Sunteți pe pagina 1din 12

Seismic Behavior of Steel Braced Frame Connections to Composite

Columns

Charles W. Roeder, Gregory MacRae and Christopher Waters


University of Washington

ABSTRACT

Concrete filled tubes (CFT) offer large axial stiffness and load capacity
and are suitable for columns in braced frames. However, brace-to-beam-to-
column connections are a concern. Past designs vary widely, and their seismic
behavior is uncertain. Past applications of CFT columns in braced frames are
summarized. The difficulties in connection design are discussed, and nonlinear
computer analyses to evaluate the seismic performance of these connections are
reviewed. An experimental investigation of the connection performance will be
performed.

INTRODUCTION

Concrete filled tubes (CFT) columns have been used for seismic resistant construction,
because they offer significant advantages over either steel or reinforced concrete. The concrete
provides compressive strength and stiffness to the steel tube and restrains local buckling. The
steel tube provides formwork to the concrete, minimizes the cost of the concrete placement,
reinforces the concrete for axial tension, bending and shear, and enhances the ductility of the
column. CFT columns are particularly well suited for braced frames, because large axial
strength and stiffness are needed. This combination is particularly desirable for seismic design,
because braced frames are very efficient for seismic design, and lateral loads place great
demands on the columns. Further, seismic design requires a combination of strength, stiffness
and ductility from the structural system.

While CFT columns are desirable for braced frames, there is little guidance regarding
the design and construction of connections between the braced frame and the CFT columns.
Variations of brace-to-beam-to-column connections (BBC connections) have been used, but
there has been no research regarding the seismic performance of these connections. As a
consequence, the seismic behaviors of the alternative BBC connections are not understood,
and the relative economy of the various alternatives is uncertain.

51
PAST APPLICATIONS

A number of buildings have been constructed with braced frames and CFT columns.
This usage appears to have initiated in the United States. CFT columns have been widely used
in Japan and other countries well before they were commonly used in the US, but the usage in
these other countries has primarily been with moment resisting frames. The first U.S. high-rise
project (Hooper et al. 1) with CFT columns was the 100 First Street Plaza in San Francisco,
California. This building is illustrated in Fig. 1. It is 27 stories tall, and it employs a braced
frame with moment connections and CFT columns at the four corners of the braced core. Since
that period, other braced frame buildings with CFT columns have been constructed. Building
heights have been up to 60 stories. A building with more than 100 stories has been designed
but not constructed. Other structures, such as the University of Washington football stadium,
also used variations of this structural system.

Figure 1. First Street Plaza Building Constructed with Braced Frames and CFT Columns

While different buildings have used braced steel frames with CFT columns, a wide range
of BBC connections have also been employed. Figure 2 schematically shows several of these
connection alternatives. The alternatives illustrated in Figs 2a and 2b employ large gusset
plates that penetrate into the CFT column. In Fig. 2a, the gusset plate has a number of shear
connectors to help distribute the brace and beam forces and moments between the steel and
concrete in the CFT column, while Fig. 2b has a plain gusset plate with no shear connectors.
Figure 3 illustrates another detail that has been used occasionally. Shear connectors are
attached to the inside of the tube to help distribute moments and forces between the steel and

52
the concrete, but the brace and the beam are attached to the steel tube with a typical steel-to-
steel connection. Braced frames with CFT columns have been frequently used in the US, but
seldom used in Japan. However, at least one recent braced frame building in Japan has
employed CFT columns, and the BBC connections for this building are schematically shown in
Fig. 2d. Japanese engineers commonly use CFT columns with moment resisting frames, and
they use an internal diaphragm connection. Figure 2d is a variation of the internal diaphragm
connection. The diaphragms penetrate into the tube and interlock with the concrete fill. This
interlock should help distribute brace and beam loads to the steel and concrete of the CFT
column. This later connection requires four complete joint penetration welds around the
perimeter of the tube at each beam-comlumn intersection.

Figure 2. Typical Brace-Beam- Column Connections for Braced Frames with CFT Columns

53
SEIMSMIC DEMANDS

Seismic design is performed so that the building remains elastic and completely
servicable during and immediately after small frequent earthquakes. Significant inelastic
deformation is permitted during large earthquakes that have a small probability of occurrence.
The inelastic deformation changes the stiffness of structure and dissipates energy, and these
combine to provide moderate building response during major earthquakes at significantly
reduced initial cost of the structure. The consequence of this inelastic deformation is permanent
yielding and deformation of the structure. Therefore, the engineer must assure that the
structure retains its basic integrity during these severe earthquakes without building collapse or
loss of life. This multi-level performance design procedure requires that the engineer consider
the strength, stiffness, ductility and inelastic performance of the system. The inelastic
performance of braced frames is dominated by yielding and buckling of the braces. Columns
assist in resisting the lateral loads, but they also support gravity loads and prevent collapse of
the structure. Connections transfer moments and forces from member to member. Premature
failure or fracture of the columns or connections results in poor performance which defeats the
goals of the seismic design approach. As a consequence, seismic design requirements (AISC
2) for braced frames require that the brace be the weak link. The connections and the columns
are designed to be strong enough to sustain the full compressive buckling load and the full
tensile yield load of the brace. Inelastic deformation of the braces is tolerable, since it

- retains the economy of the design,


- limits forces delivered to columns, connections and other elements, and
- benefits the inelastic response of the structure.

However, the inelastic buckling and deformation of the brace places additional demands on the
connection. All of these requirements must be satisfied with full consideration of the inelastic
structural deformation and the uncertainty in material properties and performance of the as-built
structure.

As a consequence of this design method, one key issue in the seismic design of braced
frames with the CFT columns is the determination of whether the BBC connections are able to
achieve these objectives. These connections must develop and retain the resistance necessary
to transfer brace and beam forces and moments, while the structure and the bracing members
sustain large inelastic deformations. The BBC connection must also distribute these forces and
moments between the steel tube and the concrete fill if the CFT column is to perform as
required. Multiple requirements must be achieved if this performance goal is to be met. First,
the steel tube serves as tensile reinforcing to the CFT column, and so the tensile force must be
distributed to the steel tube. Second, the concrete fill stiffens and strengthens the steel for
compressive load, and so the compressive force must be distributed between the concrete and
the steel. Third, the steel tube serves as sole flexural reinforcement to the concrete, and
therefore the steel tube must either carry the full bending capacity of the column or it must at
least resist all tensile stress due to bending moments. Finally, the steel acts as shear
reinforcement to the concrete, and so shear forces must be distributed between the steel and
concrete in some appropriate manner. This complex distribution of force is not magically
achieved. It requires that there either be reliable shear stress transfer between the steel tube
and the concrete fill as illustrated in Fig. 3, or as an alternative, the BBC connections must aid in
the force distribution and transfer.

54
Figure 3. Stress Transfer Between Steel and Concrete

POTENTIAL PROBLEMS IN DESIGN OF THESE CONNECTIONS

The ability to achieve the shear stress transfer required by equilibrium and illustrated in Fig.
3 is an issue of some concern. Past research evaluated this shear transfer (Roeder et al. 3, and
Roeder 4), and the shear transfer demands are found to be large and very localized as
illustrated in Fig. 4. This figure shows that the force or moment that is applied to either the steel
tube or the concrete fill must be appropriately shared with the other material in a very short
distance (usually much less than one tube diameter). Braced frames place large demands at
every BBC connection. If the shear stress transfer capacity between the steel tube and
concrete does not exceed the demand, slip occurs and causes permanent changes to the steel
concrete interface.

Figure 4. Distribution of Stress Transfer Demands

55
Figure 5. Proposed Bond Stress Evaluation Model for CFT

The shear stress transfer capacity was evaluated (Roeder et al. 3, and Roeder 4) for CFT
elements, and the important engineering criteria for safety, serviceability and economy are -

x shear connectors are not needed when the natural shear transfer capacity exceeds
the demand, but
x mechanical shear transfer must be employed for the full transfer requirement
whenever demand exceeds capacity.

At the ultimate load performance level, shear transfer demand can be uniformly distributed over
the perimeter of the tube and a length equal to the smaller of the column length or 3.5 times the
tube diameter as illustrated in Fig. 5. At the serviceability limit state, the triangular shear stress
demand distribution can be used over a length equal to one half the diameter of the tube. The
shear transfer capacity was shown to be a function of the tube diameter, d, and diameter to
thickness ratio, d/t. The shear stress capacity that is two standard deviations below the mean
capacity, ƒ2V, is

ƒ2V = 1.24 - 0.0702 (d/t) + 0.0029 d (in MPa and d is in mm) (1)

The transfer capacity is effectively zero for tubes with large diameter and large d/t ratio,
because of separation that may occur between the steel and concrete and the limited
confinement provided by thin tubes. Most CFT columns used in the US are large diameter
tubes with large d/t ratios, and so the natural shear stress transfer is clearly inadequate for
seismic design of most BBC connections.

INELASTIC COMPUTER ANALYSIS

The shear transfer demands needed to make the CFT column a functional part of the
braced frame as illustrated in Fig. 3 exceed the shear transfer capacity available within most
CFT columns. Therefore, the BBC connection must aid in this transfer, or mechanical transfer
devices must be employed within the critical area (see Figs. 4 and 5) of the CFT columns. A
series of non-linear analyses with the ABAQUS computer program were completed to evaluate
these options. Most BBC connections used for seismic design of buildings employed
penetrating gusset plate connections as illustrated in Figs. 2a or 2b. The penetration of this
gusset plate should help to distribute moments and forces between the steel tube and the
concrete fill, and so variations of these connections provide the basis of the nonlinear analysis.
The BBC connection employed in the U. of Washington stadium and illustrated in the photo of

56
Fig. 6 is the prototype for this analysis. This prototype was selected because it is of
intermediate scale to BBC connections used in existing structures. The connection is quite
large with an outside diameter of approximately 710 mm, but it is much smaller than
connections used on a number of the large buildings noted earlier.

Figure 6. Photograph of Prototype Connection

A detailed finite element grid illustrated in Fig. 7 modeled the connection. The elements
were 3 dimensional 8-node brick elements, and the steel was modeled as a bi-linear plastic
material with a yield stress of 345 MPa, a Poisson’s ratio of 0.3, an ultimate tensile stress of 380
MPa, and a strain hardening ratio of 4% of the elastic modulus. The tube thickness was 12.5
mm (for a d/t of approximately 57) in the basic model, but this thickness (and the resulting d/t)
was varied throughout the study. There was considerable difficulty in modeling the concrete.
Initially the ABAQUS unreinforced concrete element was employed, but this model consistently
failed to converge in regions where moderately large stress and strain developed. The ABAQUS
concrete model is not designed to handle situations in which the concrete is well confined, and
therefore the tension softening curve was artificially extended to provide some apparent ductility
(although at extremely low tensile stress) to the behavior. The tensile strength of the concrete
was limited to less than 8% of the compressive strength, but the compressive strength of the
concrete was a variable parameter in the study. The interface between the steel and concrete
surfaces were also parameters of interest. Because of the limited stress transfer capability
noted earlier, no adhesion or rigid attachment between the steel and concrete were permitted.
The interaction between steel and concrete elements was modeled using contact pairs
consisting of master (steel) and slave (concrete) surfaces. Hard contact was specified, requiring
that the clearance between the surfaces be zero before the surfaces may interact. When in
contact, the surfaces are permitted to transfer normal stress and shear stress through friction.
The coefficient of friction was initially chosen to be 0.3, but this coefficient of friction was also a
parameter considered in the study. Before any loads are placed on the model, the clearance
between all steel and concrete surfaces is zero.

57
Figure 7. Schematic of Analytical Model

The loads illustrated in Fig. 7 were monotically applied to the specimen. The gravity
loads in the CFT column were applied first, and the brace loads were later simultaneously
applied in increments to observe the nonlinear behavior of the connection. The maximum brace
loads were computed based upon the tensile yield and compressive buckling capacity of the
diagonal brace pair. The compressive load on the CFT column was a parameter of interest in
the analysis, and this load was varied between 10% and 30% of the compressive load capacity
of the column. The model utilized symmetry, and the model was split through the center of the
CFT column and gusset plate as shown in Fig. 7. Out-of-plane movements and in-plane
rotations were restrained on this plane of symmetry, but all other deflections and rotations were
permitted. The loads were applied so that the specimen was always in equilbrium, and so
supports of the assemblage were theoretically not required. However, accidental imbalance
occurs because of roundoff and other errors. As a result, a single node located at the centroid
of the column was pinned or restrained with a spring at each of the top and bottom cross-
sectional surfaces. These restraints prevented rigid body translation in the plane of symmetry
and rigid body rotation (vector directions defined by right hand rule) perpendicular to the plane
of symmetry. The plane of symmetry constrained the other components of rigid body translation
and rotation. A spring element with a stiffness of 350.4 kN/mm was applied at the bottom of the
column. The resulting model had 5100 elements and 8863 nodes. Computer analyses required
periods from as little as a few hours to more than a day for completion on Sun Sparc
Workstation.

58
PARAMETERS AFFECTING CONNECTION PERFORMANCE

Initial analyses evaluated convergence and mesh refinement of the model but are not
discussed here. Later analyses show that the connection relies primarily on the bearing capacity
of the concrete, as the primary load path from the braces to the concrete. This is illustrated in
the contours of compressive stress in the concrete below the gusset plate in Fig. 8. The
maximum bearing is delivered through the edge of the gusset plate into the concrete. The large
bearing stresses occur in the concrete immediately below the gusset plate at the corner
intersection of the plate and tube. In Fig. 8, the average compressive stress in the concrete is
slightly approximately 10.2 MPa, but the actual bearing stress is 58.9 MPa under the loaded
edge of the gusset plate and 29.2 MPa at the opposing corner. The concrete is well confined in
this area. The highly stressed area is in hydrostatic compression, and the material can probably
tolerate stresses larger stress than the uniaxial compressive capacity of the concrete, but the
limitations of the finite element model result in convergence problems at these highly stressed
locations. Stress concentrations are noted in the steel tube and the gusset plate at these
junctures, but the stresses in the steel are larger and the consequences of the high stress are
less severe in the analytical solution. Yielding is noted at several locations in the gusset plate
and the steel tube, but it is quite local. The juncture of the gusset plate and the steel tube, and
the attachment of the brace to the gusset plate are such locations.

Figure 8. Stress Contours in Concrete Below the Gusset Plate

While bearing of the gusset plate on the concrete is the critical transfer mechanism,
some load transfer occurs because of friction between the steel and concrete. There was no
attachment between the steel and concrete. As a result, friction provides the only stress
transfer between the steel tube and the concrete fill or between the side of the gusset plate and
concrete. This friction transfer was smaller than the bearing stress transfer noted above, but it
was not insignificant. In the models using a high coefficient of friction (0.4), the force
transferred by frictional shear transfers approximately 32% of the total vertical component of
brace forces. As the coefficient of friction increases and larger portions of the brace force are
distributed to the concrete through friction, the maximum bearing stress beneath the plate

59
decreases as shown in Fig. 9. The reduction is stress is not huge, but the demand is pulled
below the ultimate capacity of the concrete and reduces the local strain demand and potential
for crushing. Friction depends on the surface condition of the steel and concrete and shrinkage
effects, and these factors are not easily controlled. As a result, the extent to which this friction
transfer can be used is unclear.

Figure 9. Effect of Increased Friction on the Maximum Concrete Bearing Stress

There was significant slip between the steel and concrete. Figure 10 shows the pattern
of slip between the gusset plate and the concrete of a typical specimen. The slip occurs
because the strain in the steel and concrete are not compatible, and the shear stress transfer
does not distribute the stress between the steel and concrete rapidly enough to achieve strain
compatibility. The slip is largest near the mid height of the gusset plate near the loaded edge.
At this location, the steel gusset plate is highly stressed, while the concrete is only lightly
stressed, and the slip occurs in the region with relatively large strain differential. Slip is a
concern because it represents a potential source of deterioration within the CFT column and
connection and it decreases as the coefficient of friction increases. Slip was also noted between
the concrete and the steel tube, but this slip was more uniform for a given length.

Other parameters were evaluated. The effect of the strength of the concrete on the BBC
connection behavior was considered. Concrete with increased strength tolerates the large
bearing stress such as illustrated in Fig. 9 better, but it doesn’t reduce the slip and deformation
with the connection or otherwise improve the load transfer between steel and concrete.
Increased wall thickness of the steel tube significantly reduced the maximum bearing stress in
the concrete. A larger portion of the transfer appeared to occur between the concrete and the
steel tube, since the greater stiffness of the walls of the tube provided better confinement and
resulted in larger contact stresses and friction along that interface. Increased thickness of the
gusset plate directly lowered the maximum bearing stresses on the concrete.

The potential use of shear studs on the gusset plate as illustrated in Fig. 2a and on the
inside of the tube as illustrated in Fig. 2c was also investigated. Shear studs do not appear to
be an effective way of improving the shear transfer between the steel and concrete. Shear
studs are relatively flexible, and they develop their full shear resistance only after sustaining
(Taylor 5) a relative deformation. The slip occurring within these BBC connections provides a
small relative deformation, but it is not large enough to effectively develop shear studs. Other
mechanical transfer devices with greater stiffness may benefit the connection performance. The
capacity of the connection to transfer and distribute load between the steel and concrete is
dependent on the concrete strength and the thickness of the plate and tube. The stress

60
concentration at the junction between the tube and the bottom edge of the plate is a critical point
within the connection.

Figure 10. Slip Between the Gusset Plate and Concrete Fill

CONCLUSIONS

In order for CFT braced frames to behave well, transfer of force from the braces in the
column infill concrete is necessary. Due to the large diameter, d/t ratios and force demands
expected in typical US frames, friction between the concrete and steel was shown to be
unreliable for design. Also, the use of shear connections is unsatisfactory since excessively
large deformations are required to develop there shear transfer. Direct bearing of steel against
the concrete was shown to be a promising method for force transfer. Testing and further
analysis are presently underway and final conclusions should be available in approximately 18
months.

ACKNOWLEDGEMENTS

This work is funded by the Nat


ional Science Foundation under Grant CMS-9905797. “Concrete Filled Tube Braced Frame
Testing”. Dr. S. C. Liu and Dr. V. Gopu are the Program Managers for this study. Their support
is greatly appreciated. Mr. John Hooper and Prof. C. B. Brown are technical advisors on the
project. They are providing practical insight into the use of CFT columns in actual braced frame
construction and assisting with analysis and evaluation of the research results.

REFERENCES

1. Hooper, J.D., Roeder, C.W., Kelmencic, R., and Nordquist, K., Concrete-Filled Tubes for
High-Rise Construction, Civil Engineering, ASCE, Washington, D.C., February 1999.
2. AISC, Seismic Provisions for Structural Steel Buildings, AISC, Chicago, IL 1997.
3. Roeder, C.W., Cameron, B. and Brown, C.B., (1998) “Composite Action in Concrete Filled
Tubes,” Structural Engineering, ASCE, Vol 125, No 5, May 1999.

61
4. Roeder, C.W., “Composite Behavior Between Steel and Concrete Systems for Lateral
Loads,” ASCE Special Publication, Proceedings, Engineering Foundation Conference on
Composite Construction IV, Banff, Canada, June 2000.
5. Taylor, A.W., “A Study of the Behavior of Simply Supported Composite Beams,” A thesis
submitted in partial fulfillment of the requirements for the degree of Master of Science in Civil
Engineering, University of Washington, Seattle, WA, 1985.

62

S-ar putea să vă placă și