Sei sulla pagina 1di 7

Design of aluminium

columns
Y. F. W. Lai
Mitchel MacFarlane, Hong Kong
D. A. Nethercot
Department of Civil Engineering, University of Nottingham, UK
(Received April 1990; revised January 1991)
Theoretical results for the buckling of aluminium columns having
longitudinal and local transverse welds are presented. These are used
to assess the suitability of the procedures given for column design in
th~'~raft British Standard for the use of structural aluminium
BS 8118. As a result some modifications to these procedures are
suggested.
Keyword$: aluminium structures,
structural design, welding
The new draft British code for the design of aluminium
structures BS 8118 I, which will replace CP 1182, has
recently been circulated for the purpose of inviting
public comment.
Although the column design curves suggested in
Reference 1 are based largely on test data and accurate
numerical studies 3, their suitability is still uncertain for
some types of member, especially welded columns
and/or columns subject to flexural-torsional buckling.
This paper presents the results of an extensive comparison between the column design curves of the draft
BS 8118 and numerical results obtained by two finite
element programs INSTAF and BIAXIAL 4. All comparisons are restricted to 'compact' cross-sections for
which no local buckling occurs.

Nomenclature
A
A*

C,,
Cr
E
L
L*

L,r
L,.
P,

P~c

Pult
/5
n
n*
r

area of cross-section
area of heat-affected zone (HAZ)
buckling coefficient for flexural buckling
buckling coefficient for torsional buckling
Young's modulus
length of member
length of heat-affected zone
critical region defined as a distance extending
from 0.25L either side of point of maximum curvature when flexural buckling takes place
total length within critical region over which
heat-affected zone softening occurs
ultimate compressive strength given by draft
BS 8118
compressive capacity of cross-section
calculated compressive strength of member
nondimensional maximum compressive strength
of member (= Pult/Oo2A)
knee factor in Ramberg-Osgood formula
knee factor for HAZ in Ramberg-Osgood
formula
radius of gyration

0141-0296/92/030188-07
1992 Butterworth-Heinemann Ltd

188

Eng. Struct. 1992, Vol. 14, No 3

buckling,

columns,

stability,

ro ry radius of gyration about x- and y-axis, respectively


h , X, slenderness (L/rx, L/ry)
~ , ~, nondimensionalized slenderness ratio (Xx/r(E/

00.2)m, X,./Tr(E/ao.2)1/2)
o
a02
a*2
ault
w

normal stress
0.2% proof stress
0.2% proof stress of HAZ
tensile strength
material strength reduction factor for heataffected zone material properties.

Numerical results
All of the numerical results presented herein were
obtained by using the programs INSTAF for twodimensional response and BIAXIAL for threedimensional response 4. Both programs permit the full
load deflection curve to be traced up to collapse utilizing
sophisticated finite element approaches 56
' , which permit
the effects of longitudinal and/or transverse welds to be
allowed for. Both have been extensively verified against
test data and alternative ultimate strength analyses for
both aluminium and steel members 4.
Results have been obtained for a large number of
axially loaded columns, covering the range of different
types for which the design approach of Reference 1 is
intended. Tables la and lb list the individual cases
studied, Alloy types are characterized by the parameter
n used in the Ramberg-Osgood representation of the
stress-strain curve, with the HAZ (heat affected zone)
material being assumed to possess properties
corresponding to 0"2 = 0.500.2 and n* = 10. All columns were assumed initially bowed about the minor axis
in the form of a half sine wave with a maximum
amplitude of L/IO00.

Column design procedure of draft BS 8118


The factored axial resistance of a column, Pc, which is

Design of aluminium columns: Y. F. W. Lai and D. A. Nethercot


Table la
Class

List of theoretical column curves (nonwelded and longitudinally welded columns)


Theoretical
curve

A'
A

Type of
cross -

Program
used

Principal
results

Gult
Gult/GO'2
.....................................................................
A

C-A-I
C-A-2

1.17

C-B-I
C-B-2
C - B - 3
C-B-4
C-B-5
C-B-6

1.17

C-C-I
C-C-2

1.08
1.17

C- D- I
C-D-2

I .17

C - E- I
C-E-2

I .49
1.61

C-F-I

25

(I}

0.0
0.0

I
III

INSTAF
BIAXlAL

Figure

0.1
0.3
0.5
0.1
0.3
0.5

III

BIAXIAL

Figure 2

50
25

0.0
').0

II
IV

INSTAF
BIAXIAL

Figure 4

25

0.3
0.5

IV

BIAXIAL

Figure 5

10
8.4

0.0
0.0

I
III

INSTAF
BIAXIAL

Figure 6

1.61

8.4

0.3

III

BIAXlAL

Figure 7

C-G-I
C-G-2

1.36
1.95

13
6

0.0

IV

BIAXlAL

Figure 8

0.0

C-H-1

1.95

0.1

......................................................................
B

25

I~

lqOmm

10ram

Figure 3

~ ~

I- 80mm~l
"~ ~

lOmm
mm

{w)
BIAXIAL

11mm
7mm

(iH)

IV

~I
10mm

(It)

......................................................................
Table lb

, . 3 - -3...

Figure 9

Type of cross-section

List of theoretical column curves (transversely welded columns)

Theoretical
curve
reference
C-TW- 1
C-TW-2
C-TW-3
C-TW -4
C-TW-5
C-TW-6
C-TW-7
C-TW-8
C-TW-9
C-TW-IO
C-TW- 11

O'ult

L*/L

Type of crosssection (see


Table 1 (a))

Program
used

Principal
results

1.17

25

INSTAF

Figure 10

1.17
1.17
1 17
1.17

25
25
25
25

I
I
I
III

INSTAF
INSTAF
INSTAF
BIAXIAL

Figure
Figure
Figure
Figure

1.17

25

0.1
0.2
0.3
1.0
0.0
0.05
0.1 at both ends
L* = 50 mm at mid-height
L* = L
L* = 0
L* = 30 mm at both ends

III

BIAXIAL

Figure 15

00.2

influenced by the compressive proof stress of the


material 00.2, the area, HAZ effects, slenderness
and the degree of end fixity, thickness of the plate
elements and torsional properties of the cross-section, is
given by

P~=P,~C~

(1)

where
Psc = basic axial capacity and Cc = reduction factor for
overall flexural buckling.
In determining Psc due account must be taken of
HAZ effects, so that

Psc=GO.2[A--~

(1 - w)A*]

(2)

11
12
13
14

in which the second term in the square brackets allows


for the reduction in strength due t o the presence of HAZ
material.
The value of Cc is given as a function of a nondimensional slenderness parameter )] = (L/r)(ao.2/250)m by
the set of five column curves provided in Figure 5.9 of
Reference 1; selection of the appropriate curve for a particular column type will be discussed later. For a column
liable to fail due to torsional instability 7, the reduction
factor for torsional buckling Cr replaces Co. Cr is determined from Figure 5.10 of Reference 1, in which the
four curves are the upper four curves for flexural buckling.
The design of columns containing localized transverse
welds is covered in Appendix 5A of Reference 1. The
effect on buckling strength is assumed to depend solely
on the value of LJLcr, in which Lw is the length of
HAZ and L~r defines that part of the column containing

Eng. Struct.

1992,

Vol. 14, No 3

189

Design of aluminium columns: Y. F. W. Lai and D. A. Nethercot


Table 2

Factors used in selection of column curve in Reference 1

systematic comparisons against the 29 sets of numerical


results of Table 1.

Extent of HAZ

Column curve 1 and class A columns

Lw/Lc,

Design approach

Ignore presence of transverse


welds
Design column as if it consisted
wholly of HAZ material

>0.2

(Psc = a(~2A)
0 < Lw/Lcr < 0.2

Table 3

Interpolate between the above


t w o cases based on actual
Lw/Lc, value

Figure i shows how for the two sets of numerical results


corresponding to the most favourable class of member
curve 1 represents a safe and reasonable design basis
over the whole range of slenderness considered. For the
torsionally weaker narrow flange section III the design
curve is more conservative, particularly at medium
and high slenderness. For stocky columns the use of the
squash load P~, = %2 A means that the beneficial
effects of the continuously rising material stress-strain
curve are not utilized.

Factors used in selection of column curve in Reference 1

Column curve 1 and class B and C columns


Oult/O0.2

Alloy type
Cross-section
Welding

high n (H)
yl/Y2* < 1.2
symmetric (s)
nonwelded (NW)

O'ult/O'0.2 > 1.2


low n(L)
Yl/Y2 > 1.2
aysymmetric (A)
welded (W)

* y l and y2 are perpendicular distances from the axis of buckling


to the further and nearer extreme fibres, respectively

Table 4

Six cases of longitudinally welded members - three


each with symmetrically and non symmetrically
arranged welds - are covered in Figures 2 and 3.
Residual stresses have been included in all cases except
for unsymmetrical welds and A*/A = 0.5. Design curve
2 is safe for all but a few cases of very stocky members;
even there it overestimates strength by only a few per
cent. However, as slenderness increases the design

Allocation of cases to column design curves


I

Class
(see Table 1)
A
B
C
D
E
F
G
H

Condition

Draft BS 8 1 1 8
allocation

Proposed
allocation

H-S-NW
H-S-W
H-A-NW
H-A-W
L-S-NW
L-S-W
L-A-NW
L-A-W

1
2
2
3
3
4
4
5

1
2
2
3
2
3
3
4

1.0

~"-,~"'-~/~
" ~ ,
0.5

0.0

0.5

0.0

Lw undergoing the largest curvatures. Three cases are


identified as indicated in Table 2. Reference 1 does not
recognise the possibility of the HAZ not extending over
the full depth of the cross-section.
For end-welded columns for which the welds extend
for less than 0.05L the effect of transverse welds may be
neglected. In all cases it is, however, necessary to
ensure that the axial load does not exceed the crosssectional capacity Psc = %.2A.
All the column curves are described by a
P e r r y - R o b e r t s o n type of equation with the selection for
a particular column being based on the factors listed in
Table 3. Assessment on the basis of Table 3 permits the
columns to be graded into eight classes according to the
combinations of these conditions present and thus to be
rated according to the number of weakening conditions
as indicated in Table 4. The higher the rating number the
weaker is that class of column.

c-A-2 ~B, Ax, A, i

I .0

1.5

ix or Y
Figure 1 Comparison between design column curve 1 and
theoretical column curves (class A columns). (
), theoretical
curves; ( - - - ) , draft BS 8 1 1 8

....

1.0

-~

__ _

---~
~

\ , ~

I
0.0

C-B-2

-~- ~.--.~ -.~ - " ~

0.0

C-B-I

0.5

I .0

I .5

Comparison between design column curves of


draft BS 8118 and numerical results

Figure 2 Comparison between design column curve 2 and

The accuracy of the column design approach described


in the previous section has been assessed b y means of

theoretical column curves (class B columns, with symmetric


longitudinal welds). (
) theoretical curves (biaxial); ( - - - ) ,
draft BS 8 1 1 8

190

Eng. Struct. 1992, Vol. 14, No 3

Design of aluminium co/umns: Y. F. W. Lai and D. A. Nethercot

1.0 ____.<-=-~_.._

IO,. 0.5 -

Curve 2 is also applicable to asymmetric nonwelded


columns and Figure 4 shows that when design is controlled by flexural buckling (section II) the design procedure becomes unsafe when Xx falls below 0.65 but is
quite conservative over much of the range. When torsional buckling controls (section IV) the design curve is
always safe, possibly excessively so over much of the
range.

/c-B-.

- ' ~ ' ~ " ~ ~

Column curve 3 and class D and E columns


0.0
I
0.0

I
0.5

! I
1.0

I I
1.5

Curve 3 is applicable to either symmetrical, unwelded


shapes of low n-material or to welded, asymmetric
shapes. Both types have been considered and as shown
in Fieures 5 and 6 the design curve generally underpredicts the strength, particularly for the more slender
members. This would appear to be due to the relative
lack of yielded material present at failure for such cases.

Figure 3

Comparison
between
design column
curve 2 and
theoretical column curves (class B columns,
with asymmetric
longitudinal welds). (
), t h e o r e t i c a l c u r v e s (biaxial); ( - - - ) ,
d r a f t BS 8 1 1 8

curve becomes increasingly conservative,


because it is not necessary to make much,
allowance for HAZ as slenderness increases
applied loads become insufficient to induce
approaching trY.2over much of the member.

1.0

largely
if any,
and the
stresses

Column curve 4 and class F and G columns

The three sets of results covering welded, symmetrical,


low n-material members and nonwelded, asymmetrical,
low n-material members presented in Figures 7 and 8,

C-C-I (INSTAF)

~,.~-- ~

Z C - 2

(BIAXIAL)

la.

I~0.5

I
0.0
0.0

0.5

15

1.0

0.0i
0.0

1.

Figure 4 Comparison between aes=gn column curve 2 and


theoretical column curves (class C columns). (
curves; ( - - - ) , draft BS 8118

~\.

); theoretical

0.0

1.5

orXy
curve 3 a n d
); t h e o r e t i c a l

1.0
~

"~\\ C-F-I (BIAXIAL)

IQ0.5 --

0.5

1.0

1.5

0.0

~'~'~..

i
0.0

Comparison
between
design column
theoretical column curves (class D columns). (
c u r v e s (biaxial); ( - - - ) , draft BS 8 1 1 8

0.5

Figure 5

Comparison
between
clesign c o l u m n
t h e o r e t i c a l c o l u m n c u r v e s (class E c o l u m n s ) . (
c u r v e s ; ( - - - ), draft BS 8 1 1 8

C-D-1

0 " 5 r-

0.0

Figure 6

la.

1.0

-~x o ~ y

1.0

0.5

1.0

1. S

Y
curve 3 and
), t h e o r e t i c a l

Figure 7

Comparison
between
design column
t h e o r e t i c a l c o l u m n c u r v e s (class F c o l u m n s ) . (
c u r v e ; ( - - - ) , draft BS 8 1 1 8

Eng. S t r u c t .

1992,

curve 4 and
), t h e o r e t i c a l

Vol. 1 4 , No 3

191

Design of aluminium columns: Y. F. W. Lai and D. A. Nethercot

Column curve 5 and class H columns


1.0

When all adverse effects are combined Figure 9 again


shows the proposed column curve to be excessively conservative over the whole range. From Reference 3 it is
clear that the original positioning of curve 5 was based
on relatively few numerical or experimental results.

\
\\

.c-c-1

~.

.C-G-2

0.5

Columns containing transverse welds

0.0
0.0

0.5

1.0

Eleven cases of transversely welded columns have been


considered as detailed in Table lb. In an earlier study 4,
it had been shown that locating such welds at column
mid-height i.e. the region of greatest curvature and
highest stress, led to the largest reductions in strength as
compared with an equivalent unwelded member.
For the three cases of centrally welded columns
Figure 10 shows the strength to be only slightly greater
than that of a wholly HAZ member (L*/L = 1.0). The
design procedure ii of Table 21 deals with this case
quite well as the comparison in Figure 10 shows.
However, when the extent of the welding falls below
that corresponding to case ii (L*/L = 0.05), then Figure
11 shows that the use of interpolation as permitted by
case iii of Table 2 leads to unsafe results for the more
stocky end of the range. The reason for this is quite

1.5

Figure 8 Comparison

column curve 4 and


theoretical column curves (class G columns). (
), theoretical
curves (biaxial); ( - - - ) , draft BS 8118
between

design

respectively show that curve 4 significantly underestimates strength in all cases. The degree of conservatism is variable but approaches 100% for the most
slender members.

1.0
\\

C-H-I (BIAXIAL)

1.0

----

\
\\

0.5

0.5

1.0

I. 5

0.0
0.0

\\\.~

0.5

design

1.0

curve (INSTAF); ( - -

I .5

1.0

....

~'\/C-TW-7(L*IL=0,1

-=-~.~.~
~-~.

)(" at both ends)


,~\

\\

\\

\,

0.5

0.0I

0.5

1.0

0.0

1.5

0.0

Figure 10 Comparison with transversely welded column curves


obtained with INSTAF program (Lw/Lcr > 0.2). (
), theoretical
curves (INSTAF); ( - - - ) draft BS 8118

Eng. Struct. 1992, Vol. 14, No 3

0.5

-2

192

- ) , draft BS 8118

0.5

0.0

Figure 11 Comparison with transversely welded column curves


obtained by INSTAF program (LwlLcr = O.1). (
), theoretical

C-TW-1 {L*IL = 0.1 )\X


, ~ C - T W - 2 (L*IL =0.2),\
C-TW-3(L*/L=0.3) \
-

column curve 5 and


theoretical column curves (class H columns). (
), theoretical
curve; ( - - - ) , draft BS 8118
between

I .0

Figure 9 Comparison

C-TW-6 (L*/L=O.05)

0.5

0.0
0.0

1.0

1.5

~v

Figure 12 Comparison with

welded
column curves
by INSTAF program
(end-welded
columns).
(
),
theoretical curves (INSTAF); ( - - - ) , draft BS 8118
obtained

transversely

Design of a/uminium co/umns: Y. F. W. Lai and D. A. Nethercot

Proposed design improvements

simply that the cross-sectional capacity /'st of such


members is still, of course, only O'~.2 and that for
stocky columns only a relatively small part of the total
longitudinal stress is due to bending. Thus members failing by 'squash' as well as those for which little bending
occurs will have their strength overpredicted due to the
optimistic value of P=, used in the design process. For
slender members, in common with several other cases
considered herein, the design method of Reference 1 is
conservative since the allowance made for HAZ does
not reduce in an appropriate way.
For columns welded only at their ends Reference 1
proposed neglecting any reductions in strength. Figure
12 shows this to be clearly unsafe over much of the
range. A simple procedure to correct this is given in the
next section.
When failure is due to torsional buckling and the
extent of HAZ present is sufficient for case ii of Table
2 to apply, Figure 13 shows the use of Cr in place of Cc
to be reasonable. Figure 14 shows that for an endwelded column, limiting the strength on the basis of P=c
calculated for the HAZ material would provide the basis
for a satisfactory design treatment.

Nonwelded

or

longitudinally welded(2t)

The results presented in Figures 1 - 9, corresponding to


the 29 cases of nonwelded or longitudinally welded columns listed in Table la, suggest that the number and
spread of design curves proposed in Reference 1 is too
great. Generally speaking the proposals for the more
favourable cases are satisfactory,_ if rather too conservative at higher slendernesses 0~ > 1.0), while those
requiring the use of curves 4 and 5 are much too conservative over the whole range of slendernesses. Table 4
therefore suggests an alternative allocation of cases
using just 4 column curves. For classes A - D the allocation is as before; classes E - H all move up one curve,
with the lowest curve 5 being dropped.

Transversely welded
The proposed method distinguishes between the 2 cases
(1) P*_< P*, in which P * = o*A is the elastic limit for
pure compression
(2) P*> P*

1.0
k

C-TW-8 (L* = 50 mm

~/at

mid-height)

J~xC-TW-9(L*=L)

",.,,

0.5

0.0
0.0

0.5

(i)

1.0

P = 1 - (1 - w)A*/A

(ii)

Figure 13 Comparison with transversely welded column curves


obtained by BIAXIALprogram (centrally-welded columns). (
theoretical curves; ( - - - ) , draft BS 8118

-~\

"~ ~" ~ , , ~

'~C-TW-11

When longitudinal welds are also present they should be


allowed for when determining P=c and thus Pc*.

Conclusions

(L* = 30 m m

at both ends)

___.S"

0.0

0.0

0.5

1.0

If condition (i) is not satisfied the column is


designed as if it consisted of wholly HAZ material.

),

C-TW-10 (L*=0)

~'~~N

o.,_

The HAZ is located near the ends within a distance


of 0.25L of the supports. The column is designated
'end-welded' and may be designed as if it were
nonwelded subject to an upper limit given by

1.5

1.0

For the first case HAZ effects may be neglected. For the
second two variants are recognised:

1.5

Figure 74 Comparison with transversely welded column curves


obtained by BIAXIAL program (centrally-welded columns). (
),
theoretical curves; ( - - - ) , draft BS 8118 (modified in case of
end-welded column)

A series of numerical results for the flexural and


flexural-torsional buckling of various types of
aluminium column have been presented. The cases considered were selected to cover the full range of effects
taken into account in allocating column types to
design curves by the draft BS 8118. Detailed comparisons between the numerical results and the
appropriate column curves have revealed that
The strength of most types of column is safely
predicted by the method of the draft code.
For asymmetric cross-sections or material with
au,/%.2 > 1.2 underestimates of strength are possible, especially for columns of intermediate
slenderness.

Eng. Struct. 1992, Vol. 14, No 3

193

Design of aluminium columns: Y. F. W. Lai and D. A. Nethercot

As a general rule the reductions in load carrying


capacity obtained by passing through the different
categories A - H appears to be too great. Some
upward revision of the lower design curves coupled
with some re-allocation of classes would appear to
be justified.
Suggestions for improvements to the treatment of
transversely welded columns, particularly when the
welds are located at the column ends, have been
provided.

Acknowledgments
This work forms part of a project funded by
R.A.R.D.E.; the authors are grateful for assistance and
comments on the studies reported herein by Mr D.
Webber and Dr P. S. Bulson.

194

Eng. Struct. 1992, Vol. 14, No 3

References
1 British Standards Institution Draft, British Standard BS 8118, 'Code of
practice for the design of aluminium structures', 1985
2 British Standards Institution, CP 118: 1969, 'The structural use of
aluminium'
3 Hong, G. M. 'Aluminium column curves', Aluminium structures.
design and construction, R. Narayanan Ed., Elsevier Applied Science
Publishers, 1987, pp. 40-49
4 Lai, Y. F. W. and Nethercot, D. A., 'Strength ofaluminium members
containing local transverse welds', Engineering Structures, (in press)
5 E1 Zanaty, M. H. Murray, D. W. and Bjorhovde, R. 'Inelastic
behaviour of multistorey steel frames,' University of Alberta, Canada,
1980
6 El Khenfas, M. A. and Nethercot, D. A., 'Ultimate strength analysis
of steel beam-columns subjected to biaxial bending and torsion', Res.
Mechanica, 1989, 28, (1-4), 307-360
7 Nethercot, D. A. 'Aspects of column design in the new UK structural
aluminium code', Aluminium structures: advances, design and construction, R. Narayanan, Ed., Elsevier Applied Science Publishers,
1987, pp 50-59

Potrebbero piacerti anche