc/L(Equation)
αb (Equation)

0.8 0.4

0.6 0.3

0.4 0.2

0.2 0.1

0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 0.1 0.2 0.3 0.4 0.5 0.6
αE (GMNIA) xc/L (GMNIA)
Fig. 17. Influence of imperfection magnitude — 100 × 100 × 10 × 10 (hot rolled), all Taper Ratios.

numerical calculations covering a range of uniform columns with differ- Fig. 18(a) shows the difference, on the unsafe side, in considering
ent h/b ratios varying from 0.95 (HEA200) to 2.5 (IPE500) and slender- for the imperfection factor α the value of 0.34. A value of α = 0.45 was
ness varying from λ
y ¼ 0:1 to λ

y ¼ 2:0. The value ηnum is calculated shown to fit the reduction factor χy very accurately up to slenderness
according to Eq. (24), see also [19], in which χ is extracted numerically of 1. However, in order not to get too conservative for slenderness
and corresponds to the maximum load factor of GMNIA calculation, αb: above 1 and to take into account the buckling behavior of columns
with a welded residual stress pattern for that slenderness range,
 


a cut-off of η ¼ α λ−0:2 ≤0:27was also shown to be adequate. If


1


2χ
ηnum ¼ 1−λ 1− ð24Þ the cut-off of 0.27 is applied, for higher slenderness of about
num
χ num
y ¼ 1:5, imperfection becomes unsafe again. However, for high
λ

(a) Generalized imperfection η against slenderness λy (b) Resistance χy against slenderness λy

0.5
1
0.45
0.9
0.4
0.35 0.8
0.3
0.7
ηy 0.25 χy
0.2 0.6
0.15 0.5 α=0.34 (EC3)
0.1 α=0.34 (EC3)
α=0.45 (Best fit)
α=0.45 (Best fit) 0.4 α=0.45 + Cut-off
0.05
Cut-off Euler
0 0.3
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5
λy λy

Fig. 18. Generalized imperfection of in-plane flexural buckling of welded columns.

 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74 71

slenderness range, the column is not so sensitive to the imperfection 1

level and resistance converges to the Euler load. 21(b) illustrates the 0.9

y.
reduction factor χy against the relative slenderness λ
0.8
0.7
6.3. Procedure γ= 1.00
0.6 1.50
2.00
Eq. (12) was shown to follow accurately the buckling behavior of a β(xc) 0.5 2.50
3.00
tapered column. However, the application of this expression is not 0.4 4.00
straight forward: 5.00
0.3 6.00
1.00 (Equ)
1.50 (Equ)
• The critical location xc is needed throughout the application of 0.2 2.00 (Equ)
Eq. (12) — for this an iterative procedure is needed; 0.1 2.50 (Equ)
3.00 (Equ)
• Once xc is known, the additional imperfection factor β(xc) can be 4.00 (Equ)
0 5.00 (Equ)
calculated. However, to obtain it, the function for the critical curva- 0 0.2 0.5 1 1.5 2 2.5 3 3.5 (Equ)
6.00
ture is needed — this is not a direct procedure. λy(xc)

Assuming that the critical load multiplier, αcr, is obtained from a Fig. 20. Fitting elliptical expression for the additional imperfection factor β(xc).
numerical analysis, LEA, expressions regarding xc and β(xc) are still
needed for the direct calculation of resistance. Elliptical expressions
were shown to give good approximation for both these parameters.
Fitting equations for xc and β(xc) are illustrated in Figs. 19 and 20 re- 7. Example
spectively. Corresponding expressions are shown in Fig. 21, which il-
lustrates the complete procedure for in-plane stability verification of 7.1. Introduction
tapered columns. Note that, for higher Taper Ratios, β could be higher.
However, for safety reasons concerning the resistance multiplier αb, A tapered column composed of a IPE200 welded cross-section in the
the limit of β = 1 for all tapered ratios was chosen. smallest end with a linearly varying height and a Taper Ratio of γ =hmax/
Finally, Fig. 21 illustrates the steps to be followed. Firstly, the crit- hmin =3 is now analyzed (Fig. 23). The applied load is given by the plastic
ical position xc is determined based on the reference relative slender- axial force at the smallest end, i.e. NEd = 500 kN for a yield stress of
ness of the smallest cross-section. αcr shall be calculated numerically. fy =235 MPa. The column has a length of L = 12.9 m. In-plane buckling
Note that from this step, geometrical properties of xc are considered, resistance is calculated using several methods.
including slenderness calculation for the determined position. Imper-
fection can now be calculated by combining the imperfection effects 7.2. Elastic critical analysis
of the uniform member (ηuniform) and of the tapered member (β).
With this, the reduction factor at xc is determined and the verification A numerical linear eigenvalue analysis LEA attains a critical load
is finally made. multiplier of αcr = 1.8501. For comparison, the critical load is also cal-
culated by some of the methods described in Table 1. A negative differ-
ence illustrates a higher value of the critical load obtained in the
6.4. Comparison with numerical results
literature relatively to the numerical value. Results are summarized
in Table 6:
Fig. 22 illustrates the resistance of the numerical results χ(x0) ≡ αb
as a function of the relative slenderness λ
ðx0 Þ, concerning GMNIA
7.3. Verification of stability
analysis as well as the proposed formulation. The current EC3 curve
for uniform members that would be applied is also illustrated (i.e.
In Section 7.3.1 the proposed verification procedure is applied. The
α = 0.34; β = 1; and considering xc as the minimum cross-section as
application of other methodologies is summarized in Section 7.3.2. A
no guidelines exist at the moment). Good agreement is noted with
numerical GMNIA analysis leads to a maximum load factor of
the proposed methodology.
αb = 1.004.

7.3.1. Application of the proposed method

0.5
7.3.1.1. Calculation of slenderness at x = x0 (smaller cross-section). •
0.45 γ=
1.00
0.4 1.17 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1.33 NPl ðx0 Þ=NEd 640:3=500
1.50
ðx Þ ¼
λ ¼ ¼ 0:832
0.35 0
2.00 α cr 1:8501
2.50
0.3 3.00
4.00
xc/L 0.25 5.00
6.00
0.2 8.00 7.3.1.2. Critical cross-section relative position, xc/L.
1.00 (Equ)
0.15 1.17 (Equ)
1.33 (Equ) 0:6 0:6 1 1
• a ¼ 1:5 þ ¼ 1:5 þ ¼ 1:8 and b ¼ ¼ ¼ 0:25
0.1 1.50 (Equ)
2.00 (Equ)
γ−1 3−1 γþ1 3þ1
0.05 2.50 (Equ)
3.00 (Equ)
0 4.00 (Equ)
5.00 (Equ)
2−a ¼ 0:2≤ λ
ðx Þ ¼ 0:832≤2⇒
0 0.2 0.5 1 1.5 2 2.5 3 3.5 (Equ)
6.00
• sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
8.00 (Equ)  
λy(x0) λ
ðx Þ−2 2
0 ð0:832−2Þ2
xc =L ¼ b 1− 2
¼ 0:25 1− ¼ 0:190
a 1:82
Fig. 19. Fitting elliptical expression for the critical position, xc.
72 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74

Required data
N Rk ( x0 ) / N Ed 0.6 1
λ ( x0 ) = ; γ = hmax hmin ; a = 1.5 + ; b=
α cr γ −1 γ +1

Critical Position, x c

0, λ ( x0 ) < 2 − a or λ ( x0 ) ≤ 0.2

xc / L = b 1 −
(λ ( x ) − 2)
0
2
, 2 − a ≤ λ ( x0 ) ≤ 2
a2
b, λ ( x0 ) > 2

Slenderness at x c
N Rk ( xc ) / N Ed
λ ( xc ) =
α cr

Account for Imperfection

0, λ ( xc ) < 2 − a or λ ( xc ) ≤ 0.2
(
ηUniform = α Uniform λ ( xc ) − 0.2 )
β ( xc ) = 1−
(λ ( x ) − 2)
c
2
, 2 − a ≤ λ ( xc ) ≤ 2
Hot − Rolled Welded
a2 where α Uniform 0.34 0.45
1, λ ( xc ) > 2 ηUniform − ≤ 0.27

Reduction factor

χ ( xc ) =
1
≤ 1, with
( 2
φ ( x c ) = 0 . 5 1 + η ( xc ) + λ ( xc ) )
φ ( xc ) + φ 2 ( x c ) − λ ( xc )
2
η ( xc ) = ηUniform β ( xc )

Verification
N Ed ≤ χ ( xc ) N Rk ( xc ) ≤ N Rk ( x0 )
Fig. 21. Design proposal.

7.3.1.3. Calculation of slenderness at x = xc. 7.3.1.6. Verification.

NPl ðxc Þ ¼ Aðxc Þf y ¼ 740:5 kN⇒

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi • N b;Rd ¼ χ ðxc ÞNPl ðxc Þ ¼ 0:681  740:5 ¼ 504:2kN≤NPl ðx0 Þ
• NPl ðxc Þ=N Ed 740:5=500


λ ðxc Þ ¼ ¼ ¼ 0:895
α cr 1:8501 Nb;Rd ≥NEd →504:2 > 500→Design check verified!
• α ¼ N =N ¼ 504:2=500 ¼ 1:008ðGMNIA; α ¼ 1:004Þ
b b;Rd Ed b

7.3.1.4. Determination of imperfection, η.

7.3.2. Summary of results
Results are summarized in Tables 7 and 8.
• Additional Imperfection factor β(xc) Firstly, from results of Section 7.3.1, the resistance calculated
according to the proposed methodology of Section 6.3 is practically
coincident with the GMNIA resistance (0.4% of difference).
2−a ¼ 0:5≤
ðx Þ ¼ 0:895≤2⇒
λ Table 7 summarizes results considering the smallest cross-section for
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

c
 ffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2
λ
ðx Þ−2 2 ð0:895−2Þ verification. Note that, in this case, the case corresponding to α=0.34
c
βðxc Þ ¼ 1− ¼ 1− ¼ 0:789 leads to a smaller difference (10%) because current buckling curves for
a2 1:82
welded columns lead to unsafe results in this slenderness range. There-
fore, the comparable correct difference is 15%, which corresponds to pro-
• ηnon−unif orm ¼ ηuniform  βðxc Þ ¼ 0:27  0:789 ¼ 0:213 posed buckling curve with α=0.45 considering the cut-off of η≤0.27.
Finally, considering [11] or [9] for αcr calculation leads to different
relative slenderness λ
y ðx0 Þ. Considering [9] with the proposed verifi-
7.3.1.5. Reduction factor at x = xc.


 cation procedure leads to unsafe level of resistance.

2 ðx Þ ¼ 0:5 1 þ 0:213 þ 0:8952 ¼ 1:007
• ϕðxc Þ ¼ 0:5 1 þ η þ λ c
8. Conclusions
• χ ðxc Þ ¼ 1 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ϕðxc Þ þ ϕ2 ðxc Þ−λ
2 ðx Þ 1:007 þ 1:0072 −0:8952
c In this paper an analytical derivation of non-prismatic columns was
¼ 0:681≤1 carried out and compared against numerical simulations. It was shown
 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74 73

(a) 100x100x10x10 | Hot-Rolled | γ =6 (b) 100x100x10x10 | Welded | γ=4

1.0 1.0

0.8 0.8

0.6 0.6
αb αb
0.4 EC3-b 0.4 EC3-b
Euler Euler
0.2 GMNIA 0.2 GMNIA
Proposal Proposal
0.0 0.0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
λy(x0) λy(x0)

(c) IPE200 | Welded | γ=3 (d) HEB300 | Welded | γ=2

1.0 1.0
Example Section 7
0.8 0.8

0.6 0.6
αb αb
0.4 EC3-b 0.4 EC3-b
Euler Euler
0.2 GMNIA 0.2 GMNIA
Proposal Proposal
0.0 0.0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
λy(x0) λy(x0)

Fig. 22. Resistance αb against λ
y ðx0 Þ. Evaluation of the proposed method.

(a) In-plane stress distribution (perspective view) (b) In-plane buckling shape (front view)

Fig. 23. Analyzed tapered column.

that, concerning non-uniform columns, Eurocode rules needed to be • The column design formula had to be amended by an additional
adapted in the following aspects: factor β, which specifically takes into the second-order behavior
of tapered columns.
• A practical approach for the determination of the design position
• The current imperfection factor of clause 6.3.1 for welded sections
needed to be developed;
needed to be modified and re-calibrated.

For this purpose, a wide parametric study of more than 350 LEA
and in-plane GMNIA simulations was carried out regarding linearly
Table 6
web tapered columns with constant axial force subject to in-plane
Calculation of critical axial force of the tapered column.
buckling.
Method Approach αcr = Ncr,tapered/NEd Diff (%) After that, a proposal for the stability verification of these tapered col-
Hirt and Crisinel [9] Iy,eq (…) = 33% Iy,max 2.0219 − 9.3 umns was presented. It was noticed that, most of all, the consideration
Galambos [11] Leq (…) = 52% L 1.7079 7.7 of the most stressed position is necessary in order not to achieve over-
LEA Numerical 1.8501 –

Table 8
Table 7 Results — proposed method applied with other formulae for αcr (from literature).
Results — xc = x0 (minimum cross-section), current procedure EC3 (β = 1).
Method for αcr
ðx0 Þ ¼ 1=pffiffiffiffiffiffiffi
λ α cr αb Diff (%)
 
α (imperfection) Cutt-off: α λ
ðxc Þ−0:2 ≤0:27 αb Diff (%)
Hirt and Crisinel [9] 0.796 1.037 − 3.3
0.34 x 0.903 10.1 Galambos [11] 0.866 0.981 2.3
0.45 ✓ 0.853 15.0 LEA 0.832 1.008 − 0.4
GMNIA 1.004 – GMNIA 1.004 –
74 L. Marques et al. / Journal of Constructional Steel Research 72 (2012) 61–74
conservative levels of resistance. The above-mentioned factor β was de-
veloped based on the prior analytical formulation and calibrated with
numerical results. Finally, a new generalized imperfection for welded
uniform columns was also calibrated in order to obtain improved results
for the tapered cases. The next step is to perform a reliability analysis
of the proposal and determine γM1 [20,21]. Moreover, the same topic
will be analyzed for other tapering shapes and loading.
Acknowledgment
Financial support from the Portuguese Ministry of Science and
Higher Education (Ministério da Ciência e Ensino Superior) under con-
tract grant SFRH/BD/37866/2007 is gratefully acknowledged.
References
[1] CEN. Eurocode 3, EN-1993-1-1:2005, Eurocode 3: design of steel structures — Part
1–1: general rules and rules for buildings. Brussels, Belgium: European Committee
for Standardization; 2005.
[2] Müller C (2003). “Zum Nachweis ebener Tragwerke aus Stahl gegen seitliches
Ausweichen”, PhD Thesis, RWTH Aachen, Germany.
[3] Simões da Silva L, Marques L, Rebelo C. Numerical validation of the General Method
in EC3-1-1 for prismatic members. J Constr Steel Res 2010;66(4):575–90.
[4] Simões da Silva L, Gervásio H, Simões R. Design of steel structures. ECCS Eurocode
Design Manuals. ECCS Press and Ernst & Sohn; 2010.
[5] Simões da Silva L, Gervásio H. Manual de Dimensionamento de Estruturas Metálicas:
Métodos Avançados. Coimbra, Portugal: cmm Press; 2007.
[6] www.steelconstruct.com, in June 16th, 2010.
[7] ECCS TC8. “Resolution of ECCS/TC8 with respect to the general method in EN
1993-1-1”, ECCS TC8 Stability; 2006.
[8] Marques L, Simões da Silva L, Rebelo C. Application of the general method for the
evaluation of the stability resistance of non-uniform members. Proceedings of
ICASS, Hong Kong, 16–18 December; 2009.
[9] Hirt MA, Crisinel M. Charpentes Métaliques — Conception et Dimensionnement des
Halles et Bâtiments. Traité de Génie Civil, vol. 11. Lausanne: Press Polytechniques
et Universitaires Romandes; 2001.
[10] Lee GC, Morrell ML, Ketter RL. Design of tapered members. Weld Res Counc Bull
June 1972(173):1–32.
[11] Galambos TV, editor. Guide to Stability Design Criteria for Metal Structures. Fifth
Edition. John Wiley & Sons Inc.; 1998
[12] Petersen C. Stahlbau. Wiesbaden: Vieweg Verlag; 1993.
[13] Baptista AM, Muzeau JP. Design of tapered compression members according to
Eurocode 3. J Constr Steel Res 1998;46(1–3):146–8.
[14] Raftoyiannis I, Ermopoulos J. Stability of tapered and stepped steel columns with
initial imperfections. Eng Struct 2005;27(2005):1248–57.
[15] Naumes (2009). “Biegeknicken und Biegedrillknicken von Stäben und Stabsystemen
auf einheitlicher Grundlage”, PhD thesis, RWTH Aachen, Germany.
[16] Abaqus. v.6.10, Dassault Systems/Simulia, Providence, RI, USA; 2010.
[17] Beer H, Schulz G. Die Traglast des planmä βig mittig gedrückten Stabes mit
Imperfektionen. VDI-Zeitschrift 1969;21:1537–41 1683–1687, 1767–1772.
[18] Greiner R, Taras A. New design curves for LT and TF buckling with consistent
derivation and code-conform formulation. Steel Constr 2010;3(3):176–86.
[19] Taras A, Greiner R. New design curves for lateral-torsional buckling — proposal
based on a consistent derivation. J Constr Steel Res 2010;66:648–63.
[20] Rebelo C, Lopes N, Simões da Silva L, Nethercot D, Vila Real P. Statistical evaluation
of the lateral–torsional buckling resistance of steel I-beams — Part 1: Variability of
the Eurocode 3 design model. J Constr Steel Res 2008;65(4):818–31.
[21] Simões da Silva L, Rebelo C, Nethercot D, Marques L, Simões R, Vila Real P. Statisti-
cal evaluation of the lateral–torsional buckling resistance of steel I-beams — Part 2:
variability of steel properties. J Constr Steel Res 2008;65(4):832–49.
[22] Greiner R, Kettler M, Lechner A, Jaspart J-P, Weynand K, Ziller C, Örder R. SEMI-
COMP+: valorisation action of plastic member capacity of semi-compact steel
sections — a more economic design, RFS2-CT-2010-00023. Background Documen-
tation, Research Programme of the Research Fund for Coal and Steel — RTD; 2011.