Вы находитесь на странице: 1из 15

Buckling of built-up compression members in the plane of the connectors

C. TEMPLEAND GHADAELMAHDY
MURRAY
Deparfmenf of Civil and Environmental Engineering, Universify of Windsor, Windsor, ON N9B 3P4, Canada
Received July 24, 1992
Revised manuscript accepted March 2, 1993
An examination of the requirements for the design of built-up compression members in the North American and
European standards and specifications reveals a great variation in the allowable maximum slenderness ratio for an
individual main member, and also in the determination of an equivalent slenderness ratio. The requirements of the
Canadian standard with regard to the determination of the maximum allowable slenderness ratio of a main member
between points of connection can be a bit confusing.
This research involved a study of model built-up members that buckled about an axis perpendicular to the plane
of the connectors. Twenty-four tests were conducted on model built-up members. The theoretical analysis consisted
of a finite element analysis of the model built-up struts. In addition, an equivalent slenderness ratio was calculated
by several methods. These equivalent slenderness ratios were then used in conjunction with the requirements of the
Canadian standard to calculate a compressive resistance, which was compared with the experimental failure load.
From this research on built-up members that buckle about an axis perpendicular to the plane of the connectors it
was found that at least two connectors should be used, that the slenderness ratio of the main member between points
of connection has a significant effect on the compressive resistance, and that Timoshenko's equivalent slenderness ratio
when used in conjunction with the Canadian standard gives results that are in the best agreement with the experimental
results.
Key words: battens, built-up members, compressive loads, connectors, equivalent slenderness ratio.
Un examen des exigences relatives a la conception des elements composes cornprimes contenues dans les normes
et les spkcifications europeennes et nord-amkricaines permet de constater un grand ecart en ce qui concerne la determination de l'klancement maximal admissible pour un ClCment principal et la determination de I'elancement equivalent.
Les exigences de la norme canadienne en ce qui concerne la determination de 1'Clancement maximal admissible d'un
element principal entre les points de raccordement peuvent Cgalement pr&ter a confusion.
Cette recherche incluait 1'Ctude d'elkments composes qui subissent un flambement dans un axe perpendiculaire au
plan des dispositifs d'assemblage. Vingt-quatre essais de modtles d'tlCment compose ont kt6 realises. La partie theorique comportait une analyse par la mkthode des elements finis de modtles d7CtrCsilloncompose. De plus, l'elancement
equivalent a kt6 calculi selon plusieurs mkthodes. Ces elancements ont ensuite kt6 combines aux exigences de la norme
canadienne pour calculer une rksistance a la compression, qui a fait I'objet par la suite d'une comparaison avec la
charge experimentale ultime.
Cette ttude des ClCments composes qui subissent un flambement dans un axe perpendiculaire au plan des dispositifs
d'assemblage a permis de constater la nkcessite de recourir a deux dispositifs d'assemblage ainsi que I'importance de
I'effet de l'elancement de I'element principal entre les points de raccordement sur la resistance a la compression. Les
auteurs ont en outre observe que l'elancement equivalent de Timoshenko combine aux exigences de la norme canadienne donnait des resultats qui correspondaient davantage aux rksultats experimentaux.
Mots clPs : latte, elements composes, charges en compression, dispositifs d'assemblage, elancement equivalent.
[Traduit par la ridaction]
Can. J. Civ. Eng. 20, 895-909 (1993)

Introduction
Built-up compression members are used in structural
engineering for bridge and building columns, and as bracing
and truss members. These built-up compression members
are composed of two or more structural sections connected
by transverse members which can be batten plates, lacing
bars, or perforated plates. The function of these transverse
members is to make the built-up member act as an integral
unit, to hold the main members apart so that a larger
moment of inertia is achieved, and to form the shear connection between the main members. In this paper, built-up
members composed of two main members connected by
plates welded to the main members are studied. The transverse plates welded to the main members are often referred
to as connectors. Typical built-up members are illustrated
in Fig. 1.
NOTE: Written discussion of this paper is welcomed and will be
received by the Editor until April 30, 1994 (address inside
front cover).
Prinlcd In Canada / Imprime au Canada

Built-up compression members can be considered as either


simple or built-up struts, depending on the plane of bending.
If buckling occurs about the axis parallel to the connectors,
the X axis in Fig. la, the connectors simply move with the
main members. The connectors maintain the separation
between the main members, provide rotational restraint to
the individual main members, but transfer little or no forces.
Thus this type of strut may be referred to as a simple strut.
On the other hand, if buckling occurs about the axis perpendicular to the connectors, the Yaxis in Fig. la, the connectors deform and the effect of the shearing forces that
occur in the built-up member cannot be neglected. The consideration of the effect of shear results in the use of an
equivalent slenderness ratio. An equivalent slenderness ratio
is an imaginary slenderness ratio used to calculate the buckling load of a built-up member when buckling involves a
relative deformation of the connectors. This type of strut
is considered to be a built-up strut and will be considered
in this paper.

CAN. J. CIV. ENG. VOL. 20, 1993

(b)

FIG. 1. Typical built-up members.


In this report, the requirements, as contained in several
standards and specifications, for the maximum slenderness
ratio of the individual main members between points of connection and the equivalent slenderness ratio of the built-up
member are considered. Theoretical and experimental results
are presented. This paper concludes with recommendations
for the design of built-up compression members.
In this research, tests were carried out on welded model
built-up specimens. Model strut specimens were used rather
than full-sized struts in order to have better control over the
fabrication and testing of each specimen so that more
accurate results could be obtained. As the purpose of the
research did not include a study of torsional-flexural buckling on the behaviour of built-up columns, rectangular members were used for the main members. Figure 2 shows the
cross sections of the specimens.
Further research with regard to the connection of built-up
members is planned. A study of the existing test results of
full-scale members, augmented, if necessary, with further
tests, is planned to help clarify, or modify, the various
requirements in Clause 19 of the Canadian standard (CSA
1989) that deal with the connection requirements of built-up
members. A study is also planned t o determine the connection requirements of built-up members that buckle about
an axis parallel to the connectors.
The theoretical analysis consisted of a finite element analysis of the model built-up struts. An equivalent slenderness
ratio was also calculated by several methods. These equivalent slenderness ratios were then used to determine the compressive resistance of the model built-up struts.
Thus the purpose of this research is to examine the various
clauses of the Canadian standard (CSA 1989) that deal with
the connection of double members to form built-up compression members, and to recommend changes to these
requirements so that the compressive resistance of built-up
members that buckle about an axis perpendicular to the connectors can be predicted with greater accuracy.

FIG. 2. Cross sections of built-up test specimens: (a) zero


separation; (b) 4.02 mm separation; (c) 7.85 mm separation.
Standards and specifications
Several steel standards and specifications, including the
Canadian, German, and British standards, and the
American specifications, were examined in order to determine the requirements for built-up compression members.
It was found that there is a great variation in the specified
maximum slenderness ratio of an individual main member
between points of connection, and in the specified equivalent
slenderness ratio. Several examples of equivalent slenderness
ratio equations are given in the following sections.
Canadian Standard CAN/CSA-S16.1-M89, "Limit states
design of steel structures" (CSA 1989)
In Clause 19.1 the Standard specifies two requirements
for the maximum slenderness ratio of an individual main
member. Clause 19.1.3(c) requires that

where (KL/r)i is the maximum slenderness ratio of a main

TEMPLE AND ELMAHDY

member between points of interconnection; (KL/r), is the


slenderness ratio of the integral member with respect to the
axis perpendicular to the plane of the connectors; K is an
effective length factor; L is the length of the member; and
r is the radius of gyration.
On the other hand, Clause 19.1.16 contains slenderness
ratio requirements for battened columns, which can be
summarized as follows:
If (KL/r), 5 0.8(KL/r),,
[2a]

and

e ) i < ~ . 7 e )
X

[Zbl

e)

< 40

and

where (KL/r), is the modified slenderness ratio of the


built-up member and (KL/r), is the column slenderness
ratio of the built-up member acting as a unit.
German Standard DIN 4114-1952, "German buckling
specification " (DIN 1952)
The German standard gives two criteria for the maximum
slenderness ratio of an individual member between points
of connection, which are

'

<0.6e)
Y

where (KL/r), is the slenderness ratio of the integral member with respect to the axis parallel to the plane of the
connectors. For the model struts tested, (KL/r), was
greater than (KL/r), so the requirements of [2b] are
applicable. Thus [l] and [2b] seem to contradict each other.
That is, the maximum slenderness ratio of an individual
main member must not be greater than 40 or 60% of
(KL/r),, and at the same time (KL/r)i must be equal to or
less than (KL/r),.
The Canadian standard, in Clause 19.1.4, requires that
the equivalent slenderness ratio, (KL/r),,, for compression
members composed of two or more shapes in contact or
separated from one another by welded connectors shall be

where r is the radius of gyration of the integral member with


respect to the axis about which buckling occurs, which in
this research is the axis perpendicular to the plane of the
connectors, that is, the Y axis; a is the centre-to-centre
distance between connectors; and 5 is the minimum radius
of gyration for one of the main members.
American Specification, "Specificat ion for structural steel
buildings, allowable stress design and plastic design"
(AISC 1989)
In the allowable stress specification Chapter E specifies
that

and that at least two intermediate connectors shall be used


along the length of the built-up member. There is no requirement for an equivalent slenderness ratio. Thus it seems the
factor 0.75 was added to cover the case of buckling about
the Y axis, as well as buckling about the X axis.
American Specvication, "Load and resistance factor design
speci$cation for structural steel buildings" (AISC 1986)
Chapter E has the same requirement as [I], that is, the
slenderness ratio of the individual member between points
of connection cannot exceed the slenderness ratio of the
built-up member. This specification also states the following
requirements for welded connectors when buckling involves
relative deformation that produce shear force in the
connectors:

provided that at least two connectors are used, one at each


of the third points of the built-up column.
Clause 8.212 of the German standard requires that the
equivalent slenderness ratio for built-up columns that buckle
at right angles to the axis perpendicular to the connectors
be taken as

where m is the number of main members.


British Standard BS 5950, "Structural use of steelwork in
building" (BSI 1985)
The requirements of the British standard for the maximum
slenderness ratio of an individual main member, given in
Clause 4.7.9(c), are

Clause 4.7.9(c) also specifies the equivalent slenderness


ratio of built-up columns, about the axis perpendicular to
the plane of the connectors, as

where 4 is the clear distance between adjacent connectors.


This equation is the same as that specified by the German
standard when there are two main members, except for the
second term where the length is measured from centre to
centre of adjacent connectors in the German standard and
from the ends of adjacent connectors in the British standard.
'1t should be noted that an updated German standard has just
been released but to date this standard has not been translated into
English.

CAN. J. CIV. ENG. VOL. 20. 1993

Member
114 (37,38,39) number

dl-

n
W

A = American
B = British
C = Canadian
G = German

80

a/ri

FIG. 3. Comparison of equivalent slenderness ratios according


to various standards.
Comparison of equivalent slenderness ratio equations
Figure 3 compares the various equivalent slenderness
ratios specified in the previous section in standards and
specifications as a part of this research. The graph shows
two sets of curves, one for each of the integral slenderness
ratios used in this study, which are 120 and 80. Each set
of curves can be identified by looking at the graph when
the slenderness ratio of the main member, a/c, is equal to
zero. The curves show how the equivalent slenderness ratio
varies as a/rl is increased.
The Canadian Standard S16.1, in Clause 10.2.1, limits
the slenderness ratio of a compression member to 200. Some
of the individual slenderness ratios in the tests exceeded this
value. Thus, in the graph, some of the curves have been
extended beyond an a/rl of 200, but the curves are broken
to indicate that these slenderness ratios are not allowed by
the Canadian standard.
It should be noted that each standard contains additional
limits on the slenderness ratio of the individual main member
between points of connection. The Canadian standard, for
example, as shown in [2], limits a / c to either 40 or 50,
depending on the ratio of the slenderness ratios about the
two axes shown in Fig. la. These limits are not shown in
Fig. 3.
Theoretical analysis
The finite element method was used to calculate the ultimate compressive load-carrying capacity of the model struts,
and also their compressive behaviour as given by the nonlinear load-deflection curves. The equivalent slenderness
ratio of each model strut was determined using Timoshenko's
method (Timoshenko and Gere 1961), Bleich's method
(Bleich 1952), and the requirements of the Canadian standard (CSA 1989) and the AISC load and resistance factor
design (LRFD) specification (AISC 1986). The theoretical

(0,0,33)
FIG. 4. Finite element model.
compressive resistance was then calculated according to the
requirements of the Canadian standard using each of the
equivalent slenderness ratios, except when the equivalent
slenderness ratio was determined using the AISC LRFD
specification, in which case the compressive resistance was
then calculated using the requirements of the same
specification.
Finite element method
The finite element method was used to predict the theoretical load-carrying capacity of the model struts. This was
done using a computer program, first as an eigenvalue program to predict the critical load, and second as an iterativeincremental program to predict the nonlinear loaddeflection behaviour of the model struts.
A commercial finite element package, ABAQUS (Hibbitt
Karlsson and Sorenson, Inc. 1989), was used. Geometric
imperfections, the initial out-of-straightness of the model
strut, were included in the input for the analysis as the coordinates of the nodes used to define the initial geometric shape
of the strut. The initial shape of the unloaded main members
was defined such that each main member was parabolic in
shape with the maximum out-of-straightness at mid-height.
A linear elastic, perfectly plastic type of analysis was used
to model the material properties. Deformed geometry was
used, as this is a large deflection problem with a nonlinear
load-deflection response.

TEMPLE A N D ELMAHDY

FIG. 6. Model for Timoshenko's equivalent length formula.

+ d ,
FIG. 5. Model for replacing redundant system with determinate
framework.

A two-dimensional finite element model was used, since


buckling was confined to the plane of the connectors. The
main members, connectors, and end plates were all modelled
using two-dimensional Euler-Bernoulli beam elements with
rectangular cross sections.
The axes of the finite element model were taken such that
the cross section of the main members were in the X-Y
plane. Buckling was modelled to take place about the Yaxis,
that is, in the X-Z plane. Figure 4 shows the elements,
nodes, and degrees of freedom for the finite element model
for a strut with three connectors.
The connectors were modelled as one beam element connected at its ends to the corresponding node on each of the
main members. Each connector was assigned the properties
of two parallel connectors together and, as the distance
between the centroids of the two main members was very
small, the connectors were modelled as rigid beams.
The boundary conditions were taken as pin-ended. The
top boundary was free to displace in the Z direction in order
to allow for the shortening of the built-up strut under the
application of load, while the bottom was prevented from
displacing in the Z direction. Each end plate was modelled
as two beam elements with large moments of inertia connected to each other at a common middle node.

To determine the load-deflection curve, the load was


applied, in increments, to the finite element model on the
middle node of the top end plate in the negative Z direction.
Timoshenko's method (Timoshenko and Gere 1961)
The effect of shear forces on the deflection of a column
is greater for a built-up column than for a solid column and
thus decreases the buckling load. These shear forces bend
the main members and connectors. T o account for these
effects on the buckling load of a built-up column, the
concept of an equivalent slenderness ratio is used.
The analysis of a battened column is based on the assumption that there are hinges at the midpoints of the main
members between batten plates, and at the midpoints of the
connectors, as illustrated in Fig. 5. It is also assumed that
the deflected shape is sinusoidal.
To derive the expression for the equivalent slenderness
ratio, Timoshenko determined the effect the shear force
would have on the lateral deflection of a built-up column.
The lateral deflection caused by the bending of the connectors, 6,, and by the bending of the main members, A2, as
shown in Fig. 6, are computed. The effect of shear deformation in the main members and connectors is neglected.
The equivalent slenderness ratio, as determined by
Timoshenko, for a battened column, when the battens are
of practical proportions, is

900

C A N . J . CIV. ENG. VOL. 20, 1993

I-

FIG.

--- 4

7. Model for Bleich's equivalent length formula.

Equation [lo] shows that the equivalent slenderness ratio


is composed of two parts. The first is the square of the
slenderness ratio of the integral column, and the second is
the square of the slenderness ratio of a single main member
between adjacent connectors, multiplied by a factor of
7r2/12, which is 0.82.
Bleich 's method (Bleich 1952)
Bleich developed a formula to calculate the equivalent
slenderness ratio of a pin-ended battened column. Bleich's
derivation is based on an energy approach. The elastic strain
energy of the distorted column consists of three energy
terms, which are due to (i) the axial shortening in the two
main members due to the axial force F,, (ii) the local bending of the two main members, and (iii) the local bending
of the connectors. The forces and moments on the battened
column, in one panel, are shown in Fig. 7. The total energy,
that is, the sum of energy from all the panels, is then used
to determine the buckling load; and, subsequently, the
equivalent slenderness ratio is derived as

where I is the moment of inertia of the integral column about


the axis perpendicular to the plane of the connectors, that
is, the axis about which buckling occurs; and I, is the same
moment of inertia, neglecting the moment of inertia of the
individual main members about their own centroidal axis.
/ 2 , Ai is the crossThus I, = 2 ~ ~ ( d / 2=) ~ ~ d ~where
sectional area of one main member and d is the distance
from centroid to centroid of the main members.

FIG.

8. Member with rigid ends.

Bleich's equation differs from Timoshenko's only in the


second term where the ratio Io/I appears. This ratio of
moments of inertia is less than 1; as a result, Bleich's equivalent slenderness ratio is less than that given by Timoshenko,
and hence a slightly higher buckling load results.
Effect of the length of the end plates
The effect of the length of the end plates on the buckling
load of the model strut specimens was investigated using the
modified stability functions derived by Livesley and
Chandler (1956). These modified stability functions were
derived on the assumption that the end plates were perfectly
rigid, which is close to the real case but not precise, as the
end plates possess a little flexural flexibility. Thus the results
obtained using these modified stability functions indicate
a slightly greater effect on the load-carrying capacity than
is actually the case. It should be emphasized that this sort
of analysis applies to elastic behaviour only.

TEMPLE AND ELMAHDY

TABLE1. Results of tension tests

Specimen No.
Property
-

Average

209 000
321

199 000
327

21 1 000
339

199 000
327

204 000
328

Modulus of elasticity (MPa)


Yield stress (MPa)

The pin-ended model strut was modelled as one element


with rigid end plates of length g at each end as shown in
Fig. 8. Applying these modified stability functions to the
model built-up struts that buckled elastically, it was determined that the maximum increase in critical load for the
models studied in this research project, because of the rigid
ends, was only 0.53%.
Experimental program
Test program
An experimental program was devised to accumulate
enough data to check the validity of the requirements of the
Canadian Standard CAN/CSA S16.1-M89 with regard to
the design of built-up columns that buckle in the plane of
the connectors. The model built-up columns consisted of
two bars of rectangular cross section connected intermittently by plates welded to the main members. The material
of the bars was a hot-rolled carbon steel with an average
modulus of elasticity of 204 000 MPa and an average yield
stress of 328.4 MPa. These mechanical properties were
determined from tension tests performed on each of the bars
used to make the specimens. The tests were all designated
by a test number, such as 441-4-3. The first number indicates
the nominal length of the specimen, the second indicates the
nominal separation, and the third the number of connectors.
The mechanical properties for each tension test are listed
in Table 1, while the specimens made from the corresponding bars are identified in Table 3.
Six model struts were designed and the number of connectors was varied on each. The main members were made
from 25.4 x 6.35 mm bars. Two slenderness ratios of the
integral member, 80 and 120, were used. Thus built-up
columns with a slenderness ratio that falls in the intermediate and in the slender column range were tested. Each
slenderness ratio was tested with three separations, namely
0, 4.02, and 7.85 mm. These separations were originally
chosen to make the slenderness ratio of the individual main
members, for each of the cases listed below, equal to
(a) one half of the integral slenderness ratio for the built-up
member with zero separation and one connector,
(b) one third of the integral slenderness ratio for the built-up
member with a separation of 4.02 mm and three connectors, or,
(c) one quarter of the integral slenderness ratio of the
built-up member with a separation of 7.85 mm and three
connectors.
Because of the length of the end plates these ratios were not
precisely achieved.
For the slenderness ratio of 80, these separations correspond to strut lengths of 293.0, 441.3, and 588.0 mm,
respectively. For the slenderness ratio of 120, the three separations resulted in lengths of 440.0, 660.5, and 881.0 mm,

respectively. These lengths are measured from knife edge


to knife edge.
Each specimen was tested with four different numbers of
connectors in order to vary the slenderness ratio of the individual main members. All the specimens were tested with
zero, one, three, and seven connectors, except for the
293.0 mm specimen which, because of its short length, was
tested with zero, one, three, and five connectors. End connectors were also used at each end of the specimen to make
sure that the two main members stayed together at the ends.
As the specimens with a slenderness ratio of 120 buckled
elastically, only one specimen was required for the four tests.
The number of connectors was simply increased after each
test on the same specimen. The specimens with a slenderness
ratio of 80, however, buckled inelastically, and hence a separate specimen was required for each test. Thus, for the
24 tests, 15 different specimens were required.
The connectors were welded using the Tungsten Inert Gas
process, in two parallel planes, to the narrow width of the
main members. This type of welding was chosen, as it was
felt that the low heat generated by the welding process would
reduce the residual stresses generated from welding the connectors to the main members. The connectors all had the
same cross section, 4.76 x 12.7 mm. The length of the connectors was dependent upon the separation between the main
members, but in general they had a length equal to the separation of the main members plus an overlap of about 5 mm
on each of the main members. Details of the test specimens
with three connectors are shown in Figs. 9a-9c for each of
the three separations used in the test program.
The objective of each test was to obtain an experimental
load-deflection curve from which the load-carrying capacity
of each strut could be obtained.
Test setup
The model strut specimens were all tested with ends pinned
about the axis perpendicular to the plane of the connectors
and with essentially fixed end conditions about the axis
parallel to the plane of the connectors. These end conditions
were achieved by knife edges, one at each end of the specimen and parallel to each other. Two types of knife edges
were used, one when the separation was zero (see Fig. 90)
and the other when there was a separation between the specimens (see Figs. 9b and 9c).
The specimens were tested either in a universal testing
machine or in a small testing frame. The universal testing
machine was used to test the shorter specimens, that is, the
model built-up members with lengths of 293.0, 440.0, and
441.3 mm, while the longer specimens were tested in a small
testing frame. These are shown in Figs. 10 and 11,
respectively.
Because of the simple buckled shape of the model struts,
one of buckling in the plane of the connectors, lateral

CAN. J. CIV.

902

FIG. 9a. Details of test specimens No. 440-0-3 and 293-0-3.

ENG. VOL.

20, 1993

FIG.9c. Details of test specimens No. 881-7-3 and 588-7-3.

displacements were simply measured with a dial gauge placed


at the mid-height of one of the main members, perpendicular
to the long side of the rectangular cross section. For the specimens with zero connectors, two dial gauges were used, one
at mid-height of each main member.
The shortening of the specimens was also measured. The
results obtained were, however, much larger than those
calculated theoretically. As a result of this discrepancy, no
further attempt was made to compare the experimental and
theoretical values of the shortening of the specimens.
The dial gauge arrangement is shown in Figs. 10 and 11.
The horizontal dial gauge below the bottom knife edge in
Fig. 11 was used to ensure that no significant horizontal
displacement of the hydraulic jack occurred as the load was
applied.
I

-1

I a?,'a
I

I
i
' I

tg,
aI

la81
FIG. 96. Details of test specimens No. 660-4-3 and 441-4-3.

Test procedure
The initial out-of-straightness was determined for each
specimen. As buckling occurred in the plane of the connectors, the initial out-of-straightness was measured only in that
plane. The out-of-straightness was determined at mid-height
as well as at the quarter points. The initial out-of-straightness
varied from approximately zero to almost L/250, where L
is taken as the length of the specimen from knife edge to
knife edge. The initial out-of-straightness was used in the
iterative-incremental procedure to predict the theoretical
load-deflection curves.
After the specimens were set up in the testing frame or
machine, the specimens were loaded slowly in increments
that ranged from 2 to 8 kN, depending on the predicted
compressive resistance which was calculated with a resistance

T E M P L E A N D ELMAHDY

FIG. 10. Test setup, universal testing machine.

FIG. 11. Test setup, test frame.

factor of 1.0. As the load approached the predicted compressive resistance, the load increments were gradually
reduced to 0.5 kN.At each load increment, the dial gauge
readings were recorded once the specimen had reached equilibrium. Each specimen was loaded until the load started
to drop or until a small increment in load resulted in a relatively large increase in the lateral deflection.
Results and discussion
Geometric properties
The geometric dimensions of all the specimens were carefully measured prior to testing. The geometric properties

were calculated and are listed in Table 2. The specimens have


been designated by the nominal length and nominal separation only, as the number of connectors does not affect
these geometric properties. In the table, Ai, Ii, and 4 refer
to the area, the moment of inertia about an axis perpendicular to the connectors, and the corresponding radius of
gyration for one main member; A , I, and r refer to the same
properties but for the integral member; Ifiand I, are the
moments of inertia of one of the main members and of the
integral built-up member about the X axis; and I, is the
moment of inertia of the integral cross section about the Y
axis, neglecting the moment of inertia of the individual main
members about their own centroidal axis.

CAN. J . CIV. ENG. VOL. 20, 1993

TABLE
2. Geometric properties
Specimen
No.

L
(mm)

Ai

(mm2)

A
(mm2)

4i

Ii

(mm4)

(mm4)

Specimen No.

440-0-7

DISPLACEMENT ( m m )

10
mm4

Ix
(mm4)

I
(mm4)

r
(mm)

Ti

(mm)

Specimen No.
660-4-1

DISPLACEMENT ( m m )

FIG. 12. Load-displacement curves, Specimen No. 440-0-7.

FIG. 13. Load-displacement curves, Specimen No. 660-4-1.

Load-deflection curves
The load-deflection curves for all specimens were plotted
from the experimental results and from the output obtained
from the finite element iterative-incremental procedure.
Figure 12 shows the experimental and theoretical loaddeflection curves for Specimen No. 440-0-7, which has an
integral slenderness ratio of 120, zero separation, and seven
connectors. For specimens with zero separation, that is, the
two main members are in contact, the experimental buckling
load was about the same as or a little higher than the theoretical buckling load as determined by the finite element
method. This is probably due to the extra stiffness that
results from one main member being in contact with the
other. This was probably not correctly accounted for in the
finite element program, even with links between the main
members.
When the separation between the main members was not
zero, the theoretical buckling loads were greater than the
experimental buckling loads. Figure 13 shows the experimental and theoretical load-deflection curves for a specimen
with a separation of 4.02 mm, one connector, and a slenderness ratio of 120. The experimental and theoretical loaddeflections curves are, in general, in good agreement when
the loads are less than about one half of the failure load.
These curves are in good agreement over the entire loading
range when the number of connectors is zero or one, as

noted in Fig. 13. As the number of connectors increases,


the agreement is not as good. A typical set of load-deflection
curves illustrating a case where the agreement is not as good
is shown in Fig. 14. These curves are for a specimen with
a separation of 7.85 mm, three connectors, and a slenderness
ratio of 120. At least part of this difference may be due to
residual stresses, which were not included in the finite
element program.
Experimental results
The experimental results are summarized in Table 3.
Column 2 gives the mechanical properties of the material
used in each specimen by referring to the applicable tension
test and the corresponding properties listed in Table 1. The
initial out-of-straightness, as measured at mid-height of each
specimen, is listed in Column 3. The experimental buckling
loads are also listed.
Equivalent slenderness ratio
Columns 2-4 in Table 4 list the slenderness ratio of the
integral built-up member and the individual main members.
For the individual member between points of connection two
slenderness ratios are shown. In Column 3 the slenderness
ratio is based on the clear distance between connectors. This
is required when Clause 19.1.16 of CAN/CSA-S16.1-M89
is considered. Column 4 lists the slenderness ratio of the individual main members between points of connection based

TEMPLE A N D ELMAHDY

TABLE
3. Experimental results

Specimen
No.
(1)

Applicable
tension test
(2)

Initial
out-ofstraightness
(mm)
(3)

Experimental
failure
load
(kN)
(4)

Experimental

Specimen No.
881 - 7-3

DISPLACEMENT (mm)
FIG. 14. Load-displacement curves, Specimen No. 881-7-3.

on the centre-to-centre distance. These values are used for


all equivalent slenderness ratio calculations.
The equivalent slenderness ratios were then calculated by
using Timoshenko's and Bleich's formulas, and according
to the requirements of the Canadian standard and the AISC
LRFD specification. These are shown in Columns 5-8 of
Table 4. Timoshenko's and Bleich's formulas, as pointed
out previously, differ only in the second term. Timoshenko's
formula results in a higher equivalent slenderness ratio than
does Bleich's formula. The maximum difference between
these equivalent slenderness ratios is 9.3% and occurs in the
case of zero separation. In this case the moments of inertia
of the individual members has a significant effect on the
overall moment of inertia of the cross section about the axis
perpendicular to the connectors. In other cases the difference
was as low as 0.5%.
The equivalent slenderness ratio as calculated in accordance with the requirements of the Canadian standard are
up to 25.6% less than those obtained from Timoshenko's
formula for specimens that are sparsely connected, but as
little as 0.9% less when a specimen has seven connectors.

Calculated and experimental failure loads


The equivalent slenderness ratios were used to calculate
the compressive resistance of the specimens. These are listed
in Table 5. The compressive resistances were calculated in
accordance with the requirements of Clause 13.3.1 of the
Canadian standard, except for those shown in Column 6
which were calculated in accordance with Appendix E of
AISC LRFD specification. Column 2 lists the buckling load
predicted by the finite element iterative-incremental
procedure.
For comparison purposes the experimental failure loads
are listed in Column 7 of Table 5. Because of the large
variation in the initial out-of-straightness of these specimens
and because the compressive resistance in the Canadian
standard is based on a specimen with an out-of-straightness
of L/1000, it was decided to adjust the experimental

buckling load to better reflect the load-carrying capacity of


the specimen if the out-of-straightness had been L/1000.
This was done as follows:

is the ultimate load-carrying capacity


where PFEM,L/lOOO
predicted using the finite element program when the outis the experimental
of-straightness was set at L/1000; PEXPT
failure load; and PFEM,MOS is the ultimate load-carrying
capacity predicted using the finite element program and the
measured out-of-straightness. It cannot be proven that this
procedure correctly adjusts for the out-of-straightness, but
it is felt that this is a reasonable approach to try to minimize
the differences in the load-carrying capacity of the built-up
member due to the out-of-straightness. This procedure was
followed rather than to just use the buckling load with an
out-of-straightness of L/1000, since specimens with three,
five, and seven connectors had predicted finite element
buckling loads greater than the experimental buckling load
for specimens with the same out-of-straightness. This
adjusted experimental failure load is listed in Column 8 of
Table 5.
Comparing the finite element buckling load with the
actual experimental buckling load (Columns 2 and 7) for
built-up members with the same initial out-of-straightness
shows that the finite element buckling load tends to be in
good agreement when zero and one connector are used. In
the cases where three or seven connectors are used, the exper-

CAN. J. CIV. ENG. VOL. 20, 1993

4. Slenderness ratios
TABLE

Slenderness ratio
Individual member
Specimen
No.
(1)

Built-up
member
(2)

between
battens
(3

centre-tocentre
(4)

Equivalent slenderness ratio


Timoshenko's
formula
(5)

imental failure load is greater than the finite element buckling load for the specimens with no separation, but is less
than the finite element buckling load when the separation
is 4.02 or 7.85 mm.
It was pointed out previously that the equivalent slenderness ratio as calculated by Timoshenko's formula is greater
than that calculated with Bleich's formula. This is also
reflected in the compressive resistance using these equivalent
slenderness ratios, as shown in Columns 3 and 4 of Table 5.
The equivalent slenderness ratio as calculated according to
the Canadian standard results in a compressive resistance
that is higher than that calculated when Timoshenko's or
Bleich's equivalent slenderness ratios are used. When these
compressive resistances are compared with the adjusted
experimental failure loads, it can be seen that the Canadian
standard often results in a compressive resistance that is
greater than the adjusted experimental load for specimens
with a separation between the main members. It is realized
that the slenderness ratio of a few of the main members
exceeds the allowable as established by the Canadian standard and hence a comparison with this standard is not
applicable. It seems that using the equivalent slenderness
ratio calculated by Timoshenko's formula results in a compressive resistance that is in the best agreement with the
adjusted experimental failure load.
It should also be noted that when the compressive resistance is calculated in accordance with the AISC LRFD specification, using the equivalent slenderness ratio calculated

Bleich's
formula
(6)

Canadian
standard
(7)

AISC LRFD
specification
(8)

according to the same specification often results in compressive resistances that exceed the adjusted experimental failure
loads.
Figure 15 illustrates the difference between the compressive resistances calculated using the equivalent slenderness
ratios as determined by Timoshenko's formula and Bleich's
formula, that given in the Canadian standard, and the
adjusted experimental failure load. These results are for a
built-up member with an out-of-straightness of L/1000, an
integral slenderness ratio of 120, and three connectors. It
can be seen that for the built-up members with no separation, all three compressive resistances are less than the experimental failure load. For a separation of 4.02 mm, the
Canadian standard gives a load that is high compared with
the experimental failure load, but both Timoshenko's and
Bleich's buckling loads are very close to the experimental
failure load. With a separation of 7.85 mm, the Canadian
standard again gives a load that is too high, while the compressive resistances calculated using the equivalent slenderness ratios of Timoshenko and Bleich are also too high but
are a little closer to the adjusted experimental values. It may
be more significant in Fig. 15 to plot the compressive resistance versus the slenderness ratio of the main member
between points of connection. This has been done by adding
a second horizontal axis. This graph then clearly indicates
the significant effect that this slenderness ratio has on the
equivalent slenderness ratio and hence the predicted compressive resistance, according to the different methods.

TEMPLE AND

907

ELMAHDY

TABLE5. Calculated and experimental failure loads


Calculated compressive resistance
Specimen
No.
(1)

Finite element
method
(kN)
(2)

Timoshenko's
(kN)
(3)

Bleich's
(kN)
(4)

Canadian
standard
(kN)
(5)

When buckling occurs about an axis perpendicular to the


connectors, the requirement that the slenderness ratio of the
individual member between points of connection cannot
exceed the slenderness ratio of the integral built-up member
does not ensure that a load-carrying capacity is achieved
which is equal to the compressive resistance determined in
accordance with the Canadian standard. This conclusion has
already been reached theoretically by Libove (1985).
When buckling occurs about an axis perpendicular to the
connectors, the number of connectors significantly affects
the load-carrying capacity of the built-up member. The
information in Table 5 indicates that when the number of
connectors is increased from one to three, for built-up
members with a separation between the main members that
is greater than zero, the load-carrying capacity is increased
by a factor of anywhere from 1.4 to 2.0. Figure 16 shows
the theoretical load-deflection curves of a built-up member
with an integral slenderness ratio of 80, an initial out-ofstraightness of 0.1 mm, and no separation for the cases of
zero, one, two, three, and five connectors. The zero and one
connector cases result in the same buckling load, but as
additional connectors are used, the load-carrying capacity
is increased. Increasing the number of connectors from one
to two results in an increase in the buckling load of some
14%. Thus it is recommended that for built-up members
that buckle about an axis perpendicular to the connectors,
at least two connectors, one at each of the third points,
should be used.

AISC LRFD
specification
(kN)
(6)

Experimental
failure
load
(kN)
(7)

Adjusted experimental
failure load
(L/ 1000)
(kN)
(8)

Clause 19.1.17 of the Canadian standard contains a


requirement for the length of a batten plate, a connector.
The length of the batten plate is the dimension of the batten
parallel to the longitudinal axis of the built-up member.
Timoshenko's equation (1961) for the critical load of a
built-up member, from which the equivalent length equation
is derived, is

where E is Young's modulus of elasticity and Ib is the


moment of inertia of the connector about an axis perpendicular to the plane of bending.
It can be seen from this equation that the greater the
moment of inertia of the batten, the greater is the critical
load. According to Bleich (1952), however, the term containing the moment of inertia of the batten is small compared with the other terms in the denominator and can be
neglected for a properly designed batten. Thus, once a batten
with sufficient flexural rigidity is selected, any further
increase in the moment of inertia of the batten is of insignificant importance. The origin of Clause 19.1.17 is not
known and will be the subject of further research.
The connector used in Specimen 881-7-3 does not meet
the length requirements of Clause 19.1.17 of S16.1. Thus
it was decided to check, theoretically, to see what effect it
would have on the compressive resistance if the length was

CAN. J.

908

C I V . ENG. VOL. 20. 1993

Adjusted Expt.
Canadian Std.
~imoshenko'sEq.
Bleich's Eq.

Specimens with

15

- three

connectors
FIG. 17. Predicted critical load vs. moment of inertia of the
batten.

SEPARATION (mrn)

L / r OF MAIN MEMBER
FIG. 15. Buckling load vs. separation and slenderness ratio of
main member.

90

i-

No. of connectors
zero
and I

DISPLACEMENT (rnrn)
FIG. 16. Theoretical load-deflection curves for various
numbers of connectors.

changed to that required by the Standard. For the specimens


with a separation of 7.85 mm between the main members,
the battens should have a length parallel to the main members of about 17.9 mm. The actual length of the batten was
only 12.7 mm. The moment of inertia of the batten, Ib,is

813 mm4, while the Canadian standard would require an Ib


of 2256 mm4. Equation [13] was used to calculate an
equivalent length factor. The compressive resistance was
then calculated in accordance with Clause 13.3.1 of S16.1.
The length, which should probably be called a depth in the
Standard, was varied to see what effect the moment of
inertia of the batten has on the compressive resistance of
the built-up member. The results are shown in Fig. 17.
Changing the length of the batten from 12.7 to 17.9 mm
(a 40% increase) changes the moment of inertia from 813
to 2256 mm4 (a 175% increase) and the predicted compressive resistance from 23.08 to 23.34 kN (an increase of only
1%). Thus it seems that the requirement in Clause 19.1.17
of 516.1 requiring the battens to have a length of not less
than the distance between the lines of welds may be unnecessarily restrictive.
Conclusions
The following conclusions may be stated from this
research for built-up members that buckle about an axis
perpendicular to the connectors.
1. The slenderness ratio of the main member between
points of connection has a significant effect on the compressive resistance of the built-up member.
2. A minimum of two intermediate connectors, one at
each of the third points, should be used.
3. Clause 19.1.4 of the Canadian standard should be
changed to the equivalent length formula derived by
Timoshenko. This equivalent length formula together with
the compressive resistance calculated in accordance with
Clause 13.3.1 gives the best agreement with the experimental
loads.
4. The requirement that the slenderness ratio of the individual main members between points of connection be equal
to or less than the slenderness ratio of the integral member
is not applicable to these members and does not ensure that
the required load-carrying capacity is achieved.

TEMPLE AND ELMAHDY

AISC. 1986. Load and resistance factor design specification for


structural steel buildings. American Institute of Steel Construction, Chicago, Ill.
AISC. 1989. Specification for structural steel buildings, allowable
stress design and plastic design. American Institute of Steel
Construction, Chicago, Ill.
Bleich, H.H. 1952. Buckling strength of metal structures. McGrawHill Book Company, Inc., New York, N.Y.
BSI. 1985. Structural use of steelwork in building. BS 5950, Part 1,
British Standards Institution, London, England.
CSA. 1989. Limit states design of steel structures. CAN/CSA
S16.1-M89, Canadian Standards Association, Rexdale, Ont.
DIN. 1952. German buckling specification. Deutscher Normenausschuss, Beuth Vertrieb G.M.b.H., Berlin and Cologne, Germany.
Translated by J. Jones and T.V. Galambos. Column Research
Council (now Structural Stability Research Council), Lehigh
University, Bethlehem, PA.
Version 4-8,
Hibbitt, Karlsson and Sorenson, Inc. 1989. ABAQUS,
Vol. 1: Theory manual; Vol. 2: Verification manual; Vol. 3:
User's manual; Vol. 4: Example problem manual.
Libove, M. 1985. Sparsely connected built-up columns. ASCE
Journal of Structural Engineering, lll(3): 609-627.
Livesley, R.K., and Chandler, D.B. 1956. Stability functions for
structural frameworks. Manchester University Press, Manchester,
England.
Timoshenko, S., and Gere, J.M. 1961. Theory of elastic stability.
3rd ed. McGraw-Hill Book Company, New York, N.Y.

List of symbols
total cross-sectional area
cross-sectional area of one main member
centre-to-centre distance between connectors
clear distance between adjacent connectors
compressive resistance
distance from centroid t o centroid of the main
members
Young's modulus of elasticity
axial force in one main member in panel r
length of rigid end plates
moment of inertia of the integral cross section
about the axis perpendicular t o the plane of
the connectors
moment of inertia of a connector about the
horizontal centroidal axis perpendicular t o the
plane of the connectors
moment of inertia of one main member about
its centroidal axis perpendicular t o the plane
of the connectors
moment of inertia of the integral cross section
about the axis perpendicular to the connectors,
the axis about which buckling occurs, neglecting the moment of inertia of the individual
main members about their own centroidal axis
(= ~ ; d ~ / 2 )

moment of inertia of the integral built-up


member about the X axis
moment of inertia of one main member about
its X axis
effective length factor
length of member
length of centre part of column between the
rigid ends
end moments applied at ends a and b, respectively, t o produce disturbance of member
moment in connector r
moment in main member in panel r
number of main members
external load
critical load
experimental failure load
adjusted experimental buckling load
ultimate load-carrying capacity predicted using
the finite element program when the out-ofstraightness was set at L/1000
ultimate load-carrying capacity predicted using
the finite element program and the measured
out-of-straightness
lateral shear force
shear force in connector in panel r
shear force in panel r
radius of gyration of the integral built-up
member about the axis perpendicular t o the
plane of the connectors
minimum radius of gyration for one of the
main members
coordinate axes
displacement i n X a n d Z directions,
respectively
lateral deflection caused by bending of the
connectors
lateral deflection caused by bending of the
main members
rotational degree of freedom
disturbances applied at ends a and b, respectively, of a member
Subscripts
equivalent
individual main member
modified
integral built-up member
about X o r Y axis, respectively

Вам также может понравиться