Anda di halaman 1dari 15

Journal of Constructional Steel Research 61 (2005) 12261240

www.elsevier.com/locate/jcsr
On the incorporation of equivalent member
imperfections in the in-plane design of steel frames
Rodrigo Gonalves
a
, Dinar Camotim
b,
a
ESTB, Polytechnic Institute of Setbal, R. Stinville 14, 2830-114 Barreiro, Portugal
b
IST, Civil Engineering Department, Technical University of Lisbon, Av. Rovisco Pais,
1049-001 Lisbon, Portugal
Received 6 September 2004; accepted 27 January 2005
Abstract
This work deals with the incorporation of equivalent member imperfections in the global
analysis of steel frames and, in particular, is intended to clarify the Eurocode 3 (EC3) provisions
involved in such procedure. In fact, these provisions stem from the well-known European column
buckling curves, which means that they are based on the behaviour of simply supported isolated
members under uniform compression (columns). First, one addresses the geometrically non-
linear behaviour of isolated columns displaying arbitrary support conditions and different initial
geometrical congurations. Then, the results obtained are used to propose a systematic and rational
method to evaluate the appropriate equivalent initial imperfections that need to be incorporated in
the second-order global elastic analysis of a frame or isolated compressed member. This method
(i) is fully consistent with the EC3 column buckling curves and (ii) adopts critical buckling
mode shape initial imperfections with amplitudes determined by means of closed-form analytical
expressions. In order to enable a better grasp of the concepts involved and also to illustrate the
capabilities of the proposed methodology, several numerical examples are presented and discussed
throughout the paper.
2005 Elsevier Ltd. All rights reserved.
Keywords: Steel structures; Compressed members; Member imperfections; Equivalent geometrical imperfections;
Column buckling curves; Eurocode 3

Corresponding author. Tel.: +351 1 8418403; fax: +351 1 8497650.


E-mail address: dcamotim@civil.ist.utl.pt (D. Camotim).
0143-974X/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcsr.2005.01.006
R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240 1227
1. Introduction
The design and safety checking of steel compressed members (columns or
beamcolumns) must always allow for the detrimental effect of member imperfections,
which may be of a geometrical (lack of straightness, plate thickness variation, etc.) and/or
material (residual stresses, yield stress variation, etc.) nature. In the particular case of
Eurocode 3 (EC3) [3,4], member imperfections can be allowed for either
(i) indirectly, by using the interaction formulae (e.g., the exural buckling resistance of
uniformcolumns is evaluated by means of the well-known European column buckling
curves [6]), or
(ii) directly, by including member equivalent initial bow imperfections in the second-
order global analysis, which amounts to increasing the internal forces and moments to
take into account the effect of member imperfections.
1
The concept of equivalent imperfection can be traced back as far as the nineteenth
century when Ayrton and Perry [1] specied the initial curvatures and end eccentricities
that should be considered in column design. Much more recently, Rondal and Maquoi [10]
introduced the concept of equivalent (initial) bow imperfection, which is embedded
in the current Eurocode 3 (EC3) buckling curves and corresponds to the adoption of a
sinusoidal initial imperfection with an amplitude equal to
e
0
= (

0.2)
W
A
, (1)
where is an imperfection factor dening a particular buckling curve, W is the elastic
section modulus, A is the cross-section area and

=
_
Af
y
/N
cr
is the column normalised
slenderness ( f
y
is the characteristic yield stress and N
cr
is the elastic critical axial
force). This equivalent bow imperfection amplitude is obtained from the PerryRobertson
formula
2
[10], where the values were calibrated on the basis of a large number of
column test results. However, there is no clear information available on the extension of
this equivalent imperfection concept to more general cases, such as (i) isolated columns
with arbitrary support conditions or (ii) columns integrated in frames, a fact leading to a
fair amount of ambiguity and also to rather frequent misinterpretations of the EC3 rules. In
order to clarify this matter and contribute to overcome the above problems, the authors
recently proposed a general and consistent method to evaluate appropriate equivalent
critical mode shape initial imperfections complying with the EC3 provisions, which
is based on using closed-form analytical expressions to evaluate their amplitudes [7,9].
In fact, one may say that the method included in the recent stage 49 draft of EC3-prEN
(preliminary version of the upcoming European Norm) [4] is quite similar to the one
developed earlier by the authors.
1
As discussed further ahead, the equivalent imperfection prescribed by EC3-ENV (European Pre-Norm) [3]
could not be applied, in a straightforward fashion, to other than isolated and simply supported compressed
members. This limitation has been overcome by the inclusion of new provisions in EC3-prEN [4].
2
Based on the second-order elastic analysis of simply supported columns with sinusoidal initial imperfections.
It is assumed that collapse corresponds to the onset of yielding at the most compressed bre (mid-span cross-
section).
1228 R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240
Fig. 1. (a) Initial (P = 0) and (b) acquired (P = 0) deformed congurations of an isolated column with arbitrary
end support conditions.
The objective of this paper is to provide a contribution to clarify the EC3-prEN
provisions concerning the incorporation of equivalent member imperfections in the elastic
second-order global analysis of (i) isolated columns with arbitrary end support conditions
and (ii) columns integrated in frames. First, one presents and discusses analytical results
dealing with the geometrically non-linear behaviour of initially crooked isolated uniform
columns. Different initial geometrical imperfection shapes are considered and their
inuence on the column behaviour and strength is assessed. The above analytical results
are then used (i) to clarify the equivalent initial bow imperfection concept presented in
both EC3-ENV and EC3-prEN and (ii) to propose a coherent extension of this concept
for members integrated in frames. This extension (i) is consistent with the EC3 column
buckling curves, (ii) adopts critical mode shape initial geometrical imperfections and
(iii) is based on the use of closed-formanalytical expressions to determine their amplitudes.
In order to provide a better grasp of the concepts and procedures involved in the proposed
methodology, and also to illustrate its application and potential, several numerical results
are presented throughout the paper.
2. Geometrically non-linear behaviour of isolated columns
Consider the initially crooked isolated column depicted in Fig. 1(a), which has length L
and exural stiffness EI and exhibits end support conditions that include two rotational and
one translational springs (stiffness K
A
, K
B
and K

, respectively). The initial geometrical


imperfection (crookedness) shape is dened by the function w
0
(x). When acted on by an
horizontal compressive force P, the column undergoes additional lateral displacements
w(x), as shown in Fig. 1(b).
The additional displacements w(x) can be readily found by solving the well-known
fourth-order differential equilibrium equation [5]
EI
d
4
w
dx
4
+ P
d
2
w
dx
2
= P
d
2
w
0
dx
2
. (2)
For practical purposes, the initial geometrical imperfection w
0
(x) is usually expressed
as a linear combination of the column buckling modes, i.e., one has
w
0
(x) =

n=1

(n)
0

(n)
(x), (3)
R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240 1229
where
(n)
(x) and
(n)
0
are the shape and scale factor of the column nth-order buckling
mode. Although this is not essential for subsequent derivations, it is assumed that all modes
have unit amplitude within the column length, i.e., that Max|
(n)
(x)| = 1 for 0 x L.
This automatically implies that
(n)
0
is the maximum eccentricity associated with the
nth-order mode. Incorporating (3) into (2) and keeping in mind that P < P
cr
, one is led to
the general solution [5]
w(x) =

n=1

(n)

(n)
0

(n)
(x), (4)
where
(n)
, designated as nth-order mode amplication factor, is given by

(n)
=
P
P
(n)
b
P
=
P

2
EI
L
(n)
2
b
P
. (5)
The quantities P
(n)
b
and L
(n)
b
are the bifurcation load and buckling length associated
with the column nth-order buckling mode (L
(n)
b
may be viewed as the length of a simply
supported and otherwise identical column having a critical bifurcation load equal to P
(n)
b
).
For simplicity, only the rst term is retained in Eq. (3), which amounts to assuming a
critical mode shape initial geometrical imperfection. Then, making
(1)
and P
b
= P
cr
,
one has
w
0
(x) =
0
(x) =
0
_
A
1
sin
_
x
L
b
_
+ A
2
cos
_
x
L
b
_
+ A
3
x + A
4
_
, (6)
where A
1
A
4
are constants that depend on the boundary conditions. If a computer program
is used to determine the critical buckling load and mode, these constants may be directly
calculated from the nodal displacements and rotations, by solving the system
_
_
_
_
0 1 0 1
sin(L/L
b
) cos(L/L
b
) L 1
/L
b
0 1 0
(/L
b
) cos(L/L
b
) (/L
b
) sin(L/L
b
) 1 0
_

_
_

_
A
1
A
2
A
3
A
4
_

_
=
_

_
(0)
(L)
d(0)/dx
d(L)/dx
_

_
. (7)
However, in order to ensure that Max|(x)| = 1 (0 x L), it is necessary to multiply
the end result by the scale factor

= 1/Max|(x)| (0 x L). Then, the solution of
Eq. (2) and the corresponding bending moment distribution and maximum moment value
are given by
w(x) =
0
(x) =
0
_
A
1
sin
_
x
L
b
_
+ A
2
cos
_
x
L
b
_
+ A
3
x + A
4
_
(8)
M(x) = EI
d
2
w
dx
2
=
P
0
1 P/P
cr
_
A
1
sin
_
x
L
b
_
+ A
2
cos
_
x
L
b
__
(9)
M
max
=
P
0
1 P/P
cr
Max

A
1
sin
_
x
L
b
_
+ A
2
cos
_
x
L
b
_

, (0 x L). (10)
1230 R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240
Note that if the maximum moment occurs at a point corresponding to dM/dx = 0
(0 x L), the previous expression simplies to
M
max
=
P
0
1 P/P
cr
_
A
2
1
+ A
2
2
. (11)
Otherwise, M
max
occurs at one of the column end sections and one has
M
max
= Max{|M(0)|, |M(L)|} =
P
0
1 P/P
cr
Max
_
| A
2
|,

A
1
sin
_
L
L
b
_
+ A
2
cos
_
L
L
b
_

_
. (12)
Since M(x) is a sinusoidal function with period 2L
b
and frequency /L
b
(see (9)), one
readily concludes that, when L
b
L, the maximummoment may always be obtained from
Eq. (11). In simply supported columns with critical mode shape initial imperfections, one
has w
0
(x) =
0
sin(x/L) and the previous expressions yield the well-known results
w(x) =
0
sin
_
x
L
_
, M
max
=
P
0
1 P/P
cr
. (13)
However, it is worth noting that critical mode shape initial imperfections are not
necessarily the most detrimental ones, in the sense that they may not be associated with
the largest second-order bending moments. This statement can be illustrated through the
consideration of the two initially imperfect built-in columns depicted in Fig. 2(a) and (b).
Their initial imperfections display a parabolic shape (Fig. 2(a)) and the critical mode shape
(Fig. 2(b)), both having the same amplitude value
0
. For the parabolic conguration, the
analytical solution reads [7]
w(x) =
4
0
kL
_
kx
2
L
kx +sin(kx) +
cos(kx) 1
tan(kL/2)
_
,
M
max
=

0
P
cr

2
_
2
kL
tan(kL/2)
_
,
(14)
where k =

P/EI . On the other hand, the maximum moment of the column with a
critical mode shape conguration is given by M
max
= 0.5P
0
/(1P/P
cr
) (this result will
be obtained further ahead in the paperSection 3.2). The ratio between the two previous
expressions,
M
max,parab
M
max,cr
=
4

2
1 P/P
cr
P/P
cr
_
1

P/P
cr
tan(

P/P
cr
)
_
, (15)
is a function of the load ratio P/P
cr
and has the values presented in Table 1. One observes
that the parabolic initial imperfection leads to second-order bending moments that are
higher for P/P
cr
< 0.67, which conrms the previous claim that the column strength
may be more affected by initial geometrical imperfections with non-critical mode shapes.
R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240 1231
Fig. 2. Built-in columns with (a) parabolic and (b) critical mode shape initial imperfections.
Table 1
Variation of M
max,parab
/M
max,cr
with P/P
cr
P/P
cr
0.001 0.1 0.2 0.4 0.6 0.8
M
max,parab
/M
max,cr
1.33 1.29 1.24 1.14 1.04 0.93
3. Equivalent member imperfections complying with Eurocode 3
As previously mentioned, both EC3-ENV [3] and EC3-prEN [4] state that the inuence
of member imperfections may be allowed for either (i) in the individual member buckling
checks indirect method or (ii) by incorporating member equivalent geometric
imperfections in the second-order global analysis direct method.
In uniform columns (uniformly compressed prismatic members), the indirect method
leads to a buckling resistance given by N
b.Rd
= Af
y
/
M1
, where is the reduction
factor for exural buckling, which accounts for the member imperfections and is provided
by the appropriate buckling curve. Thus, the buckling check merely consists of comparing
the design value of the axial load N
Ed
with the column buckling resistance N
b.Rd
, i.e.,
N
Ed
/N
b.Rd
1.
In the direct method, the initial geometrical imperfections are incorporated into the
global analysis model and, thus, their effect is reected in the second-order bending
moment design values. Therefore, this method requires using simple (and appropriate)
cross-section interaction equations e.g., the use of simple elastic or linear-plastic criteria
for in-plane behaviour leads to
N
Ed
Af
y
/
M1
+
M
II
Ed
W f
y
/
M1
1 (16)
where it is assumed that N
II
Ed
N
I
Ed
= N
Ed
, M
II
Ed
includes second-order effects, W is the
elastic or plastic section modulus and
M1
is a partial safety factor.
The amplitude of the equivalent initial bow imperfection must be calculated in such
a way that both methods yield the same buckling strength, i.e., the maximum design
axial load value must be always equal to Af
y
/
M1
= N
b.Rd
. If a critical mode shape
imperfection is adopted, M
II
Ed
, given by Eq. (10), reads
M
II
Ed
=
N
Ed

0.d
1 N
Ed
/N
cr
Max

A
1
sin
_
x
L
b
_
+ A
2
cos
_
x
L
b
_

, 0 x L, (17)
where
0.d
is the design imperfection scale factor. A closed-form expression for
0.d
can
then be obtained, by incorporating (17) into (16) and making N
Ed
= N
b.Rd
= Af
y
/
M1
,
1232 R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240
Fig. 3. Variation of the values yielded by Eq. (16) with N
Ed
/N
b.Rd
(curve a and
M1
= 1.1).
which leads to

0.d
=
1

W
A
1

2
/
M1
Max

A
1
sin
_
x
L
b
_
+ A
2
cos
_
x
L
b
_

, 0 x L. (18)
As discussed earlier, whenever the maximum bending moment occurs at a point for
which dM/dx = 0 (0 x L), a condition ensured by L
b
L, the previous expression
simplies to

0.d
=
1

W
A
1

2
/
M1
_
A
2
1
+ A
2
2
. (19)
However, it is important to bear in mind that the formulas used in the two methods are
only equivalent for the limit situation, i.e., for N
Ed
= N
b,Rd
. In order to illustrate this
fact, the curves shown in Fig. 3 plot the values yielded by the direct method (cross-section
interaction formula with second-order bending moments) as a function of the axial load
ratio N
Ed
/N
b,Rd
, (i) for several

values, (ii) taking
M1
= 1.1 and (iii) assuming that the
strength of the column is governed by the EC3 column buckling curve a ( = 0.21). The
curve associated with

= 0 is plotted for comparison purposes, since it provides the values
yielded by the indirect method (formula N
Ed
/N
b,Rd
a straight line with unit slope). One
observes that, as

increases, the curves become progressively more non-linear and the
differences between the values yielded by the direct and indirect methods increase.
In the following subsections, one will address the consequences of applying Eq. (18) to
(i) simply supported columns, (ii) isolated columns with arbitrary end support conditions
and (iii) columns integrated in frames.
3.1. Simply supported columns
In simply supported columns, the unit amplitude critical mode shape initial imperfection
is characterised by A
1
= 1, A
2
= A
3
= A
4
= 0, yielding, from Eq. (18),
R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240 1233
Fig. 4. Variation of the normalised equivalent bow imperfection amplitude e
0
A/W with

.

0.d
=
1

W
A
(1

2
/
M1
) = e
0.d
, (20)
which is nothing other than the amplitude of the so-called equivalent initial bow
imperfection e
0.d
, associated with a column maximum bending moment given by
M
II
Ed
=
N
Ed

0.d
1 N
Ed
/N
cr
=
N
Ed
e
0.d
1 N
Ed
/N
cr
. (21)
The curves displayed in Fig. 4 provide the variation of the exact e
0.d
values, yielded by
Eq. (20) and normalised with respect to W/A (e
0.d
A/W), with

, for curves ad and, once
again,
M1
= 1.1.
Certainly for the sake of simplicity, the EC3-ENV proposed a different expression for
e
0.d
, namely
e
0.d
= (

0.2)
W
A
k

, (22)
which aims at preserving the similarity with the equivalent initial imperfection format used
in the derivation of the column buckling curves (see Eq. (1)). Equating (20) and (22), one
is led to the expression of the correction factor k

, which reads
k

=
1

2
/
M1
1

2
(23)
and, naturally, yields k

= 1 for
M1
= 1. Although the EC3-prEN already includes this
expression, the EC3-ENV proposed
k

= 1 k

+2

1, k

= k

(,
M1
), (24)
which is slightly inexact but clearly much simpler than (23). The plots k

versus

presented in Fig. 5 make it possible to compare the values yielded by the two expressions,
for curves a ( = 0.21) and d ( = 0.76) and
M1
= 1.1. One observes that the two sets
of values are very similar.
1234 R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240
Fig. 5. Comparison between the exact coefcient k

values and the ones proposed by EC3-ENV.


3.2. Isolated columns with arbitrary support conditions
For isolated columns with arbitrary end support conditions, one must consider the
general
0.d
expression, given by (18). However, using Eq. (20), this expression can be
simplied to

0.d
=
e
0.d
Max

A
1
sin
_
x
L
b
_
+ A
2
cos
_
x
L
b
_

, 0 x L, (25)
which immediately shows that, in the general case,
0.d
and e
0.d
are different. Once again,
if the maximum bending moment occurs at 0 x L (dM/dx = 0 stationary point),
the above expression can still be further simplied and one obtains

0.d
=
e
0.d
_
A
2
1
+ A
2
2
. (26)
Using Eqs. (17) and (25), it is possible to reach an expression providing the maximum
bending moment occurring within the column length. It reads
M
II
Ed
=
N
Ed

0.d
1 N
Ed
/N
cr
Max

A
1
sin
_
x
L
b
_
+ A
2
cos
_
x
L
b
_

=
N
Ed
e
0.d
1 N
Ed
/N
cr
, (27)
and is, obviously, identical to the one concerning the equivalent simply supported column
(with the same cross-section, steel grade and length L
b
), which was given in Eq. (21).
Table 2 shows the equivalent initial imperfection characteristics concerning columns
with four commonly used end support conditions, which are depicted in Fig. 6. It is
interesting to notice that, in all these cases, e
0.d
corresponds exactly to the maximum
vertical distance between w
0
(x) and an imaginary straight line connecting two consecutive
inection points. It can be shown that this geometrical interpretation remains valid as long
as the maximum bending moment occurs at a point for which dM/dx = 0 (recall that
this condition is ensured if L
b
L). Indeed, let us consider the initial conguration of
a column segment comprised between two consecutive inection points, as depicted in
Fig. 7. Then, one has
w(x) = w
0
(x) =
_
e
0.d
sin
_
x
L
b
_
+(
2

1
)
x
L
b
+
1
_
, (28)
R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240 1235
Table 2
Equivalent initial imperfection characteristics
Support conditions (x) L
b
A
1
A
2

0.d
Fixedxed 0.5[1 cos(x/L
b
)] L/2 0 0.5 2e
0.d
Fixedsliding 0.5[1 cos(x/L
b
)] L 0 0.5 2e
0.d
Pinnedxed 0.7326 sin(x/L
b
) + x/2L
b
0.7L 0.7326 0 1.365e
0.d
Fixedfree 1 cos(x/L
b
) 2L 0 1 e
0.d
Fig. 6. Equivalent initial imperfection congurations.
Fig. 7. Initial deformed conguration of a column segment comprised between consecutive inection points.
M(x) = EI
d
2
w
dx
2
=
N
Ed
e
0.d
1 N
Ed
/N
cr
sin
_
x
L
b
_
, M
max
=
N
Ed
e
0.d
1 N
Ed
/N
cr
, (29)
the latter expression being the required maximum bending moment, which has already
been obtained in Eqs. (21) and (27).
Fig. 8(a) and (b), on the other hand, illustrates a situation in which the above statement
is not true, namely a semi-rigid cantilever column. The maximum bending moment now
occurs at the support, where one has dM/dx = 0, and is given by
M
max
= P
0.d
( +1). (30)
Equating (30) and (21) (or (27)) yields
0.d
= e
0.d
, i.e., an imperfection that, when
K
A
< , is not the maximum vertical distance between w
0
(x) and the line connecting
two consecutive inection points, which is represented by e
max
in Fig. 8(a).
It is also important to realise that
0.d
can be quite large, a statement that can be clearly
illustrated by considering the elastically supported column displayed in Fig. 9. Indeed, it is
possible to show that, provided that the value of K

is kept below
2
EI /L
3
, a low value
1236 R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240
Fig. 8. Semi-rigid cantilever column (a) initial and (b) nal deformed congurations.
Fig. 9. Elastically supported column.
of K
A
implies a column critical mode shape very close to a rigid body rotation, which
automatically means very small bending moments [7]. In this case, an adequate bending
moment value can only be achieved by adopting rather high
0.d
values. For instance, for
K
A
= EI /L and K

= 4EI /L
3
one is led to the following results: L
b
= 1.456L,
A
1
= 0.09382, A
2
= 0.14100 (unit critical mode amplitude) and
0.d
= 5.90e
0.d
(almost 6 times the value of e
0.d
!). The adoption of such large values may lead to unrealistic
imperfections and cause modelling difculties (e.g., the displacements may be so large
that the validity of Eq. (2) becomes questionable). If this is the case, one should allow for
member imperfections indirectly.
3.3. Columns integrated in frames
On the basis of the results obtained for isolated columns, it now becomes possible to
extend the equivalent imperfection method to columns integrated in frames. One must
consider an initial geometrical imperfection with the shape of the frame critical buckling
mode and an imperfection scale factor
0.d
has to be determined for each column, by using
Eq. (25). The proposed methodology comprises the following steps:
(i) Evaluation of the frame critical load and buckling mode shape, by performing a linear
stability analysis of the whole frame, taking into account the axial force distribution
originated by the particular set of applied loads. In order to keep the procedure
relatively simple and without compromising accuracy, it sufces to consider only rst
order member axial forces
3
[2].
(ii) For each column (compressed member), evaluation of its (ii
1
) buckling length L
b
,
(ii
2
) normalised slenderness

and, using Eq. (20) or (22), (ii
3
) equivalent initial bow
imperfection e
0.d
.
3
Although the second-order axial forces generally differ from the rst-order ones, thus meaning that Eq. (4)
and all subsequent calculations are not rigorously valid, these differences are usually negligible and may be
ignored.
R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240 1237
(iii) For each column, calculation of the buckling mode constants A
1
A
4
, on the basis
of the frame overall buckling mode (inserting the column end displacements and
rotations in Eq. (7)). Then, the buckling mode shape for column j is given by

j
(x) = A
1, j
sin
_
x
L
b, j
_
+ A
2, j
cos
_
x
L
b, j
_
+ A
3, j
x + A
4, j
, (31)
where, for simplicity, the unit amplitude condition Max|
j
(x)| = 1, 0 x L
j
,
which was previously adopted, is only introduced in the next step.
(iv) For each column, evaluation of a normalised scale factor

j
ensuring that this
column will exhibit a unit amplitude initial imperfection, i.e.,

j
=
1
Max|
j
(x)|
, 0 x L
j
. (32)
(v) For each column, evaluation of the design scale factor
0.d, j
, by means of Eq. (25) and
multiplying the A
1, j
A
2, j
values by

j
. One obtains

0.d, j
=
e
0.d, j

j
Max

A
1, j
sin
_
x
L
b, j
_
+ A
2, j
cos
_
x
L
b, j
_

, 0 x L, (33)
which leads to an equivalent initial imperfection (for column j ) given by
w
0, j
(x) =
0.d, j

j
(x) =
e
0.d, j

j
(x)
Max

A
1, j
sin
_
x
L
b, j
_
+ A
2, j
cos
_
x
L
b, j
_

,
0 x L, (34)
where
0.d, j

j
is the design scale factor that must be applied to the frame buckling
mode, in order to ensure that the adequate initial imperfection appears in column j .
Since the various
0.d, j

j
values are (in general) distinct, the rigorous design or safety
checking of each column requires the consideration of a different frame imperfection
amplitude and, thus, also the performance of a different global analysis. Although this
unfortunate fact constitutes a major difculty to the application of the equivalent initial
imperfection concept to columns integrated in frames, one should also point out that, at
least in most practical situations, it is relatively easy to identify the critical compressed
members, in the sense that they govern the frame overall strength, which means that only a
few frame analyses need to be carried out [7,8]. Moreover, in the case of non-sway frames,
it may also be argued that it is acceptable to treat all the columns as isolated members
(with end restraints), which implies that it is feasible to consider only a single overall
frame imperfection, combining the individual imperfections of the various columns (each
with the appropriate design scale factor). Although this frame mixed imperfection differs
from the corresponding critical mode shape, it seems quite fair to expect negligible errors,
with respect to the exact approach.
At this stage, it should be pointed out that step (iv) of the above methodology is optional,
in the sense that it only needs to be carried out if
0.d. j
is to represent the maximum initial
deection in column j . Otherwise, because

j
obviously does not appear in the right hand
side of Eq. (34), one may simply adopt

j
= 1 in Eq. (33). Moreover, the fact that several
1238 R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240
Fig. 10. Simple frame (a) geometry and loading and (b) critical buckling mode shapes (P
1
= P
2
): (b
1
) braced
and (b
2
) unbraced.
commercial computer programs offer the possibility of automatically including buckling
mode shape imperfections in the frame global analysis may render the implementation of
the proposed methodology considerably simpler. Indeed, one just needs to obtain, from the
program output, all the column maximum bending moment values M
, j
corresponding to
the frame critical buckling mode. Since these M
, j
values are also given by the analytical
expressions
M
, j
= Max

EI
d
2

dx
2

= N
cr, j
Max

A
1, j
sin
_
x
L
b, j
_
+ A
2, j
cos
_
x
L
b, j
_

,
0 x L, (35)
the design scale factor for column j is readily obtained from Eq. (33) (one must take

j
= 1):

0.d, j
=
N
cr, j
e
0.d, j
M
, j
. (36)
In order to illustrate the proposed methodology, one considers the simple frame depicted
in Fig. 10(a), which is acted by two vertical loads P
1
and P
2
and will be deemed either
braced or unbraced. All members are assumed to be identical (same length and cross-
section) and, for P
1
= P
2
, the frame critical buckling mode shapes are shown in
Fig. 10(b
1
) (braced) and 10(b
2
) (unbraced). They provide the initial imperfection shapes
to be considered.
The application of the proposed methodology yields
Braced frame: L
b,1
= L
b,2
= 0.626L,

1
A
1,1
=

2
A
1,2
= 0.0705,

1
A
2,1
=

2
A
2,2
= 0.6026,
0.d,1
=
0.d,2
= 1.65e
0.d
Unbraced frame: L
b,1
= L
b,2
= 1.157L,

1
A
1,1
=

2
A
1,2
= 0.0000,

1
A
2,1
=

2
A
2,2
= 0.5233,
0.d,1
=
0.d,2
= 1.91e
0.d
and one observes that the imperfection amplitudes are greater than e
0.d
in both cases.
For P
1
= 2P
2
, the columns obviously have different buckling lengths and, therefore,
different imperfections must be considered
R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240 1239
Fig. 11. Braced frame (P
1
= 2P
2
) (a) buckling mode shape and (b) initial imperfection shapes for the safety
checking of (b
1
) column 1 and (b
2
) column 2.
Braced frame:
L
b,1
= 0.600L,

1
A
1,1
= 0.0466,

1
A
2,1
= 0.5694,
0.d,1
= 1.75e
0.d,1
,
L
b,2
= 0.848L,

2
A
1,2
= 0.4411,

2
A
2,2
= 1.0115,
0.d,2
= 0.91e
0.d,2
Unbraced frame:
L
b,1
= 1.002L,

1
A
1,1
= 0.0912,

1
A
2,1
= 0.3574,
0.d,1
= 2.71e
0.d,1
,
L
b,2
= 1.417L,

2
A
1,2
= 0.2580,

2
A
2,2
= 0.8527,
0.d,2
= 1.17e
0.d,2
.
In each case, the rigorous safety checking of each column requires a separate second-
order analysis of the frame, containing an imperfection obtained through the multiplication
of the frame critical buckling mode shape by the corresponding scale factor

0.d, j
(recall
that taking

j
= 1 is optional)Fig. 11 illustrates this procedure for the braced frame case.
It is important to emphasise that each second-order analysis is only applicable to the safety
checking of the corresponding column, i.e., column 1 is checked with the imperfection
shown in Fig. 11(b
1
) and column 2 with the one displayed in Fig. 11(b
2
). However, since
both columns are identical and one has P
1
> P
2
, it is obvious that column 1 is the
critical one, i.e., the only one that must be checked. Thus, only the imperfection shown
in Fig. 11(b
1
) has to be considered.
One nal word to mention that the method prescribed in EC3-prEN [4] is somewhat
equivalent to the one proposed in this paper (e.g., see Eq. (18)), although it (i) employs a
single frame relative slenderness, instead of the normalised slenderness of each column
(compressed member), and (ii) requires the identication of a critical cross-section.
Indeed, the authors have recently shown [7,8] that (i) the frame relative (normalised)
slenderness is equal to the

value of the column that is closer to reaching its plastic axial
resistance and, moreover, that (ii) this column almost always governs the frame overall
strength resistance because of this feature, it was designated as the critical column.
Therefore, the two procedures may be said to be equivalent, provided that the critical
cross-section mentioned in EC3-prENis interpreted as the critical column cross-section
subjected to the largest bending moment.
4. Concluding remarks
The incorporation of equivalent member initial geometrical imperfections in the design
and/or safety checking of steel frames was addressed. In particular, the EC3-ENVand EC3-
prEN provisions involved in this procedure were discussed and claried. An analytical
study concerning the geometrically non-linear behaviour of imperfect isolated columns
1240 R. Gonalves, D. Camotim / Journal of Constructional Steel Research 61 (2005) 12261240
with arbitrary support conditions was presented and the results obtained were used to
propose and illustrate a methodology to determine the equivalent initial imperfections
that must be incorporated in frame compression members when an elastic second-order
global analysis is performed. This method (i) is consistent with the EC3 column buckling
curves, (ii) adopts critical buckling mode shape initial imperfections and (iii) uses closed
form analytical expressions to determine their amplitudes. The illustrative examples
presented and discussed included isolated columns with several support conditions and
simple three-bar non-sway and sway frames.
One important conclusion concerning the design or safety checking of compressed
members integrated in frames is that, rigorously speaking, it is necessary to incorporate
a different frame imperfection (and, consequently, to perform a different global analysis)
for each one of them. However, this difculty is strongly attenuated by the fact that,
in most practical situations, (i) only a few compressed members govern the frame overall
strength and (ii) they can be spotted quite easily then, only the corresponding frame
imperfections must be considered and the number of analyses that have to be performed
reduces drastically. If this is not so, the designer of a multi-storey and/or multi-bay frame
is faced with a tremendous task, requiring an effort far in excess of the one involved in
the current design practice, and simpler alternative methods are clearly needed further
research work must be devoted to the development of such methods.
References
[1] Ayrton WE, Perry J. On struts. The Engineer 1886;62:464.
[2] Barreto V, Camotim D. Computer-aided design of structural steel plane frames according to Eurocode 3.
Journal of Constructional Steel Research 1998;46(13):3678 (full CD-ROM paper paper # 80).
[3] CEN ENV 1993-1-1. Eurocode 3: Design of steel structures, Part 1.1: general rules and rules for buildings.
1992.
[4] CEN prEN 1993-1-1. Eurocode 3: design of steel structures, Part 1.1: general rules and rules for buildings.
2004 (stage 49 draft, June 2004).
[5] Chen WF, Atsuta T. Theory of beamcolumns. In: In-plane behaviour and design, vol. 1. USA: McGraw-
Hill; 1976.
[6] European Convention for Constructional Steelwork ECCS. Manual on stability of steel structures. 1976.
Publ. 22.
[7] Gonalves R. Member imperfections in steel structures. MSc thesis, Instituto Superior Tcnico, Technical
University of Lisbon; 1999 [in Portuguese].
[8] Gonalves R, Camotim D. First and second-order design or safety checking of steel frames: Allowance for
member imperfections. In: Proceedings of SSRC annual technical session and meeting. 2001, p. 333 52.
[9] Gonalves R, Camotim D. On the allowance for member imperfections in the design/safety checking of
steel frames. In: Proceedings of 9th nordic steel construction conference. 2001, p. 5764.
[10] Rondal J, Maquoi R. Le Flambement des Collonnes en Acier. Notice 1091, Chambre Syndicale des
Fabricants de Tubes dAcier, Paris, France; 1980.

Anda mungkin juga menyukai