Professional Documents
Culture Documents
Abstract: Composite concrete filled tube 共CFT兲 columns are advantageous for use in seismic resistant braced frame construction since
these CFT columns are relatively inexpensive, stiff and strong. However, to utilize the composite behavior of the columns, the vertical
component of brace force must be transferred through the beam-brace-column connection and distributed to both the concrete and steel
in the column. Beam-brace-column connection details are critical for development of this composite action. This paper describes ana-
lytical and experimental studies carried out to better understand the transfer and distribution of force to the joint. A range of beam-brace-
column gusset plate connections were considered. It is shown that the majority of force transferred from the steel into the concrete occurs
by bearing rather than by friction. Steel gusset plates with horizontal ribs, or gusset plates with holes, allow more force transfer and have
more composite action than plain gusset plates. Slip deformations between the steel and concrete are likely to be too small to mobilize the
strength of shear studs in these connections.
DOI: 10.1061/共ASCE兲0733-9445共2004兲130:2共233兲
CE Database subject headings: Beam columns; Composite columns; Connections; Seismic properties.
Introduction penetrates into the CFT column and shear connectors on the gus-
set plate distribute the brace and beam forces between the steel
Concrete filled tube 共CFT兲 columns have been used for the con-
and concrete. Braces and beams may also be attached to the out-
struction of concentrically braced frames 共CBFs兲 up to 63 stories
side of the steel tube using steel-to-steel connections and shear
tall in regions of moderate to high seismicity in the U.S. CFT
connectors may distribute axial force to the concrete as shown in
columns are used because they offer significant advantages over
Fig. 1共b兲. Fig. 1共c兲 shows a connection similar to Fig. 1共a兲, but
either reinforced concrete or steel columns. Concrete placed in-
force transfer occurs solely through friction and bearing at the
side the CFT columns provides compressive strength and stiffness
bottom of the plate. Japanese CFT column connections use inter-
to the steel tube and restrains local buckling. The steel tube pro-
nal diaphragm connections as shown in Fig. 1共d兲. This connection
vides formwork to the concrete, minimizes the cost of the con-
requires four complete penetration welds around the perimeter of
crete placement, reinforces the concrete for axial tension, bending
the tube at each beam-column intersection as well as complete
and shear, and enhances the ductility of the column. To obtain
penetration welds in each beam flange.
these benefits, composite action between the steel and concrete in
While research on the connection of beams to CFT columns
the CFT columns must be developed and the vertical component
has been carried out for moment connections 共e.g., Schneider
of brace force must be distributed to the steel and concrete
1996; Azizinamini 1997; Ricles and Lu 1997; Yura and Jirsa
through the beam-brace-column 共BBC兲 connection. Appropriate
1997; Hajjar, Schiller and Molodan 1998a,b兲, and studies of bond
design of this BBC connection is therefore essential for good
at the steel-tube and concrete-core interface have been conducted
seismic performance.
by Virdi and Dowling 共1975兲; Shakir-Khalil 共1991兲 and Cameron
A wide range of BBC connection types have been used to
共1997兲, a study of factors controlling the ability of braces to trans-
connect braces and beams to composite columns 共Roeder et al.
fer force into the concrete and steel of CBF-CFT columns has not
2000兲. Fig. 1共a兲 shows one BBC connection where a gusset plate
yet been conducted.
1
Seismic design of CFT-CBF frames uses provisions such as
Associate Professor, Civil Engineering, Univ. of Washington, the International Building Code 共ICC 2000兲 or the 1997 Uniform
Seattle, WA 98195-2700 共corresponding author兲. E-mail: macrae@
Building Code 共ICBO 1997兲 combined with the AISC 共1997兲
u.washington.edu
2
Professor, Civil Engineering, Univ. of Washington, Seattle, WA Steel Seismic Provisions which references AISC-LRFD 共1994兲.
98195-2700. The CBFs are generally designed as special concentrically braced
3
Engineer, Toltz, King, Duvall, Anderson & Assoc., Inc., St. Paul, MN frames 共SCBFs兲 which are expected to develop inelastic action
55101-2140. primarily through tension yielding and buckling of the braces
4
Associate Researcher, Arch. Eng., Tokyo Inst. of Tech., 2-12-1 共AISC 1997兲.
Meguro-ku, Tokyo 152-8552, Japan. AISC-LRFD 共Section I 2.4兲 requires that transfer of force from
Note. Associate Editor: Sherif El-Tawil. Discussion open until July 1, the brace to the concrete be accomplished by direct bearing of
2004. Separate discussions must be submitted for individual papers. To steel on the concrete or through a mechanical shear device. If the
extend the closing date by one month, a written request must be filed with bearing area is A B , this transfer capability, b P b , is
the ASCE Managing Editor. The manuscript for this paper was submitted
for review and possible publication on September 11, 2002; approved on b P b ⫽1.7 b f ⬘c A B (1)
March 24, 2003. This paper is part of the Journal of Structural Engi-
neering, Vol. 130, No. 2, February 1, 2004. ©ASCE, ISSN 0733-9445/ where b ⫽0.60 and f ⬘c ⫽concrete cylinder strength. Section I5.3
2004/2-233–243/$18.00. provides an equation to assess the strength of stud shear connec-
Connection Modeling
or friction between the steel and concrete may also be significant. the right hand edge, because the brace loading occurs on that
As a result, the friction coefficient, , is increased to 0.4 to allow edge. There is no friction between the steel and the concrete for
greater bond transfer for this Case 2 model. Case 3 was consid- Case 3. The force that must be transferred from the steel plate by
ered to be the worst case model since the steel tube size was bearing to the concrete will cause an average stress of 352 MPa
decreased and friction was ignored. Cases 4 and 5 considered the 共51.1 MPa兲 at the bottom of the steel plate according to equilib-
possibility of column axial force variation between 0 and 50% of rium considerations. This average stress demand is larger than the
the crushing capacity. yield stress of 345 MPa 共50 ksi兲 but it is less than the ultimate
tensile stress. From the ABAQUS model the stress at the bottom
center of the steel plate from the contour plot is 161–185 MPa
Analytical Predictions 共23.4 –26.8 ksi兲. This is larger than that for Case 1 even though
the applied brace force in the Case 3 analysis is only 90% of that
The steel plate axial stress, given in Table 1, was found in the
used in the other analyses because of convergence problems. The
steel elements directly above the concrete at the center of the tube
at the maximum force as shown in the steel gusset plate axial maximum axial stress from the contour plot at the bottom of the
stress contour plot in Fig. 3共a兲. Here it may be seen that stresses steel plate is 321 MPa 共46.6 ksi兲. Since this stress is only 92% of
increase toward the bottom edge of the plate where the plate bears the required average bearing stress, the ABAQUS analysis does
on the concrete. The axial stress at the bottom and center of the not seem to model the bearing stresses well. In Case 5, the mid-
plate is between 113 and 132 MPa 共16.4 and 19.1 ksi兲 for Case 1 plate axial stresses were as high as 247 MPa 共35.8 ksi兲. The
when the peak compressive axial stress is 245 MPa 共35.5 ksi兲 at greatest plate principal stress of 350 MPa 共50.8 ksi兲 occurred in
Case 5. Peak compressive principal stress, like the peak axial slip of 2.5 mm 共0.05 in兲. Shear studs are therefore unlikely to be
stress, occurred at the bottom of the plate on the loaded side. The effective in connection details such as shown in Fig. 1共a兲.
peak tensile principal stress, shown in Fig. 3共b兲 for Case 1, was Concrete strains at three locations around the section 共Loca-
localized in the plate near the connection to the tension brace. tions A, B and C兲 for Case 1 are shown in Fig. 4共a兲. Strains given
Equilibrium dictates that the bearing stresses at the steel plate/ are at the centers of the concrete elements which were approxi-
concrete interface be equal in magnitude and opposite in direc- mately 12.7 mm 共1/2 in.兲 from the tube or plate. Significant con-
tion. The plate axial stresses are high with peak values approach- crete tensile strains occurred in the concrete above the bottom of
ing the steel yield stress. Concrete axial stresses in the elements the gusset plate. Strains were compressive below the plate and
directly beneath the plate at the center of the tube are about an they were greatest at Location C beside the gusset plate on the
order of magnitude less than the steel plate axial stresses as given side of the section where the force was applied. The strains be-
in Table 1. This is because the concrete stresses are computed at come uniform toward the bottom of the column. The length over
the mid-height of the 76.2-mm-deep 共3 in.兲 concrete elements which strain becomes uniform was defined as the distance from
rather than at the steel-concrete interface. If bearing force is trans- the bottom of the gusset plate to the location where the difference
ferred to the concrete at an angle of 2.5 units horizontal to one between the maximum and minimum strains at Locations A, B,
unit vertical, then the stress in the concrete at the element mid- and C becomes less than 5% of the average strains at this loca-
height would be expected to be at least an order of magnitude tion. It was approximately 1.8 times the column diameter for
smaller than the stress at the bearing interface. The stress distri-
Cases 1, 2, and 3 but it increased with increasing axial force as
bution of the concrete below the plate is given for Case 1 in Fig.
shown in Table 1.
3共c兲. The maximum concrete principal compressive stress demand
Force was only transferred to the concrete by two mechanisms
was often close to, and sometimes greater than, f ⬘c 共68.95 MPa or
in the model: 共i兲 bearing under the gusset plate or 共ii兲 friction
10 ksi兲 but it never exceeded the maximum possible stress of
between the steel plate and the concrete, or tube and the concrete.
1.16⫻68.95 MPa⫽79.98 MPa 共11.6 ksi兲 for the concrete model.
The percentage of vertical brace force carried by friction, as
From the results above, it may be seen that ability of the concrete
shown in Table 1, is responsible for as much as 30% of the total
to carry the high bearing stresses at the plate interface was not
represented well by the computer model and mesh size used. force transfer at the peak force in Case 4. Generally it is much
Since the concrete behavior beneath the plate is a critical part of less than 30% indicating that bearing of the gusset plate on the
the force transfer mechanism, and the actual concrete stresses concrete is the critical transfer mechanism for the force levels
beneath the plate are of critical concern, care is needed with in- considered. Fig. 4共b兲 indicates that the force carried by friction
terpretation of the results. increases at higher brace forces when concrete normal forces in-
Slip occurred between the steel and concrete because frictional crease at the steel-concrete interface. Friction between the steel
stress transfer was inadequate to assure strain compatibility. The tube and concrete contributes significantly to the total friction
relative steel-concrete vertical displacements for Location B force, even though at some locations the tube may move away
shown in Fig. 4共a兲 at the top of the column, at the top of the from the concrete due to the higher Poisson’s ratio of the steel
gusset plate, at the bottom of the gusset plate and at the bottom of tube than of the concrete.
the column are given in Table 1. In general, slip decreased toward In an actual column, boundary conditions may be different
the column ends, and the slip magnitude above the gusset plate from that assumed in the analyses described above. For example,
was slightly greater than that below the plate. The maximum slip at the base of the structure slip is restrained so the steel tube will
between the gusset plate and the neighboring concrete occurred carry a greater proportion of the total force at the first level than
near the loading point on the right hand side of Fig. 3共d兲. that assumed in the analysis above. The greater steel force will
The concrete-steel slips were generally less than 0.5 mm 共0.02 cause a greater separation of steel and concrete due to Poisson’s
in.兲. These slips are so small that they are unlikely to fully acti- ratio effects, so friction would be expected to be less than that
vate shear connectors. Taylor 共1985兲 has shown that many typical obtained from the models above. Also, concrete shrinkage 共Cam-
shear connectors do not reach their full capacity until there is a eron 1997兲 and degradation of bond from cyclic loading make it
unclear as to how much force transfer will occur due to friction in The test units were designed and built to simulate practice
an actual column. employed in recent prototype CFT-BBC connections with a scale
of about 2/3 and dimensions shown in Fig. 6. They were 1.193 m
共47 in.兲 long with an outside diameter, D, of 610 mm 共24 in.兲 and
Experimental Program wall thickness of 11.1 mm 共7/16 in.兲. These tubes had a nominal
strength of 248 MPa 共36 ksi兲 according to ASTM A252 G-2.
The analytical model used above could not simulate the BBC These properties satisfy the LRFD 共AISC 1994兲 requirements for
behavior well. In particular, the gusset plate-concrete interface composite members that:
behavior could not be appropriately modeled. Since this behavior • The steel tube cross-sectional area is greater than 4% of the
significantly affects the overall behavior of BBC connections, total composite area of the column;
total reliance on the analysis results to represent the behavior of • F y is less than 379 MPa 共55 ksi兲; and
actual CFT-BBC connections could not be justified. • The tube wall thickness is greater than or equal to D 冑( y /8),
An experimental study was therefore carried out to: where y is the tensile yield strain of the steel.
• Check the reasonableness of the analytical model by measur- ASTM A572-97 1/2 in. 共12.7 mm兲 Grade 50 gusset plates were
ing the slip and observing the posttest concrete deformation 660 mm 共26 in.兲 high and 762 mm 共30 in.兲 wide and the average
below the gusset plate; actual thickness was 13.2 mm 共0.52 in.兲. In several cases 3/4 in.
• Quantify the performance of a wide range of BBC connection 共19.3 mm兲 gusset plates with an actual thickness of 19.3 mm
types; 共0.76 in.兲 were employed. While the nominal yield stress was 345
• Evaluate the amount of composite action of each connection MPa 共50 ksi兲, the actual measured yield and ultimate strengths
type; were 343 and 510 MPa 共49.7 and 74 ksi兲 for the 12.7 mm 共1/2 in.兲
• Determine appropriate values for c 3 in Eq. 共2兲; and plate and 343 and 554 MPa 共49.7 and 80.3 ksi兲 for the 19.1 mm
• Evaluate the effect of bearing stress on the concrete. 共0.75 in.兲 plate, respectively.
Actual BBC connections with braces, beams and columns as The gusset plate fits within slots in the tube. The slots were cut
shown in Fig. 5 would be difficult and expensive to test with to a depth of 584 mm and width of 3.2 mm plus the gusset plate
realistic boundary conditions. Therefore, a number of less expen- thickness. The projection of the gusset plate above the tube was
sive simplified connections, which capture the important aspects approximately 76 mm 共3 in.兲. The plate was welded on the out-
of the force transfer, were tested. The test units represent the side of each side of the tube all the way around using 8 mm 共5/16
bottom half of the gusset plate and one half of the column clear in.兲 E70 SMAW welds.
length as shown in Fig. 5. The column clear length is assumed to These welds provide shear strength of 2.85 MN or 86% of the
equal twice the tube diameter based on prototype sizes. Force is plate axial yield strength. In an actual connection, the gusset plate
applied to the top of the gusset plate and distributed to the steel is twice as long, providing weld strength corresponding to 171%
tube through the welds and to the concrete by bearing and bond of the plate axial yield strength. The steel preparation was gener-
stresses. Uniform strain within the concrete at the base of the unit ally no more that brushing the loose debris from the inside sur-
was not expected for the length of column used according to the face, but one unit was sandblasted on the plate and inside of the
analytical study above. tube.
Before the concrete was poured, internal instrumentation was MPa 共6,581 psi兲 and 52.1 MPa 共7,563 psi兲 for Set 2 共Gunderson
attached, and the steel tube was placed on two layers of plywood 2002兲. The shrinkage of the mix design was 23⫻10⫺6 at 28 days
and held down with ropes. The bottom layer of plywood had a according to the concrete distributor. Moist burlap was used to aid
hole cut in it of the same diameter as the internal diameter as the concrete curing.
tube so that the top layer of plywood would deflect downwards A summary of the test unit properties is given in Table 2 and
making the concrete surface at the bottom of the tube convex as Fig. 8. The first test unit, Unit 1-1, was built according to the
shown in Fig. 7a. When the specimen was placed on the hydros- specifications described above. Unit 1-2 was tested with no con-
tone under the test machine, the convex surface reduced the pos- crete. Unit 1-3 had a thicker gusset plate that would be expected
sibility of trapped air pockets so that the full bearing area of the to cause more composite action as the bearing surface beneath the
concrete could be utilized. plate was larger than that for Unit 1-1. Unit 1-4 was greased to
A pea-gravel concrete was used for all test units. Concrete reduce the effect of bond and friction between the steel and con-
with a 28 day target strength of 48 MPa 共7 ksi兲 was placed crete, Unit 1-5 had two solid 508-mm-共20-in.兲-long 19-mm 共0.73-
through a 100-mm 共4-in.兲-diam polyvinyl chloride tube from a in.兲-square ribs of A572-97 Grade 50 steel placed horizontally on
bucket supported by a crane. Thorough vibration of the concrete each side and welded to the gusset plate with E70 6.35 mm 共1/4
was carried out and special care was made to ensure there were in.兲 fillet welds all around. These ribs were centered at approxi-
no gaps directly beneath the gusset plate. Two sets of test units mately the 1/3 points of the plate height 共203 mm 共8 in.兲 and 406
were tested. The 28 day strength of the concrete was 52.9 MPa mm 共16 in.兲 from the bottom of the plate, respectively兲, and this
共7,670 psi兲 and 47.54 MPa 共6,895 psi兲 for Sets 1 and 2, respec- rib spacing assured that the angle between the internal inside
tively. At the time of testing, the strengths varied between 53.3 bearing surface of one rib and the maximum outstand of the rib
MPa 共7,735 psi兲 and 72.6 MPa 共10,535 psi兲 for Set 1 and 45.4 below was less than 22.5°. A lower bound limit on the clearance
angle and rib spacing is essential if it is assumed that load is around the gusset plate in an X pattern. Unit 2-8 was similar to
spread out at an angle of 45° below each rib 共PCI 1985兲. Unit 1-6 Unit 2-7 except that 10 共nominally 32 mm diameter兲 Grade 60
had a large diameter hole in the center of the plate in the hope of 共nominally 414 MPa兲 deformed reinforcing bars were tack
increasing the bearing surface and bond resistance. A hole in the welded to the top of each of the five holes. The length of rein-
plate, rather than ribs on the plate, makes the connection easier to forcing bar in the four outside holes was 305 mm 共12 in.兲, and
fabricate. A second set of units, Set 2, was also tested in order to that in the center hole was 407 mm 共16 in.兲. Unit 2-9 was similar
verify the behavior of Set 1 tests and to investigate other configu- to Unit 2-1 but the gusset plate outstand at the top was removed
rations. Unit 2-1 was similar to Unit 1-1. The gusset plate and so that the concrete and steel could be loaded together.
inside of the steel tube were sandblasted to increase the friction Before each test a 19.1-mm 共0.75-in.兲-thick Grade 50 plate
for Unit 2-2. Unit 2-3 was similar to Unit 1-3. Unit 2-4 contained was set level on a 3.2 mm 共0.125 in.兲 layer of hydrostone under
11 Nelson 3/4⫻3-3/8 in. headed shear studs with a final length of the Baldwin 10 MN 共2300 kip兲 test machine. This plate had a hole
75 mm on each side of the gusset plate. The studs were welded to with a diameter equal to the inside diameter of the steel tubes.
the gusset plate using a Nelson TR 1800 welder with 480 V, 1600 The test units were positioned over the base with a fork lift. The
A, straight polarity and medium lift with a lifttime of 0.90 s. The hole in the steel plate was filled with wet hydrostone. The hydro-
yield strength of the studs was 458 MPa 共66.4 ksi兲 and the ulti- stone 1 h wet compressive strength was 28 MPa 共4.1 ksi兲. As the
mate strength was 489 MPa 共71 ksi兲. Unit 2-5 was similar with test units were lowered, concrete within the tube bore on the
only five shear studs on each side. Unit 2-6 was similar to Unit hydrostone pushing excess hydrostone out. As a result, the con-
1-5 except that three ribs were used on either side of the gusset crete was bearing directly on the hydrostone while the steel shell
plate. The ribs were centered 146 mm 共5.75 in.兲, 292 mm 共11.5 bore against the 19.1 mm 共0.75 in.兲 steel plate during testing as
in.兲, and 438 mm 共17.25 in.兲 from the bottom of the plate. Unit shown in Fig. 7共b兲. For the hollow tube, Unit 1-2, shims were
2-7 contained five 114-mm-共4.5-in.兲-diam holes distributed used to create uniform bearing between the tube and 19.1 mm
共0.75 in.兲 plate.
Instrumentation consisted of potentiometers measuring the
vertical displacement of the Baldwin loading head from the base,
the top of the steel tube, the top of the concrete surface and at the
bottom of the gusset plate. Strain gauges were placed vertically at
37 mm 共1.5 in.兲 below the top of the gusset plate, 15 mm 共0.6 in.兲
from the bottom of the gusset plate 共inside the tube兲, along the
outside of the tube beside the gusset plate, 25 mm 共1 in.兲 and 125
mm 共5 in.兲 below the gusset plate on the outside of the tube, on
the outside of the tube 50 mm 共2 in.兲 from the base and 215 mm
共8.5 in.兲 from the base. The number of gauges used at each loca-
tion varied slightly for the different units but typical details are
given in Fig. 6. Sometimes gauges were placed both inside and
outside the tube at the same location. A 61-mm-共2.4-in.兲-long
concrete embedment gauge measuring vertical strain was centered
215 mm 共8.5 in.兲 up the base in the center of the tube. It was
supported by a number of wires to prevent movement during
concrete placing. Strain readings of the concrete gauge tended to
Fig. 8. Idealization of configurations of units tested
be swamped by daily temperature fluctuations on the circuitry so
Practical Implications
Fig. 9. %Composite action From the analytical and experimental studies described above
some practical implications of the research are summarized
below.
amount of composite action at different levels of loading. It gen- • Friction between the tube/gusset plate and the concrete gener-
erally decreases as cohesion and friction is lost, but in some cases ally accounts for less than 30% if the total force transfer in the
it increases again possibly as bearing becomes more significant. analytical studies. Given the possibility of concrete shrinkage
The value of c 3 at Cycle 10 was computed according to Eq. 共Cameron 1997兲 and difficulties in estimating the amount of
共5兲 and is given in Table 2. It may be seen that c 3 is greater than friction in actual CFT columns, force transferred to the con-
0.6 for test units with ribs. This is significantly greater than the crete should be carried by bearing. This is consistent with
value of 0.4 suggested in current design procedures. For test units existing AISC code requirements.
with plain gusset plates c 3 is less than 0.4 at the high loads asso- • Since typical shear studs require about 2.5 mm slip before
ciated with Cycle 10. their strength is fully mobilized, and actual expected slips ex-
pected as a result of the analytical and experimental study
Slip Along Tube indicate that the slip may be less than 0.6 mm, shear studs are
not effective in transferring force between the steel and con-
The slip between the concrete and steel at the top of the tube crete. Their effect should therefore be ignored.
increased with increasing force and the relationship was approxi- • Since bearing stresses on the concrete from the gusset plate as
mately linear. When the force was removed during unloading, the large as those obtained for the prototype structure caused
slip did not return to zero. Slip tended to increase for units with cracks at 45° below the bottom of the plate, it is difficult to
less composite action as shown in Table 2. It may be seen that the justify the use of a bearing stress greater than the AISC code
average slip was about 0.7 mm, and it was 0.48 and 0.33 mm for requirement in Eq. 共1兲.
Units 1-5 and 2-6 with 4 and 6 ribs, respectively. Since the scale • The BBC connections causing the greatest amount of compos-
of the test units was 2/3, slips 1.5 times the values shown may be ite action in the column are those using horizontal ribs welded
expected in an actual structure. These levels of slip are of similar to the gusset plate. For connection types resembling those used
magnitude to that at the top of the unit analyzed as shown in in the experimental testing with welded ribs, the c 3 value in
Table 1 validating the global trends in the analyses. However, Eq. 共2兲 should be increased from the AISC recommended
they are significantly less than the level of 2.5 mm required to value of 0.4, to 0.6.
fully activate the strength of shear studs.
Conclusions
Stress at Bottom of Plate
The likely performance of a concentrically braced frame 共CBF兲
If there is no bond stress, then all force in the concrete occurs by with concrete filled tube 共CFT兲 columns designed according to
bearing and the average stress in the concrete beneath the steel current code procedures was investigated.