You are on page 1of 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/241060220

A design method for concrete-filled, hollow section, composite columns

Article · November 1997

CITATIONS READS
26 2,357

2 authors, including:

Yong C Wang
The University of Manchester
307 PUBLICATIONS   5,349 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Probabilistic distributions for use in the fire safety design of residential buildings View project

Characterisation And Multi-Scale Modelling Of Aged Concrete Under Extreme Loading Conditions View project

All content following this page was uploaded by Yong C Wang on 20 June 2014.

The user has requested enhancement of the downloaded file.


Paper: Wang/Moore

Paper

A design method for concrete-filled,


hollow section, composite columns I

Y.c. Wang, BEng, PhD, CEng, MIStructE


Building Research Establishment

D. B. Moore, BEng, PhD, CEng, MIStructE


Building Research Establishment

Synopsis BS 5950: Part 1’ (referred to as the ‘BS 5950 method’) contains a design
In this paper a simple design procedure is presented for concrete- procedure for steel columns that is safe, easy to use, and accepted by design-
filled composite columns suitable for manual calculations, based ers in the UK. Although in its present form this procedure cannot be used
on the recommendations given in BS 5950for bare steel columns. for the design of concrete-filled columns, this paper presents results which
Also presented is a series of tests on slender concrete-jilled show that, with a number of modifications, the BS 5950 method can be eas-
composite columns subjected to double curvature bending. These ily adapted to concrete-filled columns.
tests, together with acomprehensive range of published As part of the introduction and acceptance of Eurocode 4 in to the UK,
experimental data on composite columns, are compared with the the Building Research Establishment commissioned Professor Anderson at
proposed design procedure. The design procedure is also Warwick University to compare the recommendations forcomposite
compared with the traditional method given in BS 5400 and the columns in EC4 with those in BS 5400. From the results of these compar-
method described in Eurocode 4 Part l .l. The results of all of isons Anderson concluded that?
these comparisons show that the proposed method is
( 1 ) there is a significant discrepancy between the predictions from BS 5400
conservative, gives results comparable with those predicted by
and EC4;
BS 5400 and EC4, and can be used for design purposes.Howevel;
(2) this discrepancy is particularly large for slender columns with end
the simplicity ofthe proposed method and the savings in design
moments other than single curvature bending.
time are perhaps the major factors in its favour.
Unfortunately, because of the lack of experimental data on slender com-
Notation posite columns, Anderson could not resolve this problem. A series of tests
In addition to the symbols defined in BS 5950: Part I, the following sym- on slender composite columns was therefore carried out by BRE to resolve
bols appear in this paper: the reason for the discrepancy between the two Codes, to add to the avail-
ocu is theconcretecube strength able test data, and to help verify the proposed design method based on the
fck is the concrete characteristic strength (EC4) BS 5950 approach. The details of this test programme are described in full
p is the moment ratio in the column in this paper.
E,, is the concrete short-term secant modulus of elasticity (EC4) These tests, together with a comprehensive range of published experi-
Y is the material partial safety factor mental data on composite columns, are compared with the proposed
x isthe column relative slenderness = (Nu/N,,)”z BS 5950-based design procedure. The design procedure is also compared
with the two most up-to-date and best supported methods given in BS 5400
e is the column eccentricity
E is the material Young’s modulus and EC4. The results of all these comparisons show that the proposed
I isthe cross-section second moment of inertia method is conservative, gives results comparable with those predicted by
N,, is the column Euler buckling resistance BS 5400 and EC4, and can be used for design purposes. However, the sim-
Nu is the column squash resistance plicity of the method and the savings in design time are perhaps the major
factors in favour of the new approach.
Introduction
Concrete-filled composite columns consist of either a circular, square or a The proposed method
rectangular hollow steel section with a concrete core. Fig 1 shows some The proposed method is based on the recommendations given in BS 5950
examples of this type of element. They have many architectural and struc- for bare steel columns and is developed by replacing the properties of the
tural advantages, for example: bare steel section with those of the composite section.
In BS 5950, two checks are necessary: a local capacity check and an
- high load capacity with small cross-section dimensions overall buckling check. For each check, either a simplified or a more
- inherent ductility and toughness exact approach may be used. Equations for the four different combinations
- speed of construction when used as part of a composite frame in BS 5950 are as follows:
- good fire resistance
- good visual appearance - Local capacity check using the simplified approach:
F M, M
Although there are some practical concerns, in particular the joints between +-+L51 ....(l)
composite columns and other structural elements, their numerous advan- AgP, MC, Mcy
tages make them very attractive in a variety of applications’. It is therefore
surprising that they are so rarely used in the UK.
There are a number of standards in the UK that can be used to design con-
crete-filled columns. Theseinclude BS 5400: Part 5’ (referred to as the ‘BS
5400 method,’ in this paper) and Eurocode 4: Purr 1.I’ (referred to as the
‘EC4 method’ in this paper). The recommendations given in BS 5400 are
based on the work of Basu & Somerville’, modified by Virdi & Dowling‘,
while the EC4method was influenced by the CIDECT research programme.
However, both methods are complex and diffkult to use and don’t readily
lend themselves to hand calculation. Although there is readily available
computer software based on these two approaches, for conceptual design
and to develop a clear understanding of their structural behaviour it is some- Concrete
filled CHS Concrete
filled RHS
times desirable to have a reasonably accurate and simple manual approach. Fig 1. Typical examplesof concrete-filled hollow section

368 The Structural EngineerVolume 75INo 2 1 4 November 1997


Paper: Wang/Moore

TABLE l - Summary of test results The composite cross-section properties to be used in the proposed method
are:

I
Major axis Minor axis
Test load analysis - the squash load Nu= A, py in eqn (1)
no. eccentricity moment eccentricity moment - the plastic bending moment capacities M,, and Mcyin eqns (1)and (5)
(mm) ratio (mm) ratio - (Test
W (FEW
- the compressive resistance P,, and Pcy in eqn (5)
The values of Nu,M,, and Mcycan be determined in accordance with the rec-
RHS2 I 0 1 - I 55 I 0 I 246 I 218
ommendations given in BS 5400: Part 5’.
RHS3 I 55 I 0 I 110 I 0 I 172 I 161 To calculate the values of P,, and Pcy,the composite cross-section is con-
RHS4 I 55 I 0 I 110 I -1 I 238 I 223 verted into an equivalent steel section. The area A, and the second moment
RHS5 I 55 I -1 I 110 I -1 I 251 I 236
of inertia I of the equivalent steel section are:
RHS6 I 55 I 0 I 55 I 0 I 234 I 211 4=- N u
0.67E, l r,
andI=I,+I,- ...
RHS7 I 55 I -1 I 0 I - I 520 I 360 P y s 1Y s E, 15
RHS7* 55 -1 0 - 530 520 where the subscripts ‘S’and ‘c’ refer to steel and concrete components in
the compositesection. I, is thesecond moment of inertia of the uncracked
RHS8 55 0 0 - 480 340 concrete component. The factor of 0.67 relates the uniform rectangular
RHS8* 55 0 0 - 480 431 stress block to the concrete cube strength and is also used in BS 5400:
*column effective length= 3.2m. Part j2.
Column buckling curve ‘a’ is used when determining the column com-
pressive resistance P,, and Pcyin eqn (5).
- Local capacity check using the more exact approach for plastic and
compact cross sections: Tests on slender composite columns
Test set-up
A column length of 4m was chosen because this is typical of the storey
....(2) height in multistorey buildings. A hot-rolled 120 x 80 x 6.3 rectangular hol-
low section was selected to give a relative slenderness of about l .O.
- Overall buckling check using the simplified approach: Loading tests were carried out in an Amsler compressive testing machine
with a maximum load capacity of 10 000kN. A calibration of the test
F m
mM My, machine by the manufacturer confirmed that the Amsler had an accuracy of
+-+--I1 ....(3) +- 1% at the lower end of its range between 0 and 200kN and +- 0.5%
Ag P, Mb Py z y
between 200 and 1000kN.
- Overall buckling check using the more exact approach: This series of tests included two columns bent about the minor axis, two
columns bent about major axis, and four columns subjected to biaxial bend-
mM,
mM ing. The eccentricities and moment ratios for all tests are listed in Table 1 .
+ L 5 1 ...(4)
MayMax Since the Amsler was not designed to apply eccentric loading, a set of
adapter endplates was manufactured and fixed to both ends of each column
For hollow section steel columns, lateral-torsional buckling does not occur. to form a simply supported column under double curvature bending. Fig 2
Therefore: shows a typical arrangement for a column end support condition.
F F
I-- I--
Instrumentation
M , = M,, 0.5F and May = Mcy -
&X &Y
0.5F
....(5) Since the objective of these tests was to obtain column strengths, only a few
l+- l+- strain gauges and displacement transducers were placed on each column to
P,, CY obtain test data which may be used by others to conduct more detailed
This paper adopts the more exact approach in the proposed method for con- analysis of the behaviour of these columns.
crete-filled composite columns. However, since the values of M,., and Mry Three different sections (two quarter points and the centre) on each col-
in eqn (3) are not available for composite sections, eqn (l) is adopted for umn were strain-gauged and at each section four strain gauges were placed
the local capacity check. Therefore, the proposed method uses eqn (1) for around the cross-section. Fig 3 shows the positions of the four strain gauges
local capacity check and eqn (4) for overall buckling check. at each section.

1OOOmm I

t t’
looomm
- 1
l l
I I
t
/ / / / / / / / / / / / ~ / / / / / / / / / / / / (a) Along column (b) Cross-section
Fig 2. Column end support Fig 3. Strain guuge locutions

The Structural EngineerVolume75/No 21 4 November1997 3 69


Paper: Wang/Moore

550 T
450

350
h

E--
-20

-50l Deflection (mm)


-50 l Deflection (mm)

Fig 4 . Load deflection relationships: major axis Fig 5. Load deflection relationships: minoraxis

A total of 21 displacement transducers were used to measure the 3- The failure load for test 4 was 238kN. During this test, the two half-spher-
dimensional deflections at seven positions on each column. A general ical bearings at one end of the column started to bear against each other after
arrangement of three LVDTs was adopted, and the 3-dimensional deflec- initial column failure at large deflections, and therefore the results after this
tions along the three rectangular Cartesian co-ordinate directions are cal- load should be ignored.
culated from the measured deflections along these three directions*. , All eight columns were intended to be simply supported at both ends
about both axes. However, there was a variable degree of end restraint
Material properties owing to friction in the ball bearings. Since all columns failed by buckling
Grade 43A mild steel and C30 ordinary strength concrete were specified for about the minor axis, the column strengths for tests 1-6 were not affected.
each test. However, the real strengths of the test materials were much high- This is because there was substantial minor-axis bending moments in these
er than those given in eitherBS 5400 orEC4. columns, and these bending moments overcame the frictional resistance in
Four 10m steel hollow section columns from the same mill were cut into the ball bearings.
eight columns of 4m in length. From one off-cut, four steel coupons were made Tests 7 and 8 were subjected only to major-axis bending moments. The
and tensile tests
were carried out in accordance with BS EN 10002-1:1990.The very low level of minor-axis bending moments from secondary effects was
average yield stress was found to be 370N/mm2, and this value was assumed not sufficient to overcome the frictional resistance. Therefore, there was
for all the columns. In total, 27, l00mm standard concrete cubes were cast rotational restraint about the minor axis in each column. The result of this
on the day the columns were cast. Three concrete cube tests were carried additional restraint was to reduce the column effective length and to increase
out after 28 days for quality control. The average cube strength was its strength.
5 1N/mm2.Three concrete cube tests were also carried out on the day of each A finite element analysis computer program9was usedto estimate the col-
column test after approximately 6 months. The cube strengths obtained on umn effective length ratio for these two columns. This computer program
the day of each column test were very consistent (difference being less than has been developed to study the structural behaviour of steel/composite
1N/mm2), and the average cube strength of all 24 concrete cubes was frames under both the cold and fire conditions and has been extensively val-
60N/mm2. idated against a large number of test results’. An effective length ratio of 0.8
was found to give analytical results in good agreement with the test results.
Test results For comparison, the strengths for columns 1-6 were also predicted using the
The results from the eight column tests are given in Table 1. Figs 4 and 5 computer program. All the predicted column strengths using the finite ele-
show the lateral load-deflection curves measured at midheight for dis- ment analysis are given in Table 1.
placements parallel to the column’s major and minor axis directions, respec-
tively, for all eight tests. Comparison between various predictions and test results
In this section, the predicted column strengths using the following three
The failure mode for all columnswas elastic-plastic buckling about the
methods are compared with the test results:
minor axis. Except for test 8, which buckled suddenly about the minor axis,
all other columns failed in a gradual manner and there was very little resid- - the proposed method
ual deflection in the columns when the loads were taken off. This is clear- - the BS 5400 method
ly shown in Figs 4 and 5. - the EC4 method

3 70 TheStructuralEngineerVolume 75/No 21 4 November1997


Paper: Wang/Moore
~

TABLE 2 - Comparisons between predictions and test results,using Comparisons between predicted column strengths and test
measured material properties andpartial safety factors = l results
Sources of test results
A large number of concrete-filled steel circular (CHS) and square (SHS) hol-
low section, composite columns have been tested. Most of these test results
are included in a comprehensive report by Roik & Bergmann”. Columns
were subjected either to pure axial loading or single curvature bending.
More recently, Matsui et al” reported a series of similar tests. These tests
encompassed a variety of column sizes, column slenderness, steel and con-
crete material strengths, and levels of bending moment.
Tests on concrete-filled steel rectangular hollow section (RHS), com-
posite columns are rare. Shakir-Khalil & Zeghiche” and Shakir-Khalil &
Mouli” reported a series of tests on concrete-filled RHS columns with sin-
gle curvature bending moments. The authors’ eight tests are the only ones
RHS6 234 175 0.75 134 0.57 186 0.79 on slender concrete- filled composite columns under double curvature bend-
RHS7* 520 462 0.92 315 0.61 425 0.82 ing.
RHSP 480 419 0.87 281 0.58 420 0.88
Comparisonfor axially loaded composite columns, CHS and SHS sections
Ave. - - 0.77 - 0.58 - 0.76 Fig 6 presents the comparison between the predicted column strengths using
*Column effective length = 3.2m. the proposed method and the test results on axially loaded composite
columns. It shows that, for the range of column slenderness of practical
interest, the proposed method predicts column strengths which are in very
In the calculations, the average measured steel yield stress of 370N/mm2was good agreement with the test results. This suggests that the column buck-
used for all columns. The Young’s modulus of steel was assumed to be ling curve ‘a’can be used for composite columns. Comparisons between the
205kN/mm2. The measured concrete cube strength for all columns was predictions using the BS 5400 and the EC4 methods and the test results give
60N/mm2. This was used for calculating the column strength using the similar accuracy. Table3 gives a statistical summary for this comparison and
BS 5400 and the proposed method. The concrete initial modulus was shows a very good correlation between the proposed method and the two
27 OOON/mm2according to BS 5400.For calculations using EC4, concrete the existing design methods.
cube strength of 60N/mm2qualified as a C50/60 grade. Therefore, a charac-
teristic strength ofLk= 50N/mm2and a secant modulus of elasticity for short- Comparison for columns with single curvature bending,CHS and SHS
term loading of E,, = 37 OOON/mm2were used with the EC4 method. sections
The predicted column strengths from the three methods are given in The results of the comparison between the predicted column strengths using
Table 2 and they are compared with the test results. Table 2 shows that all the proposed method and the test results for CHS and SHS composite
three methods predict lower column strengths than the test results and are columns under single curvature bending are summarised in Fig 7. This fig-
therefore safe. For the eight columns, BS 5400 is most conservative. The ure ‘showsthat in most cases, the proposed method is conservative. Results
predicted column strengths from the proposed method are in reasonable in Table 3 suggest that, compared with the predictions using the proposed
agreement with the test results. As an indication of whether the accuracy of method, the predictions using BS 5400 and EC4 have a similar scatter as
the proposed method is acceptable, Table 2 shows that the proposed method indicated by the standard deviation but are slightly more accurate. On aver-
predicts results very similar to the EC4 method. age, the proposed method predicts column strengths that are about 8% and
1 1% lower than those predicted by using BS 5400 and EC4, respectively.
Further validationof the proposed approach An examination of the raw test data revealed that the proposed method
This validation study was carried out in two stages. First, the accuracy of predicted very low column strengths for very stocky columns and this
the proposed method is checked by comparing the predicted column reduced the overall accuracy of the method. The difference in predicted col-
strengths with available test results. Secondly, a systematic comparison umn strengths for very short columns is a direct result of the different N-M
between the predictions from the proposed method and those using interaction curves used in the three different methods.
BS 5400 and EC4 is carried out to check whether the proposed Fig 8 shows schematically the different interaction curves used in the
method is acceptable. three different methods. It is clear that, for a very short column with a high

1.2

-3
c

!?!
c
a.
c,
0
c

I
0 I I I I I I 1 t
0 20 40 60 80 100 120 140 160

Column slenderness
Fig 6. Comparison betweenproposed method and tests: axial load

The Structural Engineer Volume 75INo 21 4 November 1997 371


Paper: Wang/Moore

l I I l I I
0 20 40 60 80 100 120
Column slenderness
Fig 7. Comparison between proposedmethod and tests: single curvature bending

value of eccentricity, the difference in predicted axial strengths can be very For each section: six levels of eccentricity. The lowest level was for axial
large. loading, and the highest level of eccentricity resulted in
However, for columns of realistic length, the difference in predicted col- column failure in predominantly bending mode.
umn strengths using the three different methods becomes very small and the For eccentricity: three types of bending moment distribution of 13 = 1,0 and
predictions from the proposed method are very close to the test results. This -1.
trend is illustrated in Fig 7.
In all these calculations, the column height was 4m, and grade S355 steel
and C30/37 (EC4 term) concrete were used. The steel yield stress and
Comparison for concrete-filled, RHS composite columns
Young’s modulus were 355N/mm’ and 205kWmm’. The concrete cube
Fig 9 compares the predictions using the proposed method with the avail-
able test results. This figure indicates that the proposed method predicts col-
umn strengths in very good agreement with the test results. The results
given in Table 3 also suggest that the accuracy of the predictions using the 1
proposed method is very similar to those using EC4 and is better than those
using BS 5400.
0.E
Summary
Overall, the predicted column strengths from the proposed method are on
l
the safe side. These predictions are in reasonable agreement with the avail-
able test results, and the accuracy of these predictions is similar to the 0.E
BS 5400 and EC4 methods. In particular, column buckling curve ‘a’ can
be used for concrete-filled composite columns.

T
0.7
Systematic comparison between the proposed method and
BS 5400 and EC4
In this parametric study, the strengths of 216 columns were predicted
using all three methods. These columns consisted of the following arrange- 0.E
i I
ments:
12 steel sections: three SHS sections, three RHS sections and three RHS
sections bending about both the major and minor axes. 0.5
3
Each set of three sections represented the small, medium z
and large sections of each type. 1
0.4

TABLE 3 - Comparison between results~fromthe proposed method and


test results
0.3
Average Standard
Column type Design method
(prediction/test) deviation
I RS 5400 I 0.938 I 0.096 0.2
CHS & SHS,
EC4 0.96 1 0.121
axial load
BS 5950 I
0.946 0.098
BS 5400 0.835 0.136 0.1
CHS 8z SHS,
single curvature EC4 0.859 0.145
I
bending BS 5950 0.766 0. I60
.
0
~ ~~~ ~~~~ ~ ~~~

BS 5400 0.788 0.176 0 0.2 0.4 0.6 0.8 1 1.2


RHS,
EC4 0.867 0.1 12 MIMcx (MIMcy)
all tests
BS 5950 0.142 0.884 Fig 8.N-M interuction c t m v s

3 72 The Structural Engineer Volume 75/No 21 4 November 1997


Paper: Wang/Moore

1200- composite columns and is much easier to use than the methods given in
z“”1
h

Y / BS 5400 and EC4.


(3) The proposed method predicts column strengths in reasonable agreement
with test results and its accuracy is very similar to that of the BS 5400 and
EC4 methods.
(4) The correlation between the predictions using the proposed method and
those using the BS 5400 and EC4 methods is better than that between pre-
dictions using BS 5400 and EC4. Therefore, the accuracy of the proposed
method is acceptable.

References
1 . Bergmann, R., Matsui, C., Meinsma, C., and Dutta, D.: Design guide
for concrete-filled, hollow section columns under static and seismic
l I loading, CIDECT, Verlag TUV Rheinland, I995
0 200 400 600 800 1000 1200
2. ‘Design of composite bridges: use of BS 5400: Part 5: 1979 for
Test result (kN) Department of Transport Structures’, London, Department of Trans-
Fig 9. Comparison between proposed method
and tests: RHS port, December 1987
3. Basu, A. K., Somerville, W.: ‘Derivation of formulae for the design of
rectangular composite columns’, Proc. ICE, supplementary volume,
strength and Young’s modulus were 37N/mm2 and 24 800N/mm2 for cal-
paper 7206S, 1969, pp233-280
culations using BS 5400 and the proposed design method. For calculations
4. Virdi, S., Dowling, P. J.: ‘The ultimate strength of composite columns
using the EC4 method, the concrete characteristic strength and short-term
in biaxial bending’, Proc. ICE, 55, Part 2, March 1973, pp251-272
Young’s modulus were 30N/mm2and 32 000N/mm2.All material safety fac-
5. ‘Eurocode 4: Design of composite steel and concrete structures, Part
tors were set to be unity.
I . 1: General rules and rules for buildings’, Brussels, Commission of
European Communities, March I992
Results from the parametric study
6. Anderson, D.: ‘Calibration of composite columns: Eurocode 4 and
Only a summary of the results from this parametric study is given in Fig 10.
British practice’, Report to Building Research Establishment, Depart-
This figure shows the percentage distribution of three ratios: those of the pre-
ment of Environment, March 1992
dicted strengths using the BS 5400 and EC4 methods to that using the pro-
7 . BS 5950 Structural use ($steelwork in building: Part I : Code of pruc-
posed method, and the ratio of the predicted strength using EC4 to that using
tice for design in simple and continuous construction: hot rolled sec-
BS 5400. Fig 10 indicates that the discrepancy in the predicted column
tions, London, British Standards Institution, 1990
strengths between the proposed method and both the BS 5400 and EC4
8. Gibbons, C.: ‘The strength of biaxially loaded beam-columns in flex-
methods is less than 10%(0.9- 1.1) in most cases. In addition, Fig 10 shows
ibly connected steel frames’, PhD thesis, Department of Civil &
a better correlation between the proposed method with either of the two
Structural Engineering, University of Sheffield, 1990
existing methods than between the two existing methods. Therefore, the
9. Wang,Y. C., Moore, D. B.: ‘Steel frames in fire: analysis’, Engineering
accuracy of the proposed method is acceptable.
Structures, 17, No. 6, pp462-472, 1995
10. Roik, E. K., Bergmann, R.: ‘Report onEurocode4:Composite
Conclusions
columns’. Report EC4/6/89, University of Bochum, June 1989
This paper describes the development and validation of a simplified design
11. Matsui, C., Tsuda, K., and Ishibashi, Y.: ‘Slender concrete-filled steel
approach for concrete-filledcomposite columns. This method adapts the rec-
tubular columns under combined compression and bending’, Proc.
ommendations in BS 5950: Part 1’ for steel columns and is formed by
4th Pacific Structural Steel Conference, Vol. 3, Steel-Concrete Compo-
replacing the properties of the steel section with those of the composite sec-
site Structures, 1995, pp29-36
tion, which may be determined in accordance with the recommendations
12. Shakir-Khalil, H., Zeghiche, J.: ‘Experimental behaviour of concrete-
given inBS 5400: Part 5’. The findings of this paper may be summarised as
filled rolled rectangular hollow-section columns’, The Structural
follows:
Engineel; 67, No. 19, October 1989, pp346-353
( 1 ) The well-accepted design method in BS 5950 forbare steel columns can 13. Shakir-Khalil, H., Mouli, M.: ‘Further tests on concrete-filled rectan-
be adapted easily to concrete-filled composite columns. gular hollow-section columns’, The Structural Engineer, 68, No. 20,
(2) The proposed method gives a clear understanding on the behaviour of October 1990, pp405-413

0.35

<0.8 0.8-0.85
0.85-0.9
0.9-0.95
0.95-1.0 1.0-1.05 1.05-1.1 1.1-1.15 1.15-1.2 >1.2
Ratio of predicted strengths
Fig 10. Summary of parametric study results

TheStructural Engineer Volume 75/No 21 4 November 1997 373

View publication stats

You might also like