Sei sulla pagina 1di 8

6980 DISCUSSION

Computation of failure loads of


composite columns
A. K. BASU

Dr. D. S. Bondale, Fdeernan. Fox and Partners


I would like to congratulate Dr Basu on theway he has adapted forhigh speed digital
computer use the general theoretical approach to theproblem presented earlier in my
thesis.lB Further credit is also due to him for checking the validity and accuracy of
the theory against the results of the B.R.S. full-scale encased stanchion tests.
45. Dr Basu suggests in equation(2) an interesting form of a stress/strain curve of
possible use to represent the stress/strain characteristics of various steels, and yet for
computational purpose the usual elasto-plastic stress/strain curve has been retained.
It may perhaps be of interest to use equation (2) instead and study the effect of para-
meter j3 (say, with equal to 5, 10 and 50) on the column failure loads. The usual
elasto-plastic stress/strain curve assumes complete absence of residual stresses across
the steel section (and this hasbeen assumed in the theory presented); however, since
residual stress pattern is quite common in structural steel rolled and other sections
(and much more so, in U.C. and U.B. sections) it may be worthwhile to consider
equation (2) with values of 9, appreciably less than 100, for assessing accurately the
likely extent of the detrimental effect of residual stresses on the ultimate load-carrying
capacity of bare and encased columns. This should not prove difficult since the
present programme would accept the alternative stress/strain curve.
46. Referring to 8 39, it is refreshing to note that rather thancontinuing focusing
greater attention on isolated (pin-ended) columns,extensions to composite frames of
recent British and American proposals for design of bare steel frames based on ulti-
mate strength are currently being considered at Imperial College. Results of this
study would be of considerable help in understanding the behaviour of frames, both
clad and unclad.
47. Perhaps it may be relevant to mention here that the earlier work which I
carried out atImperial College, to which Dr Basu's work forms a natural sequel, has
recently been published1' inthreeparts. Briefly, Part 1 deals with the general
(comparatively) rigorous theoretical treatment of the column stability problem to-
gether with approximate design criteria based on (U) the concept of a limiting strain
at which column failure occurs, e.g. the first yield criterion in steel and (b) the inter-
actionformula/curve approachfor beam-column members. Part 2 presents the
application of the general theory to composite columns. Results of calculations for
column failure loads madeaccordingly are provided, together with those obtained in
accordance with B.S.449 (with and without suggested modifications) and C.P.114.
A brief discussion of various parameters involved and of a possible extension of the
proposed theory to frameworks then follows. Part 3 describes the tests carried out
on sixteen encased columns and concludes with a correlation between the theory and
the experiments, which is generally found to be good.
Mr A. N. Procter, Building and Civil Engineering Dept. Nottingham Regional
College of Technology
The Khrmhn-Engesser column theory was published over half a century ago.''
Since that time several important series of column tests have been carried out and
Paper published: Proc. Znrtn civ. Engrs 1967,36 (March) 557-578.
Delivered by ICEVirtualLibrary.com to:
11 305
IP: 180.191.74.12
On: Sat, 09 Oct 2010 02:30:39
DISCUSSION

l r
Mean E = 2170 000 kg/sq. cm.
l

I
4000

200[

I OM

50 75
I00
Slendernelr rario:7/r
I50
!
I75

Fig. 11. Kdrrnfrn‘s struts

considerably more information about the behaviour of compression members has


become a~ai1able.l~Indeed KBrmBn’s own tests may be used to illustrate modern
column mechanics, as illustrated inFig. 11, where the whole seriesis plotted in terms
of intensity of loading (kg/sq. cm) and slenderness ratios, I/r and I/b. It should be
remembered, however, that his tests were carried out on very small sectionstruts that
could notbe expected to give a reasonable representationof the behaviour of full-size
composite steel and concrete columns of practical size.
49. KfrrmBn’s struts 1 to 7a and b inclusive follow the Euler curve and represent
long struts or long columns in which almost all of the energy was absorbed in the
elastic deflexion, so the axial strains and lateral distortion may be ignored, as in the
Euler theory. Beyond the Euler limit a second phaseof medium-length columnsfalls
Delivered by ICEVirtualLibrary.com to:
306
IP: 180.191.74.12
On: Sat, 09 Oct 2010 02:30:39
6880
short of the Euler loads. It was in this phase that Hodgkinson noted that columns
failed by deflexion and partly by distortion of the cross-section.aO The remainder of
the diagram,CDE, is a typical curve obtainedin tests on cylinders or prisms of varying
length/breadth ratios, showinga critical point, D, beyond which most of the energy is
absorbed in direct compression,”’ although thefinal deflexions of 15a, 16 and 17 were
very large after the material had passed the yield point, as might be expected.
50. It was shown by Burraa andothers that in composite columns there is some
relative movement of the steel and concrete before failure, due possibly to creep. It
follows that a theory of the strength and stability of composite columns based on
completely compositebehaviour isnot realistic. Modernmethods of designing
columns ofsteel and concrete, whether reinforced or composite, are based on indepen-
dent action under compression. This permits the steel and the concrete to develop
their full strength before the concrete begins to crush or split along the compression
face. Full-size tests on steel and concrete encased columns show that this is what
actually happens at failure.lO-aa
51. The behaviour of a composite column under combinedor eccentric loading in
a testing machine is quite dissimilar from that of a small mild steel strut. At a
relatively small load therewould be tensile cracks formingon oneside of an eccentric-
ally loaded composite column which would become more numerous and wider as the
load increases. At the pointof failure the concrete on thecompression face suddenly
cracks and crushes, so that thecarrying capacity of the column is instantly reduced by
an appreciable amount. The sequence of events is quite different from KhrmLn’s
observations of his strut tests, so that his theory is unlikely to be suitable for the
present purpose.
52. The eccentric loading tests on the composite columns tested by Stevens at the
Building Research StationIO followed rather the same course as described above.
The tests at Imperial College were not made onconcrete encased steel, but on small
rolled sections encased in cement grout. It does not appear that control tests were
made either in compressionor in bending about either the minor or major axis of the
core. Indeed no values are quoted, either experimentalor calculated, of the moments
of resistance of the mortar-encased steelwork. It would be of interest to know what
the values, or thevalues which were assumed by Dr Basu, were for thepurpose of his
calculations.
53. The curve of major-axis bending in Fig. 4 requires perhaps an explanation
which does not appear to be forthcoming in thetext. It is assumed that the thrust on
the column has the sameeffect as a pre-stress on a beam, resulting in an increase in
the moment of resistance of the column. I venture to give the following theoretical
representation,based on an earlier example.a4 Assuming the dimensions of the
casing are 6, breadth, and d, depth, and the area of the R.S.J. is A , in an eccentrically
loaded column, there is a bending moment, M,, resisted by the steel core, causing a
bending stress, f b . The thrust in the concrete, F,, equal and opposite to a tensile
force, 6,in the steel, produces a composite bending couple, F,X a, where a is the lever
arm. Ft =ftX A and F, may be assumed with reasonable accuracy to be
X f d = L‘
-bd,
3
where fc’ is the crushing strength of the concrete. There isan additional thrust in the
steel, F,= W, the load on the column, giving a stress, fc= W/A, the concrete being
assumed fully loaded.
54. The forces and couples may be plotted on a diagram similar to Fig. 12. At
the yield point of the steel, f , =h+ft-f c so that
h = f,+f,-ft . . . . . . . .(16)
Assuming no tension in theconcrete, the totalmoment of resistance of the composite
Delivered by ICEVirtualLibrary.com to:
307
IP: 180.191.74.12
On: Sat, 09 Oct 2010 02:30:39
DISCUSSION

CROSS SECTION STRESSES STEEL CONCRETE STRESSES

Fig. 12. Composite steel and concrete columns under combined loading

member, M1= M , + Fca and a is approximately &d, as shown in the diagram. So in


terms of stresses, 2, being the section modulus of R.S.J.,
Mt = fbz, +y bda.
fc'

Substituting for fb from formula (16) above, at failure


Mt = cf,
+fc -f;)Zx +L' bda.
Since fi = F,,f,= g3 bd/A and as fc= W/A,

The free resistance moment, MO,


is approximately

so the moment of resistance of the composite column under bending moment and
thrust Mt = MO +F and the interaction formula becomes
We
-W+
P MO+ WZJA =
. . . . . . . (18)

This gives results of accuracy comparable with the formula for steel columns.

Dr P. K. Neogi, Civil Engineering Department, Imperial College


The Author is to be congratulated on an important contribution to the literature on
composite columns. The work of Bondale showed that the assumption of full inter-
action between the steel and the concrete is justified and that the behaviour of com-
posite columns is essentially similar to that of reinforced concrete columns. The
Authorhas now shown that the analysis of composite and reinforced concrete
Delivered by ICEVirtualLibrary.com to:
308
IP: 180.191.74.12
On: Sat, 09 Oct 2010 02:30:39
6980
columns can be unified, and has given an efficient method for calculating the critical
length-eccentricity relationships of a section for a given maximum load.
56. The Author’s method of calculatingthemoment-curvature curves, i.e. the
M-+P curves, is essentially similar to the one used by Changa5 and Pfrangae for
rectangular reinforced concrete columns. The innovation lies in directly determining
thecritical length/eccentricity relationships from the modified moment/curvature
curves, avoiding the calculation of the load/deflexion curves. For a given section,
once the M-+-P curves have been generated for all the specified loads, large numbers
of critical length/eccentricity relationships corresponding to these loads can then be
calculated.
57. The procedure becomes much less efficient, however, if it is used to calculate
the load/deflexion curve and the maximum load of a column having a given section,
length and eccentricity. I have showna7that thisproblem canbe solved more
efficiently by calculating the axial loadvalues for aseries of equilibrium shapesdefined
by increments of curvature at thecentral cross section. If the exact deflected shape is
used, it is determined by integrating the curvatures along the column length; if the
deflected shape is assumedto be part of a cosine wave, the central lateraldeflexion can
be related to the centralcurvature. The complete load/deflexion curve of the column,
including the falling branch, can be determined by this method.
58. For pin-ended columns with equal eccentricity of loading at both ends, the
Author is justified in assuming the deflected shape to be part of a cosine wave. My
own work has shown that for concrete-filled tubular columns this assumption lowers
the maximum load by less than 5%. The load/deflexion curves for the two cases are
virtually identical up to about 80% of the maximum load.
59. I would like to know, firstly, if the Author considers the assumption of a part
cosine wave deflected shapejustified in thecaseofunequalendeccentricities. Secondly,
can end restraints be taken into account? Thirdly, can this approach be extended to
include biaxial bending? Lastly, what approach does the Author envisage for the
practical design of encased composite columns?

Dr A. K. Basu
I would like to thank the contributors for their discussion of the Paper. I agree with
Dr Bondale about the importance of the effect of residual stresses in rolled steel
sections, but would point out that he is mistaken in thinking that the stresslstrain
curve given by equation (2) could be used for studying this effect. For calculating
the moment/curvature relationships of the section, the stress/strain characteristics of
the individual fibres are required, which remain bilinear even in the presence of residual
stresses. This should not be confused with the overall stress/strain characteristic ofa
specimen taken from the section which may indeed appear to be rounded, owing to
the fibres reaching yield at different stages of applied loadingbecause of the essentially
non-uniform distribution of the initial (residual) stresses.
61. It should be noted here that the residual stresses are not the only imperfection
present in the columns; initial curvature of the column centre-line and initial end
eccentricities are also likely to occur. However, in view of the uncertainties in the
magnitudes of the various imperfections, design rules are usually based on a suitably
chosen value of only one of these imperfections. The imperfection considered in the
A.I.S.C. specification, for example, is the presence of residual stresses, whereas that in
B.S.449 and also inthe present treatmentis the initial curvature of the column centre-
line. In Fig. 13 the C.R.C. basic strength curve,28 on which the A.I.S.C. rules are
based, is compared with the Perry-Robertson curve1. It is evident fromthe
figure that theeffect on thefailure Ioad ofthe assumption of an initial curvatureof the
magnitude given in B.S.449 isgreater than that of residual stresses. Fig. 13 also
shows thatthe failure loads of axially loadedcolumns with initial curvature as
Delivered by ICEVirtualLibrary.com to:
309
IP: 180.191.74.12
On: Sat, 09 Oct 2010 02:30:39
DISCUSSION
1.0

09 CRC basic strength curve (Ref.26 - Equation (19)


Maximum residual stress U = 0 5 U )
80 Y
08 Perry Robertron curve (Ref. I : Appendix B)

07 Present solution : B in. X 8 in. X 31 Ib U.C.


E = 3 0 x 10' Iblrq. in.
= 33WO Iblsq. in.
- 06 U,,

-i

l&=

i
05

04

03

02

01

0
Llr

Fig. 13. Failure load of axially loaded steel columns

obtained from the present computer programme are in good agreement with the
values of first-yield load given by the Perry-Robertson formula.
62. Mr Procter has ignored the evidence of agreement between theory and experi-
ment presented in the Paper in stating his opinion that von Khrmhn's column theory
is not suitable for composite columns and that the theory based on full composite
action is notrealistic. I do not follow the arguments he puts forward in support of
his opinion. The theory presented in the Paper takes into account the phenomena
described by Mr Procter, namely, the tensile cracking and the crushing under com-
pression of concrete in composite columns. His objection to thetheory is not, there-
fore, valid.
63. Mr Procter has also discussed the phenomenon of the maximum ultimate
moment of resistance of composite and reinforced concrete sections occurring at a
non-zero value of axial load. A quantitative explanation of this phenomenon will in
general require numerical workof the kind presented in the Paper;instead, a qualita-
tive explanation is given here.
64. A section of plain concrete, assumed to be a perfectly brittle material, has no
moment-carrying capacity both at zero axial load and at the squash load; somewhere
in between these twolimitingloads, themoment-carrying capacityreaches a maximum.
When an areaof steel is added to thesection to act compositely with it, the complete
section retainsthe characteristics of its concrete componentto a degree which depends
on the relative area of the steel and its distribution in the section. The theoretical
explanation offered by Mr Procter is, however, fundamentally incorrect. The stress
distributions assumed for the steel and theconcrete in Fig. 12 cannot coexist as they
correspond to twocompletely different and arbitrary strain distributions and curvature
values in thesection.
65. I am glad to note thatDr Neogi's independent theoretical investigation, using
a different approach from that given in the Paper, also confirm the validity of the
part-cosine curve assumption for computing not only the failure loads but also the
deflexions of the column at or near collapse.
Delivered by ICEVirtualLibrary.com to:
31 0
IP: 180.191.74.12
On: Sat, 09 Oct 2010 02:30:39
6980
66. In reply to his specific questions, the partcosine curve assumption is alsovalid
for the case of unequal end eccentricities. If the endeccentricities are e and Be, where
p is the eccentricity ratio (- 1 < p < + l ) , it can be shown that equation (10) takes the
following generalized form:
[(l+j3)a tan4X-(l-j3)a
tan X (1 + +
tanaX (1 - I - 2(dYm/d4)
- (ym/@- (dy,/d+)
. . (10)
67. Having evaluated the right hand side of the above equation from they, =f(+)
curve (see Q 22), values of X for a set of chosen values of j3 can be obtained by iteration.
The critical length of the column is.againgiven by equation (9) and theeccentricity is
obtained from the following equation, which is a generalized form of equation (8):
e = ym sin 2h/~(1-fl)a+4flsinzX . . . . . . (8’)
68. Fig. 14 shows that the results for the major axis bending of an 8 in. X 8 in. X
31 Ib U.C. section (see 0 28) with fl=O, obtained by the present method agree well
with the available exact solution8.
69. I think that the effect of linearly elastic end restraints can be studied by an
approximate method similar to the one given in the Paper, although the accuracy of
the solution may not be satisfactory. Extension to the case of general biaxial bend-
ing, however, appears to be difficult at the moment.
70. Referring to the question of a practical design method, work is in progress at
Imperial College under the sponsorshipof the CERAwith a view to formulating such
a method for encased stanchions and concrete filled tubes of rectangular section.
Preliminary investigations indicate that none of the methods currently available for
the design of steel or reinforced concrete columns is suitable for composite columns.

IJr=

Fig. 14.
Delivered by ICEVirtualLibrary.com to:
31 l
IP: 180.191.74.12
On: Sat, 09 Oct 2010 02:30:39
DISCUSSION
This is because the sectional properties of composite sections can vary widely. It has
been established that the reduction factor due to slenderness is not independent of
eccentricities. Neither of the simple definitions of slenderness used for the design of
steel or reinforced concrete columns appears to be suitable for composite columns.
It has also been established that the straight line interaction formula used in steel
design is not applicable to composite sections, being unsafe for slender columns and
too conservative for short columns. Incorporation of a magnification factor on the
moment to take into accounttheadditional eccentricity due to deflexion further
reduces the accuracy for short columns.
71. The approach currently being examined involves the assumption for the inter-
action curve of an equation with a number of unknown parameters, the ‘best’ values
of which for a chosen set of slenderness and eccentricity ratios are being determined
by the computer for a range of column sections. No firm conclusions have been
reached to date.

References
16. BONDALE D. S. The effect of concrete encasement on eccentrically loaded steel
columns. PhD Thesis, London University, June 1962, p. 247.
17. BONDALE D. S. Columntheory with special reference to composite columns.
Consulting Engr, London, 1966. (July) 72-77; (Aug.) 43-48; (Sept.) 68-70.
18. KARMANT. VON. ‘Untersuchungenuber Knickfestigkeit.’ Mitt. ii Forschungs-
arbeiten, 1910, 81; republished in Collected Works of Theodore von Karman,
Vol. I, 1902-1913. London, Butterworths Scientific Publications, 1956, pp.
9CL-140; see also pp. 24-35.
19. PROCTER A. N. Buckling under complex loading. Engineer, 1956, 201 (June)
629-632 and 667-669.
20. HOD~KINSON E. Experimental researches on the strength of pillars of cast iron
and other materials. Phil. Trans roy. Soc., 1840, 130, Pt. 2, 388.
21. PROCTER, A. N. Elasticmaterialsunder axial loading. J . Franklin Inst., 1958,
265 (Feb.) 125-143 and Engineering, 1959, 187, 798-800.
22. BURRW. H. Composite columns of concrete and steel. Engrs, 1912,188 (Aug.)
114-126.
23. FABER, 0. Savings to be effected by the more rational design of cased stanchions
as a result of recent full size tests. Struc. Engr, 1956, 34 (March) 88-109.
24. PROCTER A. N. Discussion: Ultimate load design of concrete structures. Proc.
Znstn cio. Engrs, 1962, 23 (Oct.) 225-221.
25. CHANGW. F. and FERGUSON P. M. Long hinged concrete columns. J. Am.
Soc. cio. Engrs, Proc., 1963, 60 (Jan.) 1.
26. PFRANG E. 0.and SIESSC. P. Analytical study of the behaviour of long re-
strained reinforced concrete columns subjected to eccentric loads. Structural
Research Series No. 214, Civil Engineering Studies, University of Illinois.
27. NEOGI P. K. Concrete filled tubular columns. PhD Thesis, University of
London, 1967.
28. Guide to design criteria for metal compression members. Column Research
Council. London, CrosbyLockwood, 1960.

Delivered by ICEVirtualLibrary.com to:


31 2
IP: 180.191.74.12
On: Sat, 09 Oct 2010 02:30:39

Potrebbero piacerti anche