Professional Documents
Culture Documents
Section Columns
Dr K. S. Virdi*
EXPERIMENTAL EVIDENCE
(L/r) E
where 2 = - - Numerous carefully controlled tests have shown that
,']-v
practical columns of different cross-sectional shapes
show different levels of deviation from the ideal
J, is often called the slenderness factor, to distinguish
behaviour. This deviation is only in part clue to tile
it from the slenderness ratio. Graphically, equation 2 is
actual shape of the section, An important factor is the
represented by the curve ABC shown in Figure 1, [t is
pattern of distribution of residual stresses within the
plain that the part of the curve above N = 1 is merely
section. The scatter in experimental results obtained by
of academic importance, since real columns would not
several investigators is further explained by the fact that
support a stress in excess o f f , if the effect of strain
nominally identical specimens can have differing levels
hardening is ignored. Thus the buckling curve of an
of residual stresses due to different manufacturing
ideally straight column made or an elastic perfectly
histories. Yet another source or scatter in experimental
plastic material would be represented by the
results is the observed fact that the measured yield
characteristic ABD in Figure I.
stress can be quite different in magnitude at different
points within a cross-section. All these factors suggest
DEVIATION FROM IDEAL BEHAVIOUR the use of multiple column curves, rather than a single
For nearly a century now, it has been accepted that the one, for design.
behaviour of a practical column deviates from that of a
similar ideal column, particularly in the range PRESENT INTERNATIONAL
corresponding approximately to 0.3 < 2 < 1.5. The RECOMMENDATIONS
foremost cause was found to be the lack of straightness Under the auspices of the European Convention for
of the column in its unloaded state. This led Perry to Constructional Steelwork (ECCS), Beer and S c h u l z ':~
propose his famous formula '7), as early as 1886, based examined the problem both theoretically and experi-
on the criterion of first yield of a column with an mentally and, as a result, the E C C S Recom-
initially curved shape in the form of a half sine wave. mendations qt~ contain three basic column buckling
This formula, with many empirical values substituted curves, termed European curves a, b, and c, together
for the magnitude of the initial bow, has survived in the with a curve selection table for various types or
codes of practice relating to steel design in Britain and columns. Under these R e c o m m e n d a t i o n s , rolled
many other countries for more than fifty years. structural hollow sections, the subject of the present
In the USA, the presence of residual stresses within paper, should be designed using curve "a'. For welded
Design of Circular and Rectangular Hollow Section Columns 37
that country, namely the American Institute of Steel Figure 2 ,4 curvilinear stress-strain curve approximated with
linear segments
Construction (AISC). has deliberately decided against
the use of multiple column curves, in order to retain the M E T H O D OF A N A L Y S I S
simplicity offered by a single column curve. Several analytical methods are available for the elastic-
In the UK, the code drafting committees of the plastic analysis of beam-columns in uniaxial bending.
British Standards Institution, dealing with the design of Many of the methods have been described in detail in
steel structures, have adopted the European column reference 12. However, none of the methods given in
curves in slightly modified forms. The modification that text allow complete generality in the choice of the
results from attempting to fit a Perry-type formula to material stress-strain curve, or in selecting the shape of
each of the three curves q"~. Thus the failure stress as the cross-section. It is believed that the following
adopted in BS 5400 : Part 3 is given by method offers a rapid analytical tool for calculating the
ultimate load of a column of arbitrary cross-section,
f= ~ (fy + (I + /'/)f~r -- having arbitrary residual stress pattern, and arbitrary
material characteristics which may be different from
¢ I E + l + ,/)f.l'- - 4f,.f,.,}) (3) zone to zone within a cross-section.
#
MATERIAL CHARACTERISTICS
Any material characteristic can be specified. In the
implemented computer program, this is represented in
and a is assigned the following values: the form of a multi-linear curve as shown in Figure 2,
although in principle, any suitable stress-strain
Curve A a = 0.0020 characteristic can be programmed easily.
Curve B a = 0.0035
Curve C a = 0.0050. CROSS-SECTIONAL SHAPE
Curves A, B, and C are the Perry-fit curves Itis assumed that the cross-section can be built up from
corresponding to the European curves, a, b, and c. a number of quadrilaterals, as illustrated in Figure 3.
The present draft of BS 5 4 0 0 : P a r t 3 (February No further subdivision of the quadrilaterals is required
1981) suggests that the appropriate curve can be
chosen on the basis of the value of r/y,, where y, is the
distance of the centroid from the extreme fibre in com-
pression. Thus,
r
0 <-- < 0.45 CurveC
)'c
r
0.50 < - - < 0.60 CurveB
Y¢
/ •
r
0.70 < - - Curve A
)'c
On this basis most, but not all, structural hollow sec-
tions manufactured by the British Steel Corporation
would be designed using curve A. The remaining,
mostly circular hollow sections, for which the r/y¢
values fall between 0.60 and 0.70, would require
interpolation between curves B and A.
In what follows, the validity of these proposals is
examined in the light of results obtained from a Figure 3 A n arbitrarily shaped eross-sectian idealised b.v a series
rigorous theoretical analysis. o f quadrilaterals
38 Journal of Constructional Steel Research: Vol. 1, No. 4: September 1981
structures, where complex residual stress patterns are Figure 5 Strain and stress distributions at a cross-section with
found. inelastic material properties
j f y dA = M. (5)
at a discrete number of points, for increasing values of
the applied loading. With increasing amount of inelastic
stresses developing within the column, the stiffness of It is clear that M and N are direct functions of ¢ and d,.
the column progressively reduces until, just before By varying d,, one can easily adjust N, so that it equals
collapse, it completely vanishes. The load correspond- the thrust applied at the column ends. Thus by altering
ing to the final deflected shape so obtained is taken as the position of the neutral axis at points along the
the ultimate load. length, it is possible to ensure that force equilibrium
The calculation of the deflected shape of the column throughout the column is maintained. This on its own
in equilibrium with some applied loading requires two does not ensure that moment equilibrium along the
basic algorithms. The first relates to the calculation of length has also been achieved. Only when the correct
the stress resultants within the section for an assumed equilibrium shape has been achieved will the curvatures
strain distribution over the cross-section. This phase is # be such that moment equilibrium throughout the
often referred to as the calculation of the moment- column length is also satisfied. This is discussed in
thrust-curvature relationship. The second algorithm detail later.
deals with improving the values of the assumed deflec- The immediate problem being posed is that of the
tions, which directly relate to the section strain distribu- evaluation of the integrals in equations 4 and 5 when
tions, so that internal stress resultants approach the cross-section is of arbitrary shape and, in addition,
equilibrium with the external forces and moments at the material stress-strain characteristic is also arbitrary.
convergence. For this phase, a variety of iteration Providing that the cross-section can be replaced by a
methods can be adopted. In this paper a rapidly con- series of quadrilaterals, each with its own residual stress
verging Newton-Raphson procedure has been pattern, use can be made of two-dimensional Gauss
described. quadrature together with some mapping using the so-
called natural coordinates, to obtain the desired
integrals.
MOMENT-THRUST-CURVATURE
RELATIONS GAUSS Q U A D R A T U R E
These relations are best understood by referring to Gauss quadrature formulae enable the evaluation of a
Figure 5. If it is assumed that plane sections before definite integral between the limits - 1 and + 1 by a
bending remain plane after bending, the strain distribu- process of weighted summation of the values of the
tion in the section is uniquely determined for any posi- integrand at prespecified points. Thus:
tion of neutral axis, d~. By referring to the material PN
stress-strain characteristic, one can easily deduce the f(O = 5" H,
stress-distribution in the section. i=1
Figure 4 Idealisation o f a circular hollow section, showing a I_ / " f(,, r/) d,~dr/ = Z "=
Z H.,Hif('~i,r/.," (6,
typical circular segment replaced by an equivalent trapezium . I I i=IJ 1
Design of Circular and Rectangular Hollow Section Columns 39
The problem of evaluating an integral over an function at any specified coordinates (~,r/). For
arbitrary quadrilateral can be successfully solved by example, if the residual strains at the four vertices p, q,
mapping the quadrilateral area to the required square r and s are Rp, Rq, R, and R,, respectively, then the
area by the following transformations: strain R at any other point can be found by the equa-
tion:
x = ÷[(l-¢')(|-~)x, + (| + 0 ( | - ~ ) x q
t ( l - O ( l - - / 7 ) R p + (l + O(|--r/).Rq
+ (l-0(l+t/)xr + (l+0(l+r/)x,[ (7)
•4'-( | - - ~ ' ) ( | "4- ?])R r + (l +~(l 4-~)).R,] (1o)
y = ¼[(1-0(I- r/)y, + (1 + 0 ( 1 - r/)yq
t C O M P U T A T I O N P R O C E D U R E FOR
+ (l-O(l+r/)y, + (l+0(l+r/)Y,I (8) M O M E N T - T H R U ST-CU RVATU R E
RELATIONS
The integrals appearing in equations 4 and 5 can be
Figure 6 illustrates the manner in which the transfor-
evaluated separately for each of the quadilateral areas
mation is affected. It is clear that over the quadrilateral
used for idealising the original cross-section. For each
area pqsr the natural coordinates have the limits - 1
quadrilateral, once the number of Gauss points m is
and 1. The elemental area dx dy can be transformed to
fixed, the coordinates of the Gauss points, and hence
the area d~dr/ by the relation
i the product IJIH,Hj as well as the interpolated residual
ely dy = IJI d~ dq strain at each Gauss point, are also fixed and remain
where IJI is the determinant of the Jacobian matrix the same throughout the analysis. This can offer con-
J given by siderable savings in computer time.
For any chosen location of the neutral axis, the
integration procedure, then, consists of five basic steps
:_- [_,,_.,
--(1 0 - - ( 1 + 0 (1--0(1+ ix-.y./1 for each Gauss point:
E Q U I L I B R I U M DEFLECTED S H A P E
Figure 7 shows the deflected shape of the beam-column
under end moments and an end thrust. Also shown is a
statically equivalent system where the end moments
have been replaced by the effective eccentricities of the
applied thrust. At any stage of loading, the initial
deflected shape represented by u0 at a distance z from
the node A, undergoes changes to enable equilibrium to
4;=-1 be maintained. If u, is the net increase in the column
deflection, the total moment a: point ,: becomes
M = N(e + uo + u.) (11)
= N(e + u)
M
or, u - N e. (12)
Figure 6 A general quadrilateral and its natural coo;'dinate system
40 Journal o f Constructional Steel Research: ~'bl. 1. ~b. 4: September 1981
A0,
- - = 0 for r < s - - I and r > s+ 1
Au~
dZu
d2u, d2tt0 CONVERGENCE
# = ------U
dz = dz" dz 2 (13) Having obtained an improved approximation to the
true solution by using equation 17, the process is
It follows that the moment M is a non-linear function repeated by replacing {u k} by {u k' =} until satisfactory
of u, assuming N to be a constant at any stage of convergence is obtained.
loading. A direct solution for the deflected shape is At this stage a question can be raised as to whether
virtually impossible, keeping in mind the arbitrary the Jacobian K should bc evaluated for each cycle, or
stress-strain characteristics assumed in this analysis. A whether the modified Ncwton-Raphson method can be
numerical method of solution, therefore, has been adopted by using previously calculated Jacobian
adopted. matrices. In the author's experience, the modified
The procedure is based on sub-dividing the column Newton-Raphson m e t h o d s o m e t i m e s leads to
into n equal segments each of length h. The equilibrium divergence, and normally offers negligible savings in
is then satisfied at the ( n - l) internal nodes, labelled, computing effort. Hence tile recommendation is to
I, 2 . . . . . ( n - I). At any station s, the curvature can be evaluatc the Jacobian at each cycle of iteration. For
calculated using the finite difference approximation: most cases, providing use is made of deflected shapes
0~ = - - ( u ~ _ t - - 2u, + u~.. O/h z. (14) calculated for previous levels of loading in predicting a
suitable trial deflected shape, convergence would be
In view of equation 12, one can write obtained within 2:~-4 cycles of iteration. If convergence
u~ - U(u~ -i, u,, u~.t) (15)
is not obtained in, say, 8 cycles, it is advisable to reduce
the loading increment by a multiplying factor of about
or even more generally, as 0. I, and then repeat the process.
{u} = {U(u I, u 2. . . . . u,, i)l. (16)
The generalised Newton-Raphson method for a system STABILITY ANALYSIS
of non-linear equations suggests that, if {U~} represents If v is defined as a load factor on the initial loading {t0},
an approximate solution to equation 16, a better the structure is analysed for varying end loads {/'].}
approximation {u~ ,i} can be obtained by the following given by
equation:
IF,} = I(v- ~)G + II I/%} (20)
{u~ ' ' } = I u~} -- (I -- K) ' { u ~ - U ~} (17) where I is an identity matrix, and G is a diagonal
matrix, tile elements of which are 1 or 0 depending
where I is an identity matrix and K is a Jacobian matrix
upon whether the corresponding load component
defined by
changes with v or not. The highest value of v for which
? Ui AU, an equilibrium shape can be obtained would be the load
K ~ = I ?uj
= At----~j" (18)
factor corresponding to the limit state of collapse.
The incremental definition is of more relevance
where, as in the present case, the derivatives have to be V A L I D A T I O N OF THE P R O P O S E D M E T H O D
calculated numerically. The elements of the Jacobian K Results o b t a i n e d f r o m a c o m p u t e r p r o g r a m
can be evaluated efficiently by first maintaining Au, the T U B Q O L S , based on the method described above, are
same at all stations. From equation 14, one observes now compared with experimental results from two
that: series of tests.
Design of Circular and Rectangular Hollow Section Columns 41
Mean 1.061
42 Journal of Constructional Steel Research: l~bl. 1, No. 4: September 1981
Table 2 Comparison of analytical results with experimental results(16) for rectangular hollow section columns
Failure end moments (kN m)
Section N/,Vy Ratio
Analytical Experimental (3)~ (4)
Mean 0.988
Recommendations, ~ however, specify that all circular reduce the ultimate capacity for all cases by a small
sections can be designed using curve 'A'. A parametric amount. The failure loads obtained are, however, still
study was, therefore, undertaken to examine the situa- above BS 5400 : Part 3, curve 'A', although just so, and
tion. well above the European curve 'a'. It therefore follows
Four sections were chosen having a range of r/.v, that it would be quite appropriate to design all circular
values from 0.610 to 0.682. All the sections were sections using the BS 5400 : Part 3, design curve 'A', or
analysed for varying effective lengths so as to cover the by the European curve "a'.
full range of slenderness factor 2 from 0 to 2.5. In all
cases an elastic-plastic stress-strain curve, with a yield P A R A M E T R I C S T U D Y ON RECTANGULAR
stress of 355 MPa and elastic modulus of 200000 MPa, H O L L O W SECTIONS
was adopted. No residual stresses were considered but The value or r/.G for the rectangular sections
it was assumed that the columns had an initial out of manufactured by the British Steel Corporation varies
straightness of L/1000. The results are given in Table 3 from 0.671 in major axis bending to 0.890 ira minor axis
and are also shown in Figure 8. It will bc seen that all bcnding. Only five sections have values of r/v~ Icss than
results lic well above the European curve "a'. To 0.70, and hence would be affected by this cut o f f point
investigate the effect of residual stresses, an axisym- for BS 5 4 0 0 : P a r t 3, curve 'A'. All these sections
metric residual stress pattern w i t h f = 0 . 2 f (Figure 4) togcthcr with a section with r/)L. value of 0.738, have
was adopted Ibr the 48.3 x 5.0 section (r/y¢ = 0.638) been analysed assuming an initial lack of straightness
and the 457 x 40.0 section (r/y, = 0.648). The results of 0.001 times thc length and no residual stress, for a
arc given in Table 4 and arc also shown in Figure 9. It range of column lengths, again covering the full range
will be seen that a residual stress of 0.2 f tends to of slenderness factor from 0 to 2.5. Material propertics
Table 3 Analytical results for circular hollow section columns compared with ECCS curve "a' and BS 5400 : Part
3, curve 'A'
Column Ultimate failure loads (N/Ny)
Sectkm length
(ram) ECCS curve 'a' BS curve "A' Analytical
Table 4 Analytical results showing the effect of residual stresses on the ultimate failure loads of circular hollow
section columns
Table 5 Analytical results for rectangular hollow section columns compared with ECCS c u r v e ' a ' a n d BS 5 4 0 0 :
P a r t 3, c u r v e ',4.'
150 × 100 × 6.3 section was adopted. This pattcrn is curopccnncs de flambcmcnt.' Construction Metallique.
September 1970. 37-57.
an idealised version of the pattern reported by Estuar 3 BI~,I[ISI[ SI'ANI)-~RI)% [NSIIIUHON Use of SIruct.ral Steel in
and Tall. ~;~ The results given in Table 6 show that tile Baihtings. London. BS 449 : Part 2, 1970.
pattern of residual stress adopted has a negligible effect 4 I|RlrlSII SIANI).kR[)% INSFIIUTION COt'NCII. I:OI/. C O l ) I S I')1:
Draft British Standard Code o f t'ractice f o r the Use
I'll.A(' ll~'li
on the computed ultimate failure loads. ¢y" Structural Steel in Buihlings. t . o n d o n . BS 5950.
CONCLUSIONS T E N S [ LI'I
31":.~ [ D U A L S'I'[{ES.~
An analytical method has been presented which makes
Y~[I H P a
it possible to analyse sections of arbitrary shape in
compression and uniaxial bending. Arbitrary material
!
F l
C051IlRE:-;S [ VE
i~;i'X"; I DL'AI. STRI':5~'J
IO ~,lt':i bit)
,+
V------ I,)
£
Table 6 Analytical results for 150 x 100 x 6.3 section with and without residual stresses
Ultimate failure loads (N/.Vy)
Column
length 2 ECCS curve "a" BS curve "A" Analytical Analytical
(ram) without with
res. Stress res. Stress
5 BRITISH STANDARDS INSTITUTION. Steel Girder Bridges. 13 ZIENKIEWICZ. O. C. The Finite Element Method. McGraw-Hill
London. BS 153 : Part 3B and 4. 1972. Book Company, London, 1977.
6 BRITISH STANDARDS INSTITUTION COUNCIL FOR CODES OF 14 WAGNER.A. L.. MUELLER. W. H. and ERZURUMLU. H. 'Ultimate
PRACTICE. Draft British Standards Code of Practice for Steel. strength of tubular beam-columns." Proceedings of American
Concrete and Composite Bridges. London. BS 5400 : Part 3. Society o f Civil Engineers. Journal of the Structural Division.
February 1981. January 1977. 9-22.
7 AYRTON.W. E. and PERRY. t, "On Struts.' The Engineer, 1886, 15 SCttULZ. G. "Der Einfluss yon Querlasten auf die Traglast
464. schlamker, imperfektionsbehaltcr Druckst~ibe aus Baustahl."
8 TALL. L. Stub-Column Test Procedure. Fritz Engineering Der Stahlbau. April 1971, I I I - I 18.
Laboratory Report No. 220A.36o February 1961. (Also given 16 DWYER.T. J. and GALAMIIOS.T. V. Plastic behaviour of tubular
in reference 103 beam-columns." Proceedings o f American Socie(v o f Civil
9 ENGESSER. F. "Ueber die Knickfestigkeit gerader St/ibe.' Engineers. Journal o f the Structural Division. August 1965,
Zeitschrift fur A rchitektur und lngenieurwesen, 1889, 455. 153-68.
I0 JOtlNSTON. B. G. led.). Guide to Stability Design Criteria for 17 ESTUAR. F. R. and TALL. I,. The column strength of hot rolh'd
Metal Structures (3rd edn.). John Wiley and Sons, New York, tubular shapes - ,,In erperimcntal evaluation. Fritz Engineer-
1977. ing Laboratory Report No. 296.1, April 1965.
II DWIf;IIT. J. B. Use of Perry formula to represent the new
European strut curves. Proceedings, Colloquium on Centrally
Compressed Struts, Paris, 1972.
12 CIII-N.W. I:. and A r s u r A . r . Theory of Beam Columns. Volume Contributions discussing this paper should be received by the Editor
I. McGraw-Ilill Book Company. New York. 1976. before I January 1982.