You are on page 1of 10

Engineering Structures 24 (2002) 29–38

www.elsevier.com/locate/engstruct

Cyclic behaviour of concrete filled steel tubular column to steel


beam connections
J. Beutel a, D. Thambiratnam a,*
, N. Perera b

a
Physical Infrastructure Center, School of Civil Engineering, Queensland University of Technology, GPO Box 2434, 2 George Street, Brisbane,
QLD 4001, Australia
b
Director, Robert Bird and Partners, Brisbane, QLD 4001, Australia

Received 31 May 2000; received in revised form 23 July 2001; accepted 3 August 2001

Abstract

An experimental investigation into the behaviour of composite column-to-beam connections using ten large-scale connections
has been conducted, four under monotonic loading and six under cyclic loading. This paper presents the details and results of the
six cyclic tests. Details and results from the monotonic tests are reported in a companion paper. All connections consisted of a
concrete-filled steel tube (CFST) column (circular), a compact universal beam section and a shop fabricated connection stub. Mono-
tonic testing was first carried out and the results, were used to conduct the cyclic tests. Each specimen tested cyclically had the
same general form, consisting of a direct connection of the beam to the tube wall (using flange connection plates, and web cleat
plates) and reinforcing bars welded to the top and bottom flanges, embedded into the concrete core. It was found that provided the
connection to the face of the column had the capacity to force the hinge into the beam itself, the specimen behaved in a very stable
fashion, and achieved ductility levels which were only limited by the beam and the nature of the test itself.  2001 Elsevier Science
Ltd. All rights reserved.

Keywords: Concrete filled steel tube; Beam column joints; Cyclic behaviour; Seismic design

1. Introduction sustained over many cycles of reversible loading, and


must not compromise the load carrying capacity of the
The current design principles used to ensure that struc- column. These connections therefore require a design
tures can sustain large seismic loads, much larger than procedure that incorporates a strength hierarchy in all of
their elastic limits, have very sound mathematical and its components and mechanisms to ensure yielding at
engineering groundings. Recent seismic events have pre-ordained locations to protect others that are more
shown that while the existing design approach is sound, critical.
there are still many issues which require further research Framing systems that use concrete filled steel tube
and study, particularly in the area of connections. (CFST) columns and compact structural steel beams are
In many structural typologies, it is the beam to column not only superior to most other types from an economic
connection that provides a structure with its lateral stiff- and construction standpoint, but also in their structural
ness, and it is also these connections that play a vital performance when elasto-plastically dissipating energy
role in building survival during a seismic event. under high strength demand. This is due to their high
Depending on the designer’s intention, these connections strength, and exceptional ductility. As a result, these
must either be strong enough to form a hinge in the con- members are very desirable for application into tall
nection beam element, or undergo yield itself while buildings located in regions of high seismic risk, parti-
maintaining its shear transfer capability and ability to cularly when combined with the efficient and innovative
generate the required bending capacity. This must be construction techniques which have been developed for
their erection, and the lower building costs that result.
Due to inadequate understanding of the behaviour of full
* Corresponding author. strength connections between these elements and vir-
E-mail address: d.thambiratnam@qut.edu.au (D. Thambiratnam). tually no guidance into their design, these components

0141-0296/02/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 1 - 0 2 9 6 ( 0 1 ) 0 0 0 8 3 - 9
30 J. Beutel et al. / Engineering Structures 24 (2002) 29–38

Nomenclature
E Young’s modulus
F⬘c Concrete compressive strength
Mc Moment at the face of the connection
Mp Plastic moment capacity of the beam section
P Beam tip load
Py Yield tip load
sy Yield tensile stress
sult Ultimate tensile stress
q Rotation of the connection (radians)
d Beam tip deflection
dy Yield beam tip deflection

have yet to reach their full potential in this application. tions are class 1 and have the capability to form a plastic
The present project was undertaken to address these hinge with adequate rotational capacity for application
shortcomings, and work towards developing a determin- in a moment resisting frame in region of high seismic
istic procedure for the design of a suitable connection. risk. The column section used for all specimens was a
406 mm diameter pipe with 6.4 mm nominal wall thick-
1.1. Previous research ness, fabricated from 350-grade plate, cold formed with
a single full strength butt weld (FSBW). The flanges of
In recent times there has been a significant amount of the beam were connected directly to the steel column
research into this area and most of the details from this via flange connection plates that were the same thickness
research have been summarised in earlier papers ([1,2]). as the beam flanges, and FSBWs. The web of the beam
It is interesting to note that many of the connections was attached to the web cleat plate via two fabrication
investigated have required a high material content and/or bolts and continuous fillet welds. All plate steel used was
an excessive amount of costly fabrication, have inad- of grade 350 MPa minimum.
equate access for fabrication and inspection, and pro- Standard ‘Y’ grade reinforcing bars with a nominal
mote air entrapment and/or damage to large embedded yield of 400 MPa were welded to the top and bottom
items during the concrete filling (pumping) operation. flanges and embedded into the concrete core. Each of
Any one of these problems could render the connection the specimens had varying diameter bars, with varying
unsuitable for most applications. anchorage details (refer to Table 1). The notation used
for each specimen gives load application type (‘C’ for
cyclic), maximum nominal bar size (except for the case
2. Connection details and testing procedure of SC410c) and the bar end detail. ‘s’, indicating a
straight bar with no special end detail, ‘c’, a cogged bar
2.1. Connection type and fabrication (standard cog to the Australian concrete design code,
AS3600 [6]) and ‘f’ bars which travelled through the
Several connection types previously investigated by tube and were welded to the opposite side of the column.
other researchers (refer to Beutel et al. [2]) were con- Each specimen was filled with a standard
sidered for this study, and the one selected was chosen 40 MPa/20 mm maximum aggregate ready-mix concrete
due to its potential for application into moment resisting from a local supplier. Tables 2 and 3 give the strength
frames with regard to performance, economy and con- characteristics for all of the materials.
structability. The connection type that was the focus of Table 4 shows the calculated moment capacities of
this experimental program shown in Fig. 1 and described each connection (using nominal material strengths), and
in Table 1, takes the same general form as the one stud- a summary of the tensile capacities of the flange connec-
ied briefly by Alostaz and Schnieder [3,4] who tested tions. This capacity is a combination of the through shear
one experimental specimen, and conducted an analytical strength of the FSBW of the flange connection plate to
investigation. Specimens SC20s, SC24s, SC24c, SC24f the tube, and reinforcement tensile strength assuming
and SC28c were fabricated using a universal beam sec- full anchorage. The bending strengths were predicted by
tion 360 mm deep, and specimen SC410c used a beam summing the expected capacities of each of the individ-
section 410 mm deep. These sections weighed 44.7 and ual components of the connection, these being the web,
53.7 kg/m, respectively, and had a nominal yield of the flange (which is in general controlled by the through
300 MPa. According to Eurocode 3 [5] both these sec- shear strength of the column wall), and the bars. Table
J. Beutel et al. / Engineering Structures 24 (2002) 29–38 31

Fig. 1. Connection configuration.

Table 1
Specimen description

Anchorage shape and


Specimen (1) Beam depth (2) Flange plate size (3) Bar size and number (4)
depth (5)

SC20s 360 150×170×10 4×Y20 Straight, 350 mm


SC24s 360 150×170×10 4×Y24 Straight, 350 mm
SC24c 360 150×170×10 4×Y24 Cogged, 350 mm
SC24t 360 150×170×10 4×Y24 Fixed to opp. side
SC28c 360 150×250×10 2×Y28+1×Y20 Cogged, 350 mm
SC410c 410 150×180×10 2×Y28+2×Y24 Cogged, 350 mm

Table 2
Steel properties

Average thickness (mm)


Element (1) E (GPa) (2) sy (MPa) (3) sult (MPa) (4)
(5)

Column 200 350 450 6.04


Beam web 200 400 525 6.87 and 7.58a
Beam flange 200 350 510 9.69 and 10.63a
Stub web 200 400 510 6.00
Stub flange 200 400 510 10.05
Y-barb 200 400 550 N/a

a
Values for 410 UB 53.7 kg/mtr.
b
Values for 20 mm diameter bar typ.

Table 3
Concrete properties

Specimen (1) SC20s (2) SC24s (3) SC24c (4) SC24t (5) SC28c (6) SC410c (7)

F⬘c (MPa) 45.0 45.3 45.6 41.9 44.0 44.2


Ec (GPa) 33.1 32.7 36.5 34.5 33.8 33.5
32 J. Beutel et al. / Engineering Structures 24 (2002) 29–38

Table 4
Specimen capacities

Flange-column weld Tensile capacity of bars Calculated capacity (kNm) Expected hinge location
Specimen (1)
capacity (%)a (2) (%)a (3) (4) (5)

SC20s 43 100 276 Conn.


SC24s 43 145 361 Beam
SC24c 43 145 361 Beam
SC24t 43 145 361 Beam
SC28c 60 125 362 Beam
SC410c 37 147 476 Beam

a
All percentages based on nominal yield capacity of flange.

4 also presents the expected hinge locations. This is strength and rotational stiffness characteristics, failure
based on whether the connection’s strength at the face sequence and important strain responses throughout the
of the column can generate 290 kNm, being the plastic specimen. These results were used to establish the cor-
bending capacity of the beam section (Mp) 100 mm past rect loading parameters for the cyclic tests according to
the end of the bars (for specimens with 360 UB ATC-24 [7], primarily the amplitude of the yield dis-
members). It is expected that the connection must have placement, dy, and specimen yield load, Py. dy was taken
a capacity greater than 335 kNm to form a hinge in as the value of the displacement at the intersection of
the beam. the initial tangent stiffness of the force displacement
It should be noted that the connections have been curve, with the plastic bending capacity of the system.
designed to be shop fabricated, and field spliced away Fig. 2 shows the typical load history that consists of
from the critical connection region. This is a popular stepwise increasing stages, each stage being made up of
design and construction technique that allows for the a number of cycles that are symmetric in peak control
most heavily loaded welds (those being the FSBWs at values. The first six cycles applied to the specimen, con-
the column face) to be completed in the shop, where sisting of two cycles each at 25, 50, and 75% of Py, were
quality can be assured. The design of welds, how they elastic and were conducted under load control. ATC-24
are performed and the construction sequence and tech- [7] suggests that force rather than deflection control is
niques employed are very important, as these can have used for these cycles as the specimen may have very
a significant influence on the performance of the connec- high initial stiffness, and deformations may be difficult
tion and should closely mirror actual practice. These to predict and control. These cycles were then followed
issues are discussed in the companion paper [2]. by inelastic cycles, which were conducted under deflec-
tion control. The amplitude for each successive step
2.2. Test set-up, loading and instrumentation increased by the yield deflection of the specimen. Three
cycles were completed for each step up to a beam tip
As only the flexural behaviour of the connection was deflection of 3×dy, after which two cycles per step were
being studied, T-shaped specimens representing an conducted as ATC-24 [7] suggests.
external frame connection, were tested. Each connection Since the actuator and data logging were manually
specimen was loaded using a reaction frame and tested controlled, each cycle’s frequency could not be con-
in the vertical plane as a cantilever beam test. A detailed trolled exactly. For each specimen’s elastic cycles the
discussion of the test set up and specimen support are actuator was load controlled at a rate of approximately
presented in the companion paper [2]. 400 N/s, after which deflection control was used at a rate
Specimen load was applied via a 150 tonne hydraulic of approximately 0.135 mm/s for the inelastic cycles.
jack at the top of the column and a 30 tonne manually These rates were kept constant for each cycle, in both
controlled hydraulic actuator at the beam tip. The col- the loading and unloading branches of each excursion,
umn load was approximately 1000 kN, which is 12.5% as suggested by ATC-24 [7]. All data logging was car-
of the squash load of the column. This load represents ried out using a Hewlett Packard 3497A Data
a typical value for an end column in such an application, acquisition/Control unit linked to a Hewlett Packard
and was kept to within 5% throughout the duration of Pentium II desktop computer. For each cycle, the data
each test. This was deemed acceptable, as the axial was logged at approximately 10 kN increments until
stresses produced throughout the specimen were com- specimen yield, then at increments of approximately
paratively low. 3 mm beam tip deflection.
As discussed in the companion paper, during the first Each specimen was fitted with a variety of devices to
stage of testing, specimens were loaded monotonically at record displacements, loads and strains. Strain gauges
the beam tip until specimen failure, giving basic flexural were also placed on the column, beam and along the bars
J. Beutel et al. / Engineering Structures 24 (2002) 29–38 33

Fig. 2. Cyclic load history.

in a number of predetermined, highly stressed locations.


Rotation transducers were placed at the column to beam
intersection point and on the centre line of the beam
100 mm from the face of the column, so that connection
rotation could be accurately assessed. Four linear voltage
displacement transducers (LVDTs) were placed through
the depth of the connection to provide an alternative
method to measure connection rotation and are also used
to determine the depth to the neutral axis within the con-
nection.

3. Test results and discussion

The results of the cyclic tests are presented and dis-


cussed in this section. The load versus beam tip deflec-
tion (P–d) response has been used to compare the per-
formance of each of the connections. In these plots load
Fig. 3. Cyclic response of SC20s.
has been normalised with respect to Py, which is the load
required to generate the full plastic bending strength of
the framing beam at the face of the column. Connection
rotation (q) is also discussed and is defined as the rela-
tive rotation of the connection between the column’s
centre line and a point on the beam 100 mm from the
column’s face. This rotation is reported as a percentage
where 1% is equal to 0.01 rad. The Flexural Ductility
Ratio (FDR) is the peak deflection for each specimen,
divided by the yield deflection of the specimen, dy. The
peak deflection is the largest deflection achieved during
an inelastic loading stage, provided its resistance has not
fallen below a specified minimum load of 0.8Py during
any of the cycles in that stage [8].

3.1. General performance

Figs. 3 and 4 show the beam tip response for speci-


mens SC20s and SC24s, respectively. Both connections
were able to mobilise Mp at the face of the column dur-
ing the second stage of inelastic loading and produce a
degree of yielding in the connection beam outside the Fig. 4. Cyclic response of SC24s.
34 J. Beutel et al. / Engineering Structures 24 (2002) 29–38

influence of the flange bars. Each was able to sustain six


inelastic cycles that were able to generate 0.8Py. Each
specimen’s initial stiffness was similar to its monotonic
counterpart, and therefore adequate to classify it as a
full strength rigid connection. Each specimen’s loading
stiffness began decreasing markedly under 2×dy, and by
the third stage of inelastic loading both their strength
and unloading stiffness had decreased to under 50% of
its yield value and the test was terminated.
SC20s underwent a very similar failure to its mono-
tonic counterpart, suffering from the formation and
propagation of a crack in the tube wall at the corner of
the beam flange to column connection in both the top
and bottom flanges. This crack began to form under
2×dy, and by 3×dy a crack had developed along the entire
length of the beam flange to tube FSBW. SC24s
underwent an altered failure mode to that experienced Fig. 6. Cyclic response SC24t.
by the same connection under monotonic loading. This
was due to anchorage failure of the bars in both flanges,
approximately 93% by the last cycle), and under 5×dy
resulting in load transfer from the bars into the flanges
their resistance had fallen below 0.8Py and the test
and excessive connection rotation. This in turn caused
was terminated.
the tube wall to tear at the corner of the flange and a
Both specimens exhibited very little damage to the
fracture across the full width of the beam flange eventu-
connection at the column face, except the yield of the
ally formed. A large amount of spalled concrete was
reinforcing bars, and the formation of a crack in the tube
uncovered around the bars, and it would seem that the
at the corner of the beam flange to column connections.
entire anchorage around both the top and bottom bars
These cracks had not propagated through the tube how-
underwent a block failure.
ever, but had only formed on the outside surface at the
Figs. 5 and 6 show the beam tip response for speci-
root of the weld. Inspection of the concrete surface
mens SC24c and SC24t, respectively. Each specimen
around the bars indicated that no appreciable damage
was able to generate upwards of 125% of Mp at the face
had occurred to the anchorages.
of the column under 3×dy and complete eleven cycles
Specimen SC28c achieved very similar results to
where specimen resistance remained greater than 0.8Py.
specimens SC24c and SC24t, as shown in Fig. 7, reach-
Each formed a plastic hinge in the beam adjacent to the
ing strengths upwards of 125% of Mp at the face of the
end of the flange bars during the second stage of inelastic
column under 3×dy and was able to complete 10 cycles
loading, and under 3×dy the flanges experienced signifi-
where specimen resistance was greater than 0.8Py. It did
cant buckling. Under 4×dy each specimen’s resistance
however, suffer from a larger decrease in specimen stiff-
gradually fell below the specimen’s yield load (to
ness during stages 3 and 4 of inelastic loading and as a

Fig. 5. Cyclic response of SC24c. Fig. 7. Cyclic response SC28c.


J. Beutel et al. / Engineering Structures 24 (2002) 29–38 35

consequence its resistance fell to 0.72Py under 4×dy and identical, the underlying reasons for its formation were
the test was terminated. By this stage the connection had different. In specimen SC20s, the original capacity of the
developed a crack at one corner of the top beam flange reinforcing bars was not adequate to keep stress levels in
to column connection at a similar location to that in the the flange below the through shear strength of the col-
previous specimens. This crack went through the thick- umn to flange connection. During cycles 7–9 (1×dy) ten-
ness of the tube and was approximately 15 mm long. sile flange strains were limited below 1000 micro-strain
After the test some evidence of anchorage failure was and all the bars had yielded in both tension and com-
found around the bars in the vicinity of the top flange pression. During cycles 10–12 (2×dy) however, flange
where substantial cracks had developed, but full anchor- strains increased above 1400 micro-strain causing the
age failure as witnessed in SC24s had not yet occurred. formation of the crack in the tube. Thus, after the top
Specimen SC410c suffered from an anchorage failure bars reached yield, load was transferred into the flange
primarily around the bottom bars resulting in a similar causing the tube to crack. By the end of cycle 12 the
failure mode to specimen SC24s with a crack forming connection’s stiffness had decreased to less than 25%
in the tube at the corner of the bottom flange connection. of its elastic stiffness, and under subsequent cycles the
This test had to be stopped prematurely, however, connection underwent excessive rotation (⬎2.5%) caus-
because the lateral restraining system failed, causing the ing the flange to tube weld to crack along its full length.
beam to buckle sideways. Once this had occurred, the Specimen SC24s had the capacity to avoid this failure
jack could not load the beam in a satisfactory manner. type under monotonic loading but under cyclic loading
Inspection after testing revealed that the anchorages had suffered from anchorage failure in both the top and bot-
suffered a large amount of cracking and tears had tom flanges. During cycles 7–9 (1×dy) the bars reached
developed in the steel column at the same location as the yield in tension, and tensile flange strains were kept
first two cyclic specimens. The specimen still generated below 1000 micro-strain. During cycles 10–12 (2×dy)
110% of the plastic strength of the beam during cycles tensile flange strains increased above 1000 micro-strain
10–12 (2×dy) and produced a degree of yielding in the and a significant reduction in the specimens initial elastic
beam section. After cycle 12, however, the peak speci- stiffness occurred (from an average of 2.3 kN/mm for
men resistance had fallen to 0.84Py, and the specimen cycles 7–9, to 1.25 kN/mm for cycles 10–12). The speci-
completed 6 cycles where specimen resistance remained men was still able to generate its yield load during these
greater than 0.8Py (refer to Fig. 8). cycles, but under three times the yield deflection, the
entire anchorage failed. This caused a volumetric
3.2. Discussion increase of the concrete behind the tension flange and
the strains in the tension flange to increase beyond 1400
Information on the cyclic performance parameters micro-strain. The failure of the anchorages had limited
achieved for each specimen is presented in Table 5. The the load transferred into the bars and increased the stress
rapid deterioration in the stiffness and strength of speci- levels in the flanges, resulting in the same failure mode
mens SC20s and SC24s resulted in each specimen only as specimen SC20s.
being able to achieve a FDR of approximately 2. While The decrease in stiffness of both specimens between
the resultant failure mechanism for each specimen was each stage of loading was primarily due to a decrease
in the rotational stiffness of the connection. Under mon-
otonic loads, connection stiffness deterioration was
strongly linked to tensile flange strains. When these
strains exceeded approximately 800 micro-strain a rapid
decrease in stiffness resulted. This strain level corre-
sponds to the maximum load the flange plate to tube
FSBW can carry based on its through shear strength
(which corresponded to a flange strain of approximately
650 micro-strain). These strain levels where not reached
in SC20s and SC24s during the elastic cycles, yet the
connection’s stiffness in the second 3 elastic cycles
decreased by 15 and 31% compared to the connection’s
initial stiffness in its first three cycles (for SC20s and
SC24s, respectively). This degradation in stiffness is due
to the cyclic loads destroying the bond between the bar
and the concrete, which explains the larger decrease in
elastic stiffness for SC24s that has much larger anchor-
age demands. This deterioration also caused the bars to
Fig. 8. Cyclic response SC410c. ‘slip’ within their anchorages under successive cycles,
36 J. Beutel et al. / Engineering Structures 24 (2002) 29–38

Table 5
Cyclic specimen performance parameters

Number of plastic cycles


Specimen (1) FDR (2) Beam hinge rotation
(5)

2×dy (3) 3×dy (4)

SC20s 1.93 0.8 NA 6


SC24s 2.01 1.3 NA 6
SC24c 4.14 0.9 3.3 11
SC24t 4.11 1.1 2.9 11
SC28c 3.04 1.1 3.3 10
SC410c 2.47 2.1 NA 6

and lead to the connections forming ‘soft regions’ in the under subsequent cycles as specimen resistance
beginning of their response for each cycle. This deterio- decreased. Each test was terminated under five times its
ration of the bond was supported by a gradual increase yield deflection due to specimen resistance decreasing
in tension flange strains at similar loads for each success- below the specified minimum. The main plastic rotations
ive cycle. occurring in the specimen were at the beam hinge which
Each specimen was only able to achieve connection achieved rotations greater than 3.5% under 4×dy. This
rotation levels of the order of 1.2–1.5% (only approxi- combined with a FDR ⬎4 indicate that these specimens
mately 0.75% plastic rotation) during the second stage as a whole had attained an excellent level of ductility.
of inelastic loading. Under 3×dy these rotations increased Connection stiffness deterioration suffered by these
to beyond 2%, which resulted in connection failure in specimens during the elastic cycles was 23 and 11% for
both specimens. It should be noted that many researchers SC24c and SC24t, respectively. The magnitude of these
in this field consider a plastic rotation of 1.5% to be the decreases during the elastic cycles was directly related
minimum required from a connection if it is to dissipate to the decrease in the stiffness of the entire specimen.
the required energy through plastic deformation needed Their performance was an improvement on that of
to justify the use of Rw=12 [9]. This rotation however SC24s indicating the beneficial effect of the improved
need not take place at the point of connectivity to achi- anchorage. SC24t had the least elastic stiffness reduction
eve the required ductility. The rotation may also occur of all the specimens using Y24 bars due to its superior
in the beam section, which has many advantages over anchorage detail. Specimen SC20s, however, with
the connection yielding including the superior ductility straight Y20 bars matched this level of deterioration.
levels capable from a compact steel section, and the pro- This would tend to suggest that as long as the bars have
tection afforded to the column elements. adequate anchorage even straight bars could withstand
Specimens SC24c and SC24t were tested to study the cyclic load application satisfactorily.
effect of improving the anchorage capacity of SC24s. Specimen SC28c was tested to assess a number of
Both of these specimens were able to achieve FDRs of different design issues, primarily the effect of altering
approximately 4.1. Tensile flange strains kept below the strength proportions of the reinforcing bars and the
1000 micro-strain during cycles 10–12 (2×dy), and the connection of the flange to the tube, the performance of
large reductions in specimen and connection stiffness the larger bars sizes, and having varied bar sizes on the
noted in the previous specimens were reduced by same specimen. The results achieved were similar to
approximately 400%. Stiffness deterioration between those of the previous two specimens, but also had some
each stage was stable resulting in the excellent form of differences. The specimen suffered a 35% decrease in
the hysteresis behaviour (refer to Figs. 5 and 6). During its connection stiffness during the elastic cycles which
cycles 13–15 (3×dy) tensile flange strains only increased was considerably higher than those of SC24c and SC24t.
to approximately 1100 micro-strain, and there was a It was only able to achieve a FDR of 3.04 as under 4×dy
large discrepancy between the stiffness deterioration suf- its resistance fell to 0.72Py. The accelerated decrease in
fered by the specimen as a whole, and the small stiffness strength suffered by this specimen was due to a larger
reduction occurring in the connection. By this stage, the decrease in the stiffness of the connection (twice the
beam hinge had developed severe local flange buckles decrease suffered by SC24t) between inelastic loading
in both the top and bottom flanges and these were caus- stages 3 and 4. After testing, the surface of the concrete
ing the decrease in stiffness and strength of the speci- surrounding the bars was inspected for any visual signs
men. of anchorage failure. Some cracking was found around
The connection achieved maximum rotations of the the top flange bars indicating the anchorage in this zone
order of 1.3% during cycles 13–15, which decreased was close to, if not it had failed. This accelerated anchor-
J. Beutel et al. / Engineering Structures 24 (2002) 29–38 37

age deterioration was the reason for this specimen’s There are a number of interesting points and con-
higher rate of stiffness deterioration and lead to the clusions that can be drawn from the results of the cyclic
reduced performance levels. tests. Roeder and Foutch [8] report that a FDR of
Flange strains behaved in a similar fashion to those of between four and six can be related approximately to a
the previous two specimens, with strains reaching 1100 Rw factor of 12, which is the required ductility demand
micro-strain during cycles 7–9. Under 2×dy the flanges from a moment resisting frame designed per the Uniform
reached yield in compression and attained 1300 micro- Building Code [10]. Results from both SC24c and SC24f
strain in tension, but during cycle 11 the tension strains would suggest that these specimens achieved these duc-
suddenly increased beyond yield. This increase tility levels, making them suitable for incorporation into
coincided with the onset of the specimen’s resistance framing systems with the highest possible ductility
deteriorating, and it is expected that this was the point demands of any structure. Whilst this should not be
at which the anchorage’s began to fail. As a result, this taken as the only measure of a connection’s/specimen’s
specimen was only able to generate 1% plastic rotation seismic performance, the excellent form of the hysteretic
from the beam hinge formed at the end of the bars. behaviour from each of these specimens would tend to
SC410c was tested to assess the cyclic performance suggest that such a system, using such details, would
characteristics of the connection using a deeper beam behave very well during a severe seismic event. It should
section, and assess the accuracy of the existing capacity also be remembered that the primary factor limiting the
prediction formula. This specimen suffered from global FDR attained by these specimens was the capability of
lateral buckling of the framing beam due to the bearings the hinge in the beam (which was subjected to severe
on the lateral restraining system running out past the localised flange buckling), and not due to failure or
channel bearer. This occurred due to inaccurate place- deterioration of the connection itself. This beam hinge
ment of the bearing plates. As a consequence the test could still achieve plastic rotation levels in the region of
had to be stopped prematurely. By the last cycle which 3%, which is more than adequate for this application
was completed (12, 2×dy) the specimens resistance had The rapid deterioration in the stiffness and strength of
already dropped to 0.84Py and was deteriorating rapidly. specimens SC20s and SC24s resulted in each specimen
It is therefore expected that the specimen would not have only being able to achieve a FDR of just over 2. The
successfully completed the next loading stage and the poor cyclic performance of these specimens and SC28c
achieved FDR of approximately 2.4 is an accurate pre- and SC410c tend to suggest that designing for connec-
diction of this specimen’s ductility. This rapid deterio- tion yield with this type of connection is not rec-
ration in the specimen’s response was due to the failure ommended. In each case where this occurred, the con-
of the bottom bar anchorage. The plastic rotations occur- nection suffered from severe localised demands on the
ring in the beam hinge were approximately 2.1% during tube wall causing tearing and leading to the flange to
cycle 10, which decreased during subsequent cycles due tube weld fracturing. This not only severely and rapidly
to the anchorage failure. diminished the capacity of the connection resulting in
poor FDRs, but harmed the load carrying capacity of
the column itself by damaging the tube and therefore
4. Design considerations destroying the lateral confining pressure which it was
providing to the concrete core. Those specimens that
The criteria for acceptable performance for the type were designed for the hinge to form in the beam, and
of connection investigated herein are dependent on its were successful, did not suffer from these effects.
intended application. For non-seismic applications, this Adequate bar anchorage is possibly the most
is based solely on its strength and stiffness character- important aspect of this connection’s design. As dis-
istics, however seismic applications also require criteria cussed above, each of the design methods used seemed
to assess the energy dissipation capacity and stability of conservative, but the likely impact of connection failure
the connection/system. due to anchorage failure would suggest this approach be
Based on the tests that were conducted on monotonic warranted. For non-seismic applications it may be poss-
specimens, the existing design approach to the connec- ible to design these connections with straight bars, pro-
tion was accurate, and the results that were gathered vided adequate anchorage can be achieved. For seismic
from the testing supported its formulation. The results applications, however, it is suggested that all bars should
show that such a connection could be applied into non- have a 90 degree cog as a minimum. Designing these
seismic applications as a rigid connection with adequate anchorages using the standard design provisions of the
strength and stiffness characteristics. The bars were very Australian concrete design code (AS3600 [6]) or other
effective in reducing stress levels in the flanges and appropriate codes will result in a conservative design due
accepting both tensile and compressive loads from the to the conservative approach of these codes, and the con-
framing beam and distributing these directly into the fining effects of the tube improving the anchorage
concrete core of the column. capacities of the bars. Recent work conducted by Esfah-
38 J. Beutel et al. / Engineering Structures 24 (2002) 29–38

ani and Rangan [11] that takes into account this enhance- 앫 When the connection’s strength was such that a full
ment gave more accurate results however, it would seem hinge formed in the framing beam, as was the case
prudent in an actual design situation not to rely on the with specimens SC24c and SC24t, the specimen’s
anchorage enhancements provided by the tube. It is well overall seismic performance improved, and resulted
known that such columns under long-term loading in FDRs of the order of 4.1 and plastic hinge rotations
undergo load shedding from the concrete to the steel of the order of 3.5%.
tube. This has the effect of reducing the confining press- 앫 These specimens suffered a very minor tube wall tear
ure on the concrete itself due to a decrease in the lateral at the corner of the beam flange to column connec-
‘bulging’ of the concrete. This effect combined with the tion, which did not however propagate entirely
action of the beam flange connection pulling the tube through the wall.
away from the face of the concrete would suggest that 앫 The strength, stiffness and ductility results would sug-
relying on confinement is not advisable. gest that such a connection/system would be suitable
A further design consideration apparent from these for applications where large seismic loads are a possi-
tests was the excellent performance of both the FSBWs bility.
of the flanges, and the fillet welds connecting the bars 앫 Based on the behaviour of the models tested, it is
to the flanges. No cracking of the FSBW connecting the possible to design a joint with a component strength
beam flange to the flange connection plates occurred for hierarchy that will rely on the ductility of the beam
any of the specimens. The only occasions when the weld only and the design approach used gives an accurate
between the flange plates and the column face failed was prediction of connection strength.
after a bar anchorage failure, or when stress levels in the
flange were too large causing a tear in the column wall.
This would suggest that the adopted FSBW details were
References
adequate, and the bars were very effective in reducing
stress levels in the flanges thereby decreasing stress lev- [1] Tubular Structures VIII — Proceedings of the Eight International
els in these welds and ensuring their survival. Symposium on Tubular Structures 1998., 1998.
[2] Beutel JG, Thambiratnam DP, Perera NJ. Monotonic behavior of
composite column to beam connections. Journal of Engineering
Structures 2001;23(9):1152–61.
5. Conclusions [3] Alostaz Y, Schneider S. Analytical behavior of connections to
concrete-filled steel tubes. Journal of Constructional Steel
Research 1996;40(2):95–127.
The first stage of this project investigated the perform- [4] Alostaz Y, Schneider S. Experimental behavior of connections to
ance of four specimens with varying size reinforcing concrete-filled steel tubes. Journal of Constructional Steel
bars welded to the top and bottom flanges and embedded Research 1998;45(3):321–52.
into the concrete core. These connections were tested [5] CEN. prENV 1993-1-1 Eurocode 3, Design of steel structures,
monotonically, the results of which are reported in a Part 1.1 — General rules and rules for buildings. Brussels: Euro-
pean Committee for Standardisation (CEN), 1993.
companion paper. During the second stage of testing [6] SAA. AS 3600 — 1990 concrete structures. Sydney: Standards
these connections were tested under cyclic load appli- Association of Australia, 1990.
cation with varying results. Some of the important con- [7] ACT — 24. Guidelines for cyclic testing of components of steel
clusions are listed below. structures. California: Applied Technology Council, 1992.
[8] Roeder CW, Foutch DA. Experimental results for seismic resist-
ant steel moment frame connections. Journal of Structural Engin-
앫 When the plastic hinge formed within the connection, eering 1996;122(6):581–8.
as was the case with specimens SC20s and SC24s, [9] Earthquake performance of civil engineering structures, COST
the performance of the specimen was poor. FDRs in C1 — Seismic working group report., 1996.
the order of only 2 were achieved, with connection [10] ACI-318. Building code requirements for reinforced concrete.
Detroit: American Concrete Institute, (1989, revised 1992).
rotation levels of approximately 1.5%. [11] Esfahani MR, Rangan BV. Evaluation of proposed revisions to
앫 Connections designed to yield at the columns face AS3600 bond strength provisions. Australian Journal of Struc-
would not be suitable in seismic applications. tural Engineering 1999;2(1):31–5.

You might also like