T-Cleat Connections to Concrete-Filled Tubular Internal

Columns
Abdel-Kareem Al-Rodan1
Downloaded from ascelibrary.org by BRUNEL UNIVERSITY LONDON on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Tests have been carried out on four full-scale composite connection specimens. The specimens were made by connecting two
unsymmetrically loaded steel beams to concrete-filled rectangular hollow section tubular columns. The columns were hence subjected to
axial loading and bending moments, and thus represented internal columns in multistory buildings. Columns and beam loads were
increased proportionately up to failure, and both column and connection behavior were observed and recorded. The experimental failure
loads and column behavior are compared against the predictions of a three-dimensional finite-element 共FE兲 model, using a general purpose
FE software.
DOI: 10.1061/共ASCE兲0733-9445共2004兲130:5共725兲
CE Database subject headings: Connections; Columns; Finite elements; Steel beams; Computer software; Axial loads; Bending
moments.

Introduction sizes of each type of steel hollow section were used. The connec-
tion was effected through welding either finplates or T-cleats at
Resent research 共Davison 1987; Wang and Nethercot 1988兲 has midheight of the columns. The beams were then bolted to the
shown that the force distribution in multistory steel frames is finplates/T-cleats stems by using 20 mm diameter, grade 8.8 bolts.
strongly influenced by the selection of the beam-to-column con- As the beam flanges are not connected, such connections are con-
nections. There have been many investigations of bare steel sidered in practice as simple connections. They are easy to manu-
frames incorporating different connections, but only limited re- facture, cost-effective, and have a moment-carrying potential.
search has been carried out on composite frames. The current The work reported here was also sponsored by BS-TP and was
British practice for the design of composite frames assumes that carried out on eight unsymmetrically loaded concrete-filled tubu-
beam ends are pinned. However, tests 共Owens and Echeta 1981; lar columns, series F. The test results of the first four specimens,
Van Dalen and Godoy 1982; Ammermann and Leon 1987兲 indi- group I of series F, have already been published 共Shakir-Khalil
cate that all beam-to-column connections possess some degree of and Al-Rodan 1995兲. Only one side beam was connected to the
rotational stiffness and moment capacity and that the utilization of tubular column of each of those four specimens. The columns and
the true connection performance in design can lead to benefits at beam-column connections of those specimens thus represented
both the servicability and ultimate limit states. edge composite columns in multistory structures. The results of
The test results of an experimental investigation involving 28 the last four tests, group II of series F, are reported here. As with
full-scale connection specimens have already been published all previous tests, each column of the current series was about 3 m
共Shakir-Khalil 1993a,b; 1994a,b兲. The test specimens in the ex- long and the beams were connected to the column at midheight.
perimental study were both symmetrical and symmetrically
loaded. The work was supported by both the Science and Engi-
neering Research Council and British Steel, Tubes, and Pipes Test Rig and Specimens
共BS-TP兲, who provided the writer with all fabricated specimens
and also a purpose built rig. Table 1 gives a summary of the six The test rig has already been described elsewhere, and a sche-
series of tests A to F in that test program. matic view is shown in Fig. 1. The rig has headroom of about
In the experimental investigation, series A–F consists of 28
specimens. The specimens were symmetrical and also symmetri-
cally loaded. The columns were 3 m long, and the beams were
connected to the columns at midheight. The concrete-filled tubu-
lar section of the column was either circular or square, and two

1
Associate Professor, Dept. of Civil Engineering, Mutah Univ. Karak,
P. O. Box 7, Jordan. E-mail: akrodan@mutah.edu.jo
Note. Associate Editor: Mark D. Bowman. Discussion open until Oc-
tober 1, 2004. Separate discussions must be submitted for individual pa-
pers. To extend the closing date by one month, a written request must be
filed with the ASCE Managing Editor. The manuscript for this paper was
submitted for review and possible publication on October 7, 2002; ap-
proved on February 4, 2003. This paper is part of the Journal of Struc-
tural Engineering, Vol. 130, No. 5, May 1, 2004. ©ASCE, ISSN 0733-
Fig. 1. Test rig and specimen
9445/2004/5-725–731/$18.00.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2004 / 725

J. Struct. Eng. 2004.130:725-731.

 Table 1. Summary of Tested Series
Series Column size Type of connection Number of specimens
A 168.3⫻5 circular hollow section Finplate 8
B 150⫻150⫻5 rectangular hollow section Finplate 8
C 219.1⫻6.3 circular hollow section Finplate 2
D 200⫻200⫻6.3 rectangular hollow section Finplate 2
E 150⫻150⫻5 rectangular hollow section T-cleat 8
F 150⫻150⫻5 rectangular hollow section T-cleat 8

3 m, is self-contained, and is bolted to the strong floor of the ment transducers were used to measure the in-plane and out-of-
Downloaded from ascelibrary.org by BRUNEL UNIVERSITY LONDON on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

laboratory. The rig was originally designed for testing symmetri- plane column displacements, and were also placed against the
cally loaded specimens, and was thus provided with no lateral column end plates to record the shortening and end rotations of
bracing. Triangulated bracing was therefore welded between the the columns 共Fig. 4兲.
vertical ties in order to ensure the lateral stability of the test rig The beam-column rotations were measured, and Fig. 5 shows
before testing the unsymmetrically loaded beam-column connec- how the rotations were obtained by placing displacement trans-
tions of test series F. ducers with magnetic bases on the side beams. Such measure-
The concrete-filled tubular section of the column was ments are relatively crude as they were affected by the deforma-
150⫻150⫻5 rectangular hollow section 共RHS兲. The beam- tions in both the T-cleat and the beam as well as by the slip in the
column connections of these specimens were effected through connection as a result of the clearance in the bolt holes. The
welding the flanges of T-cleats to the columns. The side beams results so obtained were therefore not exact measurements of the
were then bolted to the T-cleat stems by using 20 mm diameter, beam-column rotation, but gave nevertheless a good indication of
grade 8.8 bolts. Such connections are easy to manufacture, cost- the relative stiffness of the tested connections.
effective, and are considered in practice to be simple connections. Each connection was assembled by bolting the column in the
Two side beams were connected to each concrete-filled tubular upright position to the strong floor of the laboratory. The side
column. The beams were equally loaded, but the beam loads were beams were then loosely connected to the T-cleat, and their ends
applied at different distances from the column center. were then lifted upward in order to take up the slack caused by
The beam-column connections were fabricated by using the clearance in the bolt holes. Only then were the bolts hand-
T-cleats, as shown in Fig. 2. Compared to the finplate connection tightened. This procedure proved successful as it either greatly
used in previous tests 共one to three兲 and shown in Fig. 3, the reduced or eliminated the slip which was caused by the clearance
T-cleat connection is stiffer, less labor-intensive, and simpler to in the bolt holes.
manufacture. The T-cleats are cut from 305⫻127⫻48 UB 共Uni-
versal Beams兲 section and are welded to the column section using
an 18 mm fillet weld along the toe of the T-cleat flange. Each Material Properties
T-cleat is 360 mm long and is connected to the beam web by
either a single or double row of four grade 8.8, M20 bolts in 22 The details and properties of the four specimens, F5–F8, tested in
mm clearance holes. the current experimental investigation are shown in Table 2. For
The tubular column of all specimens is 2.8 m long and is made the sake of completion, the table also includes the first four speci-
of 150⫻150⫻5 RHS with 15 mm thick welded end plates. One mens of the same series, F1–F4, the results of which have already
end plate is provided with a 120 mm square hole to allow for the been published 共Shakir-Khalil and Al-Rodan 1995兲. When casting
casting of the concrete core of the tubular column. Vibrators were the concrete of each tubular column, at least four 100 mm cubes
used during the concrete casting process, and the specimens were and two solid 150 mm diameter cylinders were also prepared. The
bolted to a sturdy column which was bolted to the strong floor of concrete cube strength f cu and cylinder strength f cy are the respec-
the laboratory. When being tested, the column end plates were tive average results obtained from testing the cube and cylinders.
bolted to 40 mm thick loading plates which in turn rested on The steel tubes from which the column lengths were cut were
rollers to allow the specimens freedom of rotation in the plane of about 7 m long. This allowed the manufacturing of two speci-
loading. During the test procedure, the specimens were laterally mens, and left an adequate length for investigating the material
restrained at midheight, as shown in Fig. 1. properties of steel. Tension coupon specimens were prepared and
The test specimens were provided with electric resistance tested from all four sides of the RHS. The average of the four
strain gauges over distances equal to three times and twice the coupon tests gives the properties of steel in tension, namely de-
lateral dimension of the steel hollow section above and below the sign strength f sd , ultimate strength f su , yield strain ␧ y , and elas-
T-cleat position, respectively. Both longitudinal and transverse tic modulus E s .
strains were measured in the vicinity of the connection. Displace-

Fig. 3. Circular hollow secion and rectangular hollow section with

Fig. 2. Rectangular hollow section with T-cleat finplate

726 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2004

J. Struct. Eng. 2004.130:725-731.

 Fig. 6. Rectangular stress block
Downloaded from ascelibrary.org by BRUNEL UNIVERSITY LONDON on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

Properties of Composite Section

The three main properties of a composite section are the squash

load N u , the ultimate moment of resistance M u , and the interac-
tion curve of the section. The latter gives the relationship between
the axial load N and the uniaxial moment M, which when acting
simultaneously cause the section to reach full plasticity.
Short lengths of steel tubing were also cut from which stub
specimens 175 and 200 mm high were prepared. These were
tested respectively and gave the design strength of steel in com-
Fig. 4. Displacement transducer positions along column specimen * , and the experimental squash load of the composite
pression, f sd
section, N ue . The latter property is compared later to the pre-
dicted squash load given by
N u ⫽ f sdA s ⫹0.67f cuA c (1)
in accordance with British Standard 共BSI 1979兲. In order to com-
pare the experimental results with the BS5400 predictions, the
material partial safety factors of concrete and steel are taken equal
to unity in the expression. The same was also later applied when
calculating the ultimate moment capacity values, M u . The com-
parisons given later between prediction and experimental findings
are based on the tensile design strength f sd , whereas the compres-
* is tabulated only for the sake of compari-
sive design strength f sd
son. The predicted squash load N u and the ultimate moment car-
rying capacity M u for the column tubular section of each
specimen were calculated and are given later. These section prop-
erties are the basis of composite column design by BS5400.
Two values were calculated for the ultimate moment capacity
of each column section, namely M u1 and M u2 . These are, respec-
tively, based on the rectangular stress block 共RSB兲 approach
shown in Fig. 6 and the three-dimensional 共3D兲 finite-element
program predictions 共Hibbitt et al. 2002兲 as discussed elsewhere
Fig. 5. Beam-column rotation measurement 共Shakir-Khalil and Al-Rodan 1996兲. In accordance with BS5400,
the compressive strength of concrete in bending was taken to be
0.6f cu in both methods, and the tensile strength of concrete was
ignored in the rectangular strength stress block approach. How-

Table 2. Details of Specimens and Material Properties

e1 e2 f cu f cy f sd *
f sd f su Ec Es
Group no. Connection 共mm兲 P1 / P2 共N/mm2兲 ␧y 共kN/mm2兲
I F1 285 5 36.7 31.4 330 359 475 1,603 35.6 209
F2 285 8 45.5 34.3 330 359 475 1,603 36.7 209
F3 380 5 43.2 31.4 332 352 478 1,590 37.2 206
F4 380 8 45.1 31.3 332 352 478 1,590 35.7 206

II F5 400 200 5 41.8 32.1 340 349 379 1,640 34.1 204
F6 400 200 3 38.6 30.9 340 349 379 1,640 34.8 204
F7 600 270 5 41.7 33.4 343 351 379 1,620 33.3 203
F8 600 270 8 41.3 29.6 343 351 379 1,620 34.5 203

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2004 / 727

J. Struct. Eng. 2004.130:725-731.

Downloaded from ascelibrary.org by BRUNEL UNIVERSITY LONDON on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 7. Dimensional interaction curves

ever, as the finite-element model requires a value for the tensile

strength of concrete, this was taken to be 10% of the compressive
strength in bending, i.e., 0.06f cu . This explains the difference in
the moment capacity values obtained by the two methods and Fig. 9. Specimen finite-element model
given later.
In predicting the failure loads of the test specimens, a check
was required for the local capacity of the composite section. This The commercial multipurpose finite-element software package
was established by determining the interaction graph of the ABAQUS 共Hibbitt et al. 2002兲 is employed in this research. In the
concrete-filled section. The graph gives the combination of the finite-element models described herein, both geometrical and ma-
axial force N and bending moment M, which when acting simul- terial nonlinearities are considered. These nonlinearities tend to
taneously cause the section to reach full plasticity. The procedure create formidable computing requirements since dense meshes of
of plotting the interaction graphs is given elsewhere 共Shakir- brick elements must be used to properly model localized instabili-
Khalil and Al-Rodan 1996兲, and the resulting curves for the ties that occur in conjunction with global instabilities, both of
150⫻150⫻5 RHS are shown in Figs. 7 and 8 in dimensional and which are inelastic in nature.
nondimensional forms, respectively. The f cu and f sd values on ABAQUS is widely recognized as being a very appropriate tool
which the curves are based were, respectively, of the values of the for use in structural analysis problems involving a high degree of
tested columns, and were taken as 40 and 330 N/mm2, respec- geometric and material nonlinearity. The preprocessing and post-
tively. The results of both the rectangular stress block and finite- processing work was done by ABAQUS/CAE version 6.2, which
element methods are shown in the figures, with the former re- is an additional module written to enhance ABAQUS. It is a
ferred to as the RSB method. graphical user interface program that allows the user to execute a
finite-element 共FE兲 analysis process from start to finish. The FE
model can be viewed and checked interactively and the results
Numerical Analysis 共stress, displacements, etc.兲 can be visualized graphically.
It can be seen from the numerical model shown in Fig. 9 that
The finite element is a useful tool for analyzing problems with the side beams were not modeled, and neither were the bolts in
complex geometry, material properties, and boundary conditions. the connection. The T-cleat stem was extended to model the
The technique can produce comprehensive and reliable results if beam, and the numerical model was only used to predict the over-
the method is used correctly. It is much cheaper than the full-scale all column failure. The numerical model has 48 concrete brick
experimental testing. An accurate model to simulate the actual elements 共C3D20兲, 96 steel shell elements 共S8R兲 for the RHS
structure is necessary to accomplish a successful analysis. steel tube, 16 steel shell elements 共S8R兲 for the T-cleat flanges, 32
steel shell elements 共S8R兲 for the T-cleat stems, and 8 rigid brick
elements 共C3D20兲 for the end loading plate. In all, the model has
200 elements and 1,081 nodes. The boundary conditions at the
top of the numerical model are shown in Fig. 10. It can be seen
from the figure that, similar to the experimental end conditions,
the top end of the column was fixed in position and was allowed

Fig. 8. Nondimensional interaction curves Fig. 10. Boundary conditions at column top

728 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2004

J. Struct. Eng. 2004.130:725-731.

Downloaded from ascelibrary.org by BRUNEL UNIVERSITY LONDON on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. In-plane displacements group II 共F5兲 Fig. 13. Beam-column rotations group II

to rotate freely only about its Y axis. The only difference in the
connections F5 and F7, respectively. The initial stiffness of the
boundary conditions at the lower end of the model, and also at its
tested connections would have been influenced by the applied
midheight, is that all nodes along the Y axis were also allowed to
hand-tightening of the bolts, and this effect could have caused the
move in the Z direction, thus allowing for column shortening. The
apparent wide variation in connection stiffness. Specimens F7 and
numerical model was not used to predict the connection failure,
F8 seem to have suffered a sudden drop in their stiffness as the
which could be easily calculated in accordance with BS5950 共BSI
moment exceeded 20 kN m. This could have been as a result of
1985兲.
the applied moment overcoming the friction caused by the tight-
ening of the bolts. The stiffness of connection F5, on the other
hand, seems to have increased as the moment exceeded 10 kN m.
Test Results
This could have been as a result of the connection bolts being
lightly tightened, and that an applied moment of 10 kN m was all
The experimental in-plane displacements of specimen F5, which
that was needed to bring the bolts to bear against both the beam
is typical of the behavior of the other specimens, is shown in Fig.
web and T-cleat stem, thus taking up the slack in the clearance
11. It can be seen from the figure that at load levels up to about
holes.
1,200 kN, the stiffness of the lateral bracing was adequate to
It should be noted that the maximum beam-column moments
restrain the column at midheight. The upper and lower parts of the
reached in these tests do not represent connection failure. They
column are seen to have deflected in opposite directions. Fig. 11
represent the general failure of the column close to the top of the
shows that the column became unstable as its upper part ap-
T-cleat. This is the zone where the transducers were located and
proached its failure load and required more support from the brac-
where the bulges formed in the steel wall, thus rendering the
ing. This also caused displacements to develop at column mid-
readings of the transducers rather meaningless in the latter stages
height as a result of the flexibility of the lateral support, as seen in
of the tests.
Fig. 11. This also caused the lower part of the column to deflect in
The initial stiffness of the tested connections represents a wide
the same direction as its more severely loaded upper part. As the
spread of connection stiffness. A consistent method of assembling
upper part of the column approached the failure load, the bracing
connections and a better method of measuring the beam-column
had to sustain a dramatic increase in the axial force, as shown in
rotation should be adopted when studying the moment-rotation
Fig. 12.
relationship, M -␾, the general behavior of beam-column connec-
The beam-column rotation was measured as previously stated,
tions, and also establishing their stiffness.
and the results are shown in Fig. 13. The initial stiffness of the
The results of the connection tests are summarized in Table 3.
connections is seen to vary between 3,243 and 8,276 kN m/rad for
The table gives three predicted values for the failure loads of each
specimen, namely N p1 , N p2 , and N c . These are based, respec-
tively, on overall failure of the upper part of the column, failure
by the section reaching full plasticity just above the T-cleat posi-
tion in accordance with the interaction curves, Figs. 7 and 8, and
failure of the beam-column connection. The predicted failure load
N p of each specimen is the lowest of these three predicted values
and shown underlined in italics in the table. It should be stated
that in predicting the failure load of the upper part of the column,
N p1 , BS5400 makes no allowance for the restraining effect which
the lower part of the column has on the behavior of the upper
part. In other words, the upper part was considered to be pinned at
both ends and subjected to an axial load of P⫽ P 1 ⫹2 P 2 , and the
bending moment which varies linearly from P 2 (e 1 ⫺e 2 )/2 at the
bottom end, i.e., at midheight of the test specimen, to zero at the
top.
In calculating the connection failure load N c , all possible fail-
ure modes were investigated, including shear failure of bolts,
Fig. 12. Force in bracing group II 共F5兲
bearing failure, and yield line failure of the T-cleat. In investigat-

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2004 / 729

J. Struct. Eng. 2004.130:725-731.

 Table 3. Test Results
Nu N uc M u1 M u2 Ne Nn N p1 N p2 Nc Np
Group no. Connection no. 共kN兲 共kN m兲 共kN兲
II F5 1,532 1,771 59.3 62.0 1,436 1,378 1,202 1,277 972 972
F6 1,490 1,746 59.0 61.5 1,352 1,310 1,086 1,212 694 694
F7 1,539 1,737 59.8 62.6 1,298 1,258 1,071 1,163 605 605
F8 1,534 1,740 59.8 62.6 1,423 1,346 1,168 1,249 864 864

ing the yield line failure, the two modes shown in Fig. 14 were which tension in concrete is ignored. It can also be seen that the
considered in the analysis. In a yield line failure, one would nor- values of N p2 are between 6% and 12% higher than N p1 , and thus
Downloaded from ascelibrary.org by BRUNEL UNIVERSITY LONDON on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

mally assume the rotation to take place about the top end of the indicated that overall failure of the columns was expected to take
T-cleat stem with yield lines forming only in the cleat flange. place before the local capacity of the composite section was
However, it was noticed that all the test specimens showed that reached.
the rotation had taken place at the level of the second bolt from All specimens were expected to fail as a result of connection
the cleat top. This was later confirmed by the presence of a pres- failure as the N c values are seen to be considerably smaller that
sure point on the concrete core after the side wall of the RHS had N p1 . However, all specimens eventually failed as a result of over-
been removed. The yield line mode involved the formation of all column failure, and thus provide evidence of the conservative
yield lines in both the flange of the T-cleat and a plastic hinge in connection design in accordance with BS 5950. It is noteworthy
the T-cleat stem. A 3D view of the connection model is shown in that Table 4 shows that the experimental failure loads are between
Fig. 15. It can be seen from the figure that a gap has developed 48% and 113% higher than the connection failure loads of the test
between the cleat and the column as a result of the deformations specimens.
in the T-cleat flange. However, it should be stated that after a The experimental failure loads N c of the specimens are seen to
certain amount of rotation had taken place along the yield lines, be 19–24% higher than the predicted failure load of the upper
and a gap had consequently developed between the T-cleat flange part of the column, N p1 , and are also 11–14% higher than the
and the column, such deformations would result in the yield line failure loads based on the local capacity prediction, N p2 . How-
failure mode being replaced by a much stiffer membrane mode of ever, the experimental failure loads are seen to be in better agree-
behavior. This was observed in both the experimental and numeri- ment with the numerical predictions as the N c values are only
cal behavior of the specimens, and was caused by the fact that between 2% and 6% higher than the numerically predicted N n
excessive yield line deformations were arrested as the toes of the values.
T-cleat flange were fixed in position through being welded to the
column corners.
The numerically predicted failure load of the test specimens, Conclusion
N n , is based on analysis of the composite column model given in
Fig. 9, and thus allows for the column continuity at midheight. A The tests reported here were carried out to investigate the behav-
comparison of the results is shown in Table 4. The table shows ior of unsymmetrically loaded composite columns, and also of
that the experimental squash load, N ue , is between 13% and 17% beam-column connections in composite construction. The test re-
higher than the predicted value, N u , as based on BS 5400. It also sults show the moment carrying potential of T-cleat connections
shows that the finite-element model, which allows for tension in to concrete-filled tubular columns. When used for connecting
concrete to be taken into account, yields the ultimate moment of beams to columns of concrete-filled RHS, the yield line mode of
resistance values, M u2 , which are between 4% and 5% higher failure should not be overlooked. However, it should be stated
than those based on the rectangular stress block approach in that after a certain amount of rotation had taken place along the
yield lines, and a gap had consequently developed between the
T-cleat flange and the column, such deformations would result in
the yield line failure mode being replaced by a much stiffer mem-

Fig. 14. Yield line failure modes Fig. 15. Cleat flange deformation

730 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2004

J. Struct. Eng. 2004.130:725-731.

 Table 4. Comparison of Test Results
Group Connection N ue /N u M u2 /M u1 N p2 /N p1 N e /N p N e /N p1 N e /N p2 N e /N n
II F5 1.16 1.04 1.06 1.48 1.19 1.12 1.04
F6 1.17 1.04 1.12 1.95 1.24 1.12 1.03
F7 1.13 1.05 1.09 2.13 1.20 1.11 1.02
F8 1.13 1.05 1.07 1.65 1.22 1.14 1.06

brane mode of behavior. Although the yield line type of failure N u ,N ue ⫽ predicted and experimental squash loads of
was observed to fully develop in the test specimens, it did not stub column;
Downloaded from ascelibrary.org by BRUNEL UNIVERSITY LONDON on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

result in the eventual failure of the connection. P ⫽ total load on upper part of column ( P⫽ P 1
The test results confirmed yet again the conservative design ⫹2 P 2 );
procedure of British Standards in predicting the failure loads of P 1 , P 2 ⫽ loads applied by middle and side jacks,
beam-column connections. They also underline the importance of respectively;
providing lateral bracings with sufficient strength and stiffness to ␥ mc ,␥ ms ⫽ material partial safety factors of concrete
satisfactorily perform their function. The experimental failure
and steel, given, respectively, by 1.5 and
loads of the tested concrete-filled RHS columns were found to be
1.1 in accordance with BS 5400; and
about 20% higher than the values predicted by British Standards.
␧ y ⫽ experimental yield strain of steel in
However, the experimental results are in better agreement with
microstrain.
the finite-element analysis which allows for column continuity.
The finite-element model provided an insight into the behavior of
the tested specimens, and should prove useful in investigating References
columns under more complicated loading conditions.
Ammermann, D., and Leon, R. T. 共1987兲. ‘‘Behavior of semi-rigid com-
posite connections.’’ J. Struct. Eng., 113共8兲, 1165–1174.
Acknowledgments British Standards Institution 共BSI兲. 共1979兲. ‘‘Steel, concrete and compos-
ite Bridges.’’ Code of practice for the design of composite bridges, BS
The work reported here was sponsored by British Steel, Tubes 5400, Part 5, London.
and Pipes, Corby, England, UK. The writer wishes to acknowl- British Standards Institution 共BSI兲. 共1985兲. ‘‘Structural use of steel work
edge the assistance received from, and the technical discussions in building.’’ Code of practice for design in simple and continuous
carried out with, Dr. Shakir-Khalil and the British Steel personnel. construction: Hot rolled sections, BS 5950, Part 1, London.
Davison, J. B. 共1987兲. ‘‘Strength of beam-columns in flexibly connected
steel frames.’’ PhD thesis, Dept. of Civil and Structural Engineering,
Univ. of Sheffield, Sheffield, U.K.
Notation
Hibbitt, Karlson, and Sorenson. 共2002兲 ABAQUS user manual, Version
6.2, Part 1 & 2, Hibbitt, Karlson and Sorenson, Inc.
The following symbols are used in this paper:
Owens, G. W., and Echeta, C. B. 共1981兲. ‘‘A semi-rigid design method for
E c ,E s ⫽ elastic moduli of concrete and steel; composite frames.’’ Proc., Int. Conf. on Joints in Structural Steelwork,
e 1 ,e 2 ⫽ beam load eccentricities from face and London, 233–241.
center of column e 1 ⬎e 2 ; Shakir-Khalil, H. 共1993a兲. ‘‘Full-scale tests on composite connections.’’
f cd , f sd ⫽ design strengths of concrete and steel, given Composite Construction in Steel and Concrete II, ASCE, New York,
by respective characteristic strength f ck and 539–554.
f sk , divided by material partial safety Shakir-Khalil, H. 共1993b兲.‘‘Connection of steel beams to concrete-filled
factors ␥ mc and ␥ ms ; tubes.’’ 5th Int. Symp. on Tubular Structures, Nottingham, England,
f ck , f sk ⫽ characteristic strengths of concrete and steel, U.K., 5-ISTS, 195–203.
the former taken as 0.67f cu in accordance Shakir-Khalil, H. 共1994a兲. ‘‘Finplate connections to concrete-filled
with BS 5400; tubes.’’ 4th Int. Conf. on Steel-Concrete Composite Structure, Kosice,
f cu , f cy ⫽ characteristic 28-day cube and cylinder Slovakia, ICCS-4, 181–185.
strengths of concrete; Shakir-Khalil, H. 共1994b兲. ‘‘Beam connection to concrete-filled tubes.’’
* ⫽ design strength of steel in compression;
f sd 6th Int. Symp. on Tubular Structures, Monash, Australia, 6-ISTS,
f su , f y ⫽ ultimate and normal yield strength of steel; 357–364.
Shakir-Khalil, H., and Al-Rodan, A. 共1995兲. ‘‘Behaviour of concrete-
M u1 ,M u2 ⫽ ultimate moment capacity of column section
filled edge columns.’’ 3rd Int. Conf. on Steel and Aluminum Struc-
by rectangular stress block and finite-element
tures, ICSAS ’95, Istanbul, Turkey, 515–522.
analysis, respectively;
Shakir-Khalil, H., and Al-Rodan A. 共1996兲. ‘‘Experimental behavior and
N c ⫽ predicted failure load on basis of connection numerical modeling of concrete-filled rectangular hollow section tu-
failure; bular columns.’’ Composite Construction III Conf., Irsee, Germany,
N e ⫽ experimental failure load; 267–278.
N n ⫽ numerically calculated column failure load; Van Dalen, K., and Godoy, H. 共1982兲. ‘‘Strength and rotational behavior
N p ⫽ predicted failure load, taken as smallest of of composite beam to column connections.’’ Can. J. Civ. Eng. 9, 313–
N c , N p1 , and N p2 ; 322.
N p1 ,N p2 ⫽ predicted failure loads on basis of column Wang, Y. C., and Nethercot, D. A. 共1988兲. ‘‘Ultimate strength analysis of
failure and local capacity failure of three-dimensional column subassemblages with flexible connections.’’
composite section; J. Constr. Steel Res., 9共4兲, 235–264.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MAY 2004 / 731

J. Struct. Eng. 2004.130:725-731.