Cyclic Behavior of Buckling-Restrained

Braces for Ductile End Cross Frames

LYLE P. CARDEN, AHMAD M. ITANI, and IAN G. BUCKLE
Furthermore, serviceability requirements immediately following
an earthquake mean that minimum levels of residual
strength and stiffness must be maintained, even though the
braces are secondary members with respect to gravity loads
(in a straight bridge). This requirement is difficult to guarantee
with buckled members. Thus, many of the X-braces will
need replacement following an earthquake.
To address the limitations of angle members (suboptimal
energy dissipation, uncertain residual stiffness, and member
replacement), an end cross frame with the same or better
energy dissipation at less drift, with reliable post-earthquake
stiffness and strength and a low likelihood of need for replacement,
is desired. For this reason, the use of bucklingrestrained
braces (BRB), instead of the angle members, has
been investigated, and the results are reported in this paper.
One example of a BRB is the “unbonded” brace, developed
and manufactured by Nippon Steel Corporation in
Japan (Wada, Saeki, Takeuchi, and Watanabe, 1989), as illustrated
in Figure 1. These braces are constructed from a
cruciform or rectangular steel core of specially formulated,
low-yield-point steel. The core is surrounded by a debonding
material and encased in a steel hollow tube filled with
grout. The debonding material allows the core to slide rela-
Lyle P. Carden is a post-doctoral research scholar, department
of civil and environmental engineering, University
of Nevada, Reno, NV.
Ahmad M. Itani is associate professor, department of
civil and environmental engineering, University of Nevada,
Reno, NV.
Ian G. Buckle is professor, director of center for civil
engineering earthquake research, department of civil
and environmental engineering, University of Nevada,
Reno, NV. Fig. 1. Buckling-restrained brace.
128 / ENGINEERING JOURNAL / SECOND QUARTER / 2006
tive to the encasement. Unbonded braces used in large-scale
bridge experiments (Figure 2) were shown to improve the
energy dissipation in the end cross frames compared with
X-braces and, as degradation in stiffness is prevented, they
were less likely to require replacement soon after an earthquake,
as is necessary for single-angle X-braces (Carden et
al., 2006a).
Buckling-restrained braces designed for use in end cross
frames of bridges are typically much smaller, with shorter
core lengths, than those used for concentrically braced frames
in buildings. The objective of this paper is to evaluate the
inelastic cyclic axial behavior of braces as proposed for use
in ductile end cross frames of steel girder bridges. Different
loading histories, including slow-cyclic and dynamic-cyclic
loading were investigated.
EXPERIMENTAL SETUP
Experiments were performed on seven buckling-restrained
braces, provided by Nippon Steel Corporation, termed “unbonded”
braces. Four of the braces were used solely for axial
cyclic experiments, while the remaining three were used in
shake table experiments on the large-scale bridge model
before being tested to failure using cyclic axial loads. Axial
experiments were performed to evaluate overall brace behavior,
cumulative displacement capacity, effect of different
loading histories, and effect of dynamic loading. Loads were
applied to the braces using the load frame shown in Figure 3.
The dimensions and connection configurations for the braces
are illustrated in Figure 4. Experiments were performed on
the first brace with pin-ended connections between the gusset
plate and the load frame grips as shown in Figures 4a and
5, while the remaining braces used bolted, moment-resisting
connections, which were designed to be slip critical based on
the serviceability criteria in the AISC Load and Resistance
Factor Design Manual of Steel Construction provisions
(AISC, 2001) (Figures 4b and 6). The pin-ended connection
was intended to prevent bending moments in the brace. To
minimize the size of the fixed slip-critical connection, two
rows of bolts were used to connect the brace in the direction
of larger core plate dimension (in the plane of the end cross
frame in the bridge model), while the perpendicular legs
of the cruciform connection were connected with a single
row of bolts at each interface. The force at which slippage
was expected to occur was equal to 40.1 kips, which was
approximately two times the expected yield strength of the
brace. The total length of the brace was equal to 388 in.,
including the connection regions on the cruciform section
(Figure 3). The core length was equal to 20a in. between the
centers of the transition between the maximum cruciform
and minimum core sections at either end of the brace. The
cross-sectional dimensions of the core plate were 1 in. s in.
Fig. 2. Large-scale steel plate girder bridge model with
buckling-restrained braces in the end cross frames.
Fig. 3. Load frame for applying reversed cyclic
axial loads to the buckling-restrained braces.
ENGINEERING JOURNAL / SECOND QUARTER / 2006 / 129
Three of the unbonded braces were instrumented with
strain gages to measure strains on each side of the plate at
each end of the core length. The braces were also instrumented
with two displacement transducers on either side of
the brace to measure displacements across the core length of
the brace as shown in Figure 3. The displacement across the
entire length of the brace, including connection regions, was
also measured using the actuator, and the axial force was
measured by the actuator load cell.
The loading history for each of the braces differed. Brace
A was subjected to a modified ATC 24 loading history (ATC,
1992). The modification involved subjecting the brace to two
cycles at each excursion instead of three cycles as in ATC
24. The history was based on the applied force until yielding
occurred, then increased based on the displacement at yield.
As slippage in the pinned connection was considerable compared
with the deformations in the core at low-displacement
amplitudes, the history was further modified to achieve more
uniform displacements in the core. The loading rate was
slow. The resulting displacement time history for Brace A is
given in Figure 7.
The loading history for the Brace B (Figure 7) was based
on the loading history in the Draft Recommended Provisions
for Buckling-Restrained Braced Frames (BRBF) (SEAONC,
2003). This is similar to that recommended by Sabelli
(2004). This loading history is defined by a number of
cycles at different levels of displacement ductilities. After
the prescribed cycles at increasing displacement amplitude,
the specimen was cycled at constant amplitude until failure
at the maximum displacement amplitude. The displacements
were controlled using the total displacement measured in the
actuator. The yield displacement, y, used to define the loading
history was based on the estimated displacement calculated
at a 0.2% offset strain, over the core length for Brace A,
that is, 0.064 in. No slippage was observed before yielding
and thus did not affect the yield displacement. It is important
to note that the displacement at 0.2% offset strain was
considerably higher than would be estimated if a method
Fig. 4. Buckling-restrained braces setup for axial cyclic experiments with
(a) pin-ended connections and (b) moment-resisting, slip-critical connections.
130 / ENGINEERING JOURNAL / SECOND QUARTER / 2006
such as the ATC-24 procedure (based on the slope at 75% of
the yield force) was used. There is an even larger difference
when compared with a theoretical yield displacement using
a theoretical yield force and elastic modulus. As the BRBF
guidelines (SEAONC, 2003) do not define the way in which
the yield displacement is to be estimated, the method that
resulted in the more severe displacements, that is, at 0.2%
offset strain, was used. The other amplitudes were defined
based on this displacement at 2y, 4y, and 6y. The loading
history for Brace C (Figure 7) was the reverse of that for
Brace B, followed by additional constant amplitude loading
until failure applied at a slow rate. For Brace D the same
loading history was used (Figure 7) but applied dynamically
to the brace at a frequency of 2 Hz to simulate a bridge with
a 0.5-s effective period. This was equivalent to a maximum
strain rate of 25% per second.
Brace E was the brace first used in the north end of the
bridge model (Carden, Itani, and Buckle, 2006b) and had already
been subject to some inelastic deformation, although it
had not failed. It was subjected to further inelastic behavior
using axial cyclic loading with the same loading history as
that applied to Brace D, at a dynamic rate of 2 Hz (Figure
8). Brace F was the brace used in the south end of the bridge
model and was cycled to failure at constant amplitude of
±3y; with y as defined previously, at a slow rate (Figure 8).
The final brace, Brace G, was the brace used at the midspan
of the bridge model. It was heavily worked in the bridge
model and subsequently was subjected to just one cycle 9y
until failure at a slow rate of loading (Figure 8).
CYCLIC BEHAVIOR OF THE
BUCKLING-RESTRAINED BRACES
Hysteretic Properties and Connection Slippage
Excluding Brace A, the braces were designed to have slipcritical
connections. The displacement time histories in
Figures 7 and 8 indicate that where slippage occurred, it
sometimes resulted in a large difference between the total
displacement in the brace and the displacement measured
across the length of the core. As expected, slippage occurred
in the pin-ended connections of Brace A, indicated by the
large difference in the total and core displacements in Figure
7. In contrast, for other braces such as B and C, there was
little difference because slippage was prevented, or in the
case of Brace B, delayed until late in the history. Prior to
slippage, the small difference between the two measured
displacements can be attributed to elastic deformations in
the connection region. Slippage was also observed in the
dynamically loaded Braces D and E and with the large amplitude
displacements in Brace G.
Fig. 5. Buckling-restrained brace with pinned-end connections. Fig. 6. Buckling-restrained brace with fixed-end connections.
ENGINEERING JOURNAL / SECOND QUARTER / 2006 / 131
The axial force-displacement curves for each unbonded
brace are shown in Figures 9 and 10. Overall, the hysteresis
loops exhibit good energy dissipation characteristics and
repeatable behavior. The effect of slippage can also be observed
in these figures, when comparing the plots for the total
displacement with those for the displacement across the core
of the brace. The slip force for the slip-critical connections
was calculated at 40 kips based on AISC serviceability criteria
(AISC, 2001), while the apparent slip force, determined
Fig. 7. Axial displacement loading
history for Braces A, B, C, and D.
Fig. 8. Axial displacement loading history for
Braces E, F, and G in the load frame after being
subjected to deformations in the bridge model.
132 / ENGINEERING JOURNAL / SECOND QUARTER / 2006
by inspection of the hysteresis loops, was less than this value—
at typically around 30 kips. The calibrated wrench pretensioning
method is known to result in significant scatter,
due to a number of factors described in the Specification for
Structural Joints Using ASTM A325 or A490 Bolts (AISC,
2001), which is supported by these experimental observations.
Another explanation may be that the faying surface, a
clean Class A surface, did not provide the friction coefficient
assumed in design (0.33).
Fig. 9. Force-displacement hysteresis loops
for Braces A, B, C, and D.
Fig. 10. Force-displacement hysteresis
loops for Braces E, F, and G.
ENGINEERING JOURNAL / SECOND QUARTER / 2006 / 133
Maximum and Cumulative Displacements
The maximum displacements, converted to average axial
strains in the length of the core for the unbonded braces
during each of the component experiments, are given in
Table 1. For Braces E, F, and G, which were subjected to
deformations in the bridge model, the maximum axial strain
observed during these experiments is also given in Table 1.
Axial strain is used to define the maximum brace displacements
instead of ductility as it is independent of the yield
displacement and yield strength and therefore can provide
a more consistent measure between braces. The maximum
strain in brace A, with increasing amplitude cycles until failure
was 3.7%. For the BRBF history (SEAONC, 2003) the
maximum strain was approximately 1.9%, although slightly
less in cases where slippage occurred.
Cumulative displacement capacity was evaluated for the
braces using measures of cumulative plastic strain and cumulative
plastic ductility (Table 1). The cumulative plastic
strain was calculated by estimating the total plastic axial
displacement during a loading history, using a simple algorithm,
divided by the effective length of the core (20.4 in.).
The cumulative plastic ductility was calculated from the cumulative
plastic displacement divided by the yield displacement,
included for comparison with previous studies such
as Black, Makris, and Aiken (2002). As the yield strength of
each of the braces was the same, these two measures were
equivalent. Both require an estimation of the yield displacement,
which was assumed to be equal to the theoretical yield
displacement, y, over the length of the core plate given by
where
Fye = expected yield strength of the core plate
(= 32.6 ksi)
L = length of the core plate (= 20.4 in.)
E = elastic modulus (= 29,000 ksi)
This resulted in a theoretical yield displacement of
0.0229 in. This estimate of the yield displacement was
around 29% less than the average yield displacement mea-
Table 1. Maximum Strains and Cumulative Plastic Displacement Capacity of Unbonded Braces
Maximum Strain,
%
Cumulative Plastic Strain,
%
Cumulative Plastic
Brace Loading History Ductility
Bridge Load Frame Bridge Load Frame Bridge Load Frame
A Modified ATC-24 – 3.71 – 54 – 479
B BRBF Normal Static – 1.86 – 98 – 875
C BRBF Reversed Static – 1.87 – 66 – 588
D
BRBF Reversed
Dynamic
– 1.71 – 38 – 337
E
BRBF Reversed
Dynamica 1.76 1.77 41 40 366 352
F Constant Staticb 2.22 0.92 23 125 205 1109
G Constant Staticc 2.29 2.51 76 22 674 192
aLoading history applied in the load frame after the braces were removed from the south end of the bridge model following shake table
experiments.
bLoading history applied in the load frame after the braces were removed from the north end of the bridge model following shake table

experiments.
cLoading history applied in the load frame after the braces were removed from the mid-span of the bridge model following shake table

experiments.
y
FyeL
E
= (1)
134 / ENGINEERING JOURNAL / SECOND QUARTER / 2006
sured in the braces using the ATC 24 (ATC, 1992) procedure
and is approximately 36% of the displacement calculated at
the 0.2% offset strain used to define the loading history. The
large difference highlights the importance of clearly defining
the method in which yield displacement is calculated. The
theoretical yield displacement was consistent, computable,
and comparable for other systems.
Table 1 shows that the cumulative plastic strains vary greatly
between the different specimens with a range between 38
and 148% (when combining the estimated cumulative plastic
strains from the bridge excitations and load frame axial
deformations). Much of this variation is attributed to the inherent
variability of fracture that cannot be easily quantified;
however, there are certain factors that can be judged to affect
the cumulative plastic strain when comparing the different
braces. While the aim was to find a measure of cumulative
displacement that was independent of the maximum strain,
comparing the braces in Table 1 shows that the braces that
were subjected to the smallest strain amplitudes tended to
have larger cumulative plastic strains. Comparing Braces B
and C with the normal and reversed slowly applied BRBF
loading histories resulted in a 35% lower cumulative plastic
strain in the brace with reversed loading. This demonstrates
that having large amplitude cycles at the beginning of the
history reduces the overall cumulative strain capacity compared
with increasing amplitude cycles. Furthermore, applying
the loading history dynamically, as in Brace D, further
reduced the cumulative strain capacity to 39% of that in
Brace B. This reduction may be partly due to the increased
forces observed in the dynamically loaded specimen, as discussed
in the following section, but could also be attributed
to slippage that, when dynamic loads were applied, caused
a lowering of the displacement capacity. While it is difficult
to substantiate conclusions based on only three observations,
conservatism suggests that the cumulative displacement capacity
for a brace based on increasing amplitude slow cyclic
loads should be reduced for a real earthquake where the
dynamic reversed loading history is more representative—at
least until further research results are available.
From these experiments it is recommended that ductile
end cross frames using buckling-restrained braces should be
designed for a maximum deformation during an earthquake
not exceeding 2.0%. Therefore, to be consistent with the
design of isolation systems (AASHTO, 1999) and also the
BRBF recommendations (SEAONC, 2003), the maximum
considered earthquake strain is no more than 3.0% (1.5 times
the design level strain). The maximum strain limit of 3.0%
is less than the 3.7% maximum strain measured in the brace
with increasing amplitude cycles of loading. Although the
braces in the bridge model were not deformed up to strains
of 3.7% during shake table experiments, an analytical model
of the bridge was used to determine the cumulative plastic
strain in the brace for a series of different earthquakes scaled
to result in a 3.0% strain in the braces. Resulting cumulative
plastic strains for 10 different earthquake excitations
are listed in Table 2. The average cumulative plastic strain
was 26% and the mean plus one standard deviation cumulative
plastic strain was 40%. This is approximately equal to
the most conservative value of the cumulative plastic strain
(38%) measured in the axial experiments. Therefore, a 3.0%
maximum strain is considered an appropriate level for the
maximum considered earthquake and 2.0% strain is appropriate
for the design level earthquake.
The axial strains in the braces used in ductile end cross
frames of bridges can be converted to girder drifts where the
girder drift is defined by the lateral displacement in the cross
frames divided by the depth of the girders. This enables the
displacement limits to be described using the same terms as
used in building provisions. However, different girder sizes
and configurations will change the axial strain to drift ratio,
therefore using axial strain that is directly related to inelastic
deformations in the brace as the basis for defining design
displacements is more rational. This is synonymous with the
specified link rotation limits for eccentrically braced frames
in the AISC Seismic Provisions for Structural Steel Buildings
(AISC, 2002). In addition, the use of axial strain for
calculating displacement limits implies the same deformation
capacity in lower strength steels than for higher strength
steels. If ductility is used as a basis for determining a displacement
capacity, the implication is that higher strength
steels with equal stiffness have a larger displacement capacity
than lower strength steels, when, in fact, the reverse is
generally true.
Axial Yield and Ultimate Forces
Two coupon tests were performed on the LYP-225 steel used
in the manufacturing of the unbonded braces provided by
Nippon Steel (Nippon, 2002). The measured yield strengths
(32.3 and 34.1 ksi) were within 5% of the nominal yield
strength of 32.6 ksi (225 MPa). Unlike most U.S. steels,
where the nominal strength is generally defined so that the
majority of specimens are stronger than the nominal strength,
the nominal strength for the LYP-225 steel is defined in the
middle of a relatively tight range (29.7 to 35.5 ksi) of acceptable
yield strengths. Therefore, the nominal strength of 32.6
ksi was assumed to be the expected strength for this material.
The ultimate strength of the LYP-225 steel was around 35%
higher than the yield strength. The ultimate strain was much
higher at around 65% for the LYP-225 steel compared with
typical values of 30 to 35% for A36 steel (Carden, Itani, and
Buckle, 2006c).
The expected yield force for the unbonded braces, Pye,
was defined as
Pye = Fye Asc (2)
ENGINEERING JOURNAL / SECOND QUARTER / 2006 / 135
where
Fye = expected yield strength (= 32.6 ksi)
Asc = area of the steel core (= 0.625 in.2)
Thus the expected yield force was equal to 20.4 kips. The
yield force for each unbonded brace was estimated using the
force at a 0.2% offset strain and was found for each brace.
Table 3 lists the measured yield force for each unbonded
brace and shows that they were slightly larger than the expected
yield forces but within 12% for all braces.
The overstrength factor, , which describes the maximum
tensile force measured in the brace divided by the expected
yield strength, was calculated for each brace and is given in
Table 3. The factor was generally around 1.35, similar to
the difference between the yield and ultimate strengths measured
in the coupon tests. The overstrength factor was larger
in some instances, such as Brace D, which had the largest
overstrength factor. This is attributed to the high strain rate
in experiments on Brace D that increased the force in the
brace, particularly during the first cycle of the loading. This
can be observed in Figure 9 when comparing the response
to the same loading history applied slowly for Brace C and
dynamically for Braces D and E. Brace D had no previously
applied deformations and the resulting maximum force in
the first cycle was 30% larger than the corresponding force
in the first cycle of Brace C. However, the strain rate effect
was most pronounced in the first cycle of loading as subsequent
cycles stabilized to around a 10 to 15% difference
when comparing Braces C and D. Brace E had been previously
loaded in the bridge model and, as a result, the first
cycle of loading was not as high as that in Brace D, although
each cycle exhibited maximum forces around 15% higher
than in Brace C. Braces A and G, which had larger maximum
strains in the core than the other braces, also exhibited
factors greater than 1.35.
The compression strength adjustment factor, , is defined
as the maximum compression force divided by the maximum
tension force, and is calculated and listed in Table 3 for each
of the braces. The compression strength adjustment factor
was typically around 1.10 to 1.15, indicating a maximum
compression force 10 to 15% larger than the tension force.
Thus, was largest for the braces with the largest maximum
strains as shown for Braces A and G, while the reverse is
true for smaller strains as in Brace F, indicating that the compression
overstrength tended to increase as the core strain
increased. This is attributed to high mode buckling and the
Poisson effect causing increased friction between the core
and the surrounding encasement at high strains. For Brace D
the maximum compression force was actually less than the
maximum tension force due to a large observed tension force
Table 2. Cumulative Plastic Strains and Ductilities in Braces from an Analytical Model of a Bridge
Subjected to Different Earthquakes Resulting in a Maximum Brace Strain of 3.00%
Earthquake Station and Component
Amplitude
Scale Factor
Maximum
Brace Strain,
%
Cumulative
Plastic Strain,
%
Cumulative
Plastic Ductility
1940 Imperial Valley El Centro Array#9 NS 2.80 3.04 37.1 286
1966 Parkfield California Array#2 NS 1.50 3.05 20.1 155
1971 San Fernando Pacoima Dam NS 1.75 2.94 28.4 219
1977 Bucharest Building Res. Inst. NS 3.70 3.00 18.3 141
1979 Imperial Valley Array #7 230˚ 1.60 3.12 21.8 168
1992 Landers Lucerne EW 2.50 2.93 13.7 106
Sylmar Hospital NS 1.10 3.00 11.8 91
1994 Northridge
Rinaldi 229˚ 0.80 2.95 13.1 101
1995 Kobe KJMA NS 1.30 2.94 46.3 357
1999 Taiwan Chi-Chi TCU084 EW 0.80 3.13 49.2 380
Average 3.01 26.0 200
Average + 1 Standard Deviation 3.08 39.7 307
136 / ENGINEERING JOURNAL / SECOND QUARTER / 2006
in the first reversal of the dynamically loaded specimen.
Energy Dissipation in the Unbonded Braces
The area enclosed by each hysteresis loop was calculated
for Braces A through D at different cycles of loading during
the component experiments. The displacement was based
on that measured across the core length. The hysteretic area
was compared with the circumscribing rectangle defined
by the maximum positive and negative forces and displacements
(“ideal” area). The normalized energy dissipated per
cycle is plotted against maximum axial strain amplitude
for each cycle in Figure 11. This figure demonstrates that
once significant yielding had occurred, the area inside the
hysteresis loop was typically around 80% of the rectangular
area. This is much higher than the same ratio for concentric
braces, such as single angle braces, which exhibit buckling
and elongation (Carden et al., 2006c). Unlike the situation
for the single angles, the effective energy dissipation by the
buckling-restrained braces increased as the displacement
amplitude increased. Hence these braces were much more
efficient than the single angles for energy dissipation, imply-
Table 3.Yield and Maximum Tension and Compression Forces in the Unbonded Braces
Brace Loading History
Expected
Yield
Force,
kips
Measured
Yield
Force,
kips
Maximum
Tension
Force,
kips
Maximum
Compression
Force,
kips
Overstrength
Factor,
􀁚
Compression
Strength
Adjustment
Factor, 􀁚
A Modified ATC-24 21.6 29.3 􀁚36.9 1.44 1.26
B BRBF Normal Static 21.1 27.7 􀁚30.6 1.36 1.11
C BRBF Reversed Static 22.7 27.4 􀁚31.0 1.35 1.13
D
BRBF Reversed
Dynamic
22.8 33.0 􀁚32.7 1.62 0.99
E
BRBF Reversed
Dynamica 21.7 31.5 􀁚37.6 1.55 1.19
F Constant Staticb 21.7 28.0 􀁚30.3 1.37 1.08
G Constant Staticc
20.4
22.0 30.0 􀁚34.6 1.47 1.15
aLoading history applied in the load frame after the braces removed from the south end of the bridge model following shake table experiments.
bLoading history applied in the load frame after the braces removed from the north end of the bridge model following shake table experiments.
cLoading history applied in the load frame after the braces removed from the mid-span of the bridge model following shake table experiments.

Fig. 11. Energy dissipated per cycle as a ratio

of “ideal” for Braces A, B, C, and D.
ENGINEERING JOURNAL / SECOND QUARTER / 2006 / 137
ing that there should be smaller drifts in ductile end cross
frames that use buckling-restrained braces compared with
those that use single angles for the same level of forces, as
was observed by Carden et al. (2006c).
The hysteretic energy reported in Figure 11 is based on
the displacements across the core length. It might be expected
that slippage in the connections would reduce the effective
energy dissipated per cycle. To quantify this effect, the
hysteretic energy dissipated per cycle was calculated based
on the total displacements in the brace including the connections
for a hysteretic area based on the total brace displacements
including the connections for Brace A, the brace
exhibiting the most slippage with the pinned-end connections.
It was found that the measured energy dissipation as
a ratio of circumscribing area was reduced by around 5%
at any given level of strain, thus connection slippage had a
relatively small effect on the efficiency of the brace. Consequently,
in terms of energy dissipation there is no apparent
advantage in providing slip-critical connections in unbonded
braces for seismic loading.
Strain Gage Readings
Strain gages were placed on Braces A and B, with gages
on either side of the core plate near each end of the core
length. A typical strain gage measurement in Brace B is plotted
against the axial force in Figure 12. This figure shows
that there are both tensile and compressive plastic strains in
the brace, although there is some bias toward tensile strains.
This can be explained by the higher mode buckling that is
known to occur in the core section of buckling-restrained
braces. Consequently, the maximum measured strain was
higher, at 4.83%, compared with the maximum average
computed strain from the displacement across the core of
the brace of 1.86%. Brace A shows similar strains, indicating
that the pinned-end connection did not adversely affect the
localized strains.
While in the elastic range, the strain gages were used to
estimate the forces in the unbonded brace using the elastic
modulus, assumed to be equal to 29,000 ksi, and the area
of the core plate. Good correlation was shown between the
measured and computed axial forces, where the difference
between the two was less than 8%, demonstrating that the
gages could be used to monitor elastic force levels in the
braces after installation in a bridge.
ANALYTICAL MODEL
Black et al. (2002) showed that the hysteretic properties of a
buckling-restrained brace can be captured well with a Bouc-
Wen model (Wen, 1976). However, for design using such
methods as a capacity spectrum analysis, it is more useful to
characterize the behavior using a bilinear model, although
some loss of accuracy is expected. A bilinear model was
shown to fit the experimental data reasonably well. The initial
stiffness, Ki, was calculated based on the cross-sectional
properties of the core and the core length, giving an initial
stiffness of 889 kip/in. The post-yield stiffness ratio, , was
assumed to be equal to 0.025 in order to fit the backbone
curve for the increasing amplitude cyclic displacements.
The yield force was assumed to be equal to the expected
yield of 20.4 kips for the static cases, and was amplified by
15% to 23.5 kips for the dynamically loaded braces.
The analytical model was subjected to the same displacement
time history as the axial cyclic experiments. The resulting
hysteresis loop for Brace B, as an example, is shown
in Figure 13. The bilinear model is shown to fit the cyclic
backbone curve for the experimental data well, although it
was conservative for the response in compression. Without
a separate model for the tension and compression properties,
Fig. 12. Typical strain gage measurement for Brace B.
Fig. 13. Analytical hysteresis loop compared with backbone
curve and experimental data for Brace B.
138 / ENGINEERING JOURNAL / SECOND QUARTER / 2006
this is unavoidable. The dynamic strain rate effect is evident
in Brace D. However, increasing the yield force by 15% resulted
in a reasonable comparison between the experimental
and analytical results. This was the basis for the inelastic
properties of the analytical bridge model used to calculate
cumulative displacement responses described in the earlier
section.
While the bilinear model captures the elastic loading and
unloading and cyclic backbone curve of the unbonded braces
well, it loses some accuracy in modeling the initial reversal
and post-yield stiffness of an individual cycle of loading
in the braces, as observed in Figure 13. The main effect of
this is a difference in the area inside the hysteresis loops. A
comparison of the hysteretic energy dissipated per cycle for
the experimental data and the bilinear model shows that the
hysteretic areas are similar at moderate strains, but larger
differences are observed at higher amplitudes. At strains of
around 2%, the bilinear model typically underestimates the
energy dissipation by around 20%, making the bilinear analytical
response conservative.
CONCLUSION
The cyclic inelastic behavior of buckling-restrained braces
was studied for potential implementation in ductile end cross
frames at the supports of steel plate girder bridges. The unbonded
braces exhibited excellent hysteretic behavior with
similar properties in tension and compression. Slippage in
the connections was observed in some of the braces, although
this did not have a significant effect on the energy dissipation
characteristics of the braces. The braces were subjected
to different loading histories with displacements up to an
equivalent axial strain in the core of the brace of 3.7%. Various
loading histories show that a history with large amplitude
cycles at the beginning followed by smaller cycles, like
a typical earthquake loading history, tends to result in lower
cumulative plastic strain capacity than an increasing amplitude
history. Dynamic loading also lowers the cumulative
plastic strain capacity compared with static loading, which
is attributed to increased forces during dynamic loading and
also the dynamic effects of slippage.
The yield tension force of the unbonded braces is within
12% of that predicted. The maximum forces due to strain and
cyclic hardening at a design strain of 2% are typically 35%
greater than the expected yield strength. Dynamic loading
increases these forces by typically 15% and up to 30% for
the first cycle of loading. The maximum compression force is
generally 10 to 15% greater than the tension force. The area
of the hysteresis loops for the unbonded braces is around
80% of the circumscribing rectangular area and increases as
strains increase unlike the degradation seen in typical concentric
braces after buckling. The response of the braces can
be approximately modeled with a bilinear representation to
capture the maximum forces, displacements and, more conservatively,
the energy dissipation in the braces.
From this study it is recommended that a maximum brace
strain of 2.0% be used for design of the buckling-restrained
braces when used in ductile end cross frames. Results from
an analytical study of a bridge model showed for 10 earthquakes
that the mean plus one standard deviation cumulative
plastic strain demand, for maximum displacements of
1.5 times the design displacement, was equal to the lowest
cumulative plastic strain capacity measured during experiments.
ACKNOWLEDGMENTS
This project was funded by the Federal Highway Administration
through the Highway Project at the Multidisciplinary
Center for Earthquake Engineering, University of Buffalo
under contract DTFH61-98-C-00094. Additional funding
was provided by the California Department of Transportation
under contract 59Y564. Grateful acknowledgment is made
of both agencies for their generous support. The donation
of the unbonded braces by Nippon Steel Corporation and
assistance of Dr. Ian Aiken in facilitating these experiments
is also greatly appreciated.
REFERENCES
American Association of State Highway and Transportation
Officials (AASHTO) (1999), Guide Specifications for
Seismic Isolation Design, Second Edition including 2000
Interim Revisions, AASHTO, Washington, DC.
AISC (2001), Load Factor and Resistance Design Manual
of Steel Construction, 3rd Edition, American Institute of
Steel Construction, Inc., Chicago, IL.
AISC (2002), Seismic Provisions for Structural Steel Buildings,
American Institute of Steel Construction, Inc., Chicago,
IL.
ATC (1992), ATC 24–Guidelines for Cyclic Seismic Testing
of Components of Steel Structures, Applied Technology
Council, Redwood City, CA.
Astaneh-Asl, A., Bolt, B., McMullin, K.M., Donikian, R.R.,
Modjtahedi, D., and Cho, S. (1994), Seismic Performance
of Steel Bridges During the 1994 Northridge Earthquake,
UCB/CE-STEEL-94/01, Department of Civil Engineering,
University of California at Berkeley, Berkeley, CA.
Black, C., Makris, N., and Aiken, I. (2002), Component Testing,
Stability Analysis and Characterization of Buckling
Restrained Unbonded Braces,TM Technical Report: PEER
2002/08, Berkeley, CA.
Bruneau, M., Wilson, J.W., and Tremblay, R. (1996), “Performance
of Steel Bridges during the 1995 Hyogoken-
Nanbu (Kobe, Japan) Earthquake,” Canadian Journal of
Civil Engineering, Vol. 23, No. 3, pp. 678–713.
ENGINEERING JOURNAL / SECOND QUARTER / 2006 / 139
Carden, L.P., Itani, A.M., and Buckle, I.G. (2006a), “Seismic
Performance of Steel Girder Bridges with Ductile Cross
Frames Using Single Angle X Braces,” ASCE, Journal of
Structural Engineering, Vol. 132, No. 3, March 1.
Carden, L.P., Itani, A.M., and Buckle, I.G. (2006b), “Seismic
Performance of Steel Girder Bridges with Ductile Cross
Frames using Buckling-Restrained Braces,” ASCE, Journal
of Structural Engineering, Vol. 132, No. 3, March 1.
Carden, L.P., Itani, A.M., and Buckle, I.G. (2006c), “Cyclic
Behavior of Single Angles for Ductile End Cross Frames
in Steel Girder Bridges,” AISC, Engineering Journal, Vol.
43, No. 1, 4th Quarter.
Clark, P.W., Aiken, I.D., Tajirian, F.F., Kasai, K., Ko, E., and
Kimura, I. (1999), “Design Procedures for Buildings Incorporating
Hysteretic Damping Devices,” International
Post-SMiRT Conference Seminar on Seismic Isolation,
Passive Energy Dissipation and Active Control of Vibrations
of Structures, Cheju, Korea, August 23–25.
Kim, J. and Choi, H. (2004), “Behavior and Design of Structures
with Buckling-Restrained Braces,” Engineering
Structures, Vol. 26, pp. 696–706.
Lopez, W.A., Gwie, D.S., Lauck, T.W., and Saunders, C.M.
(2004), “Structural Design and Experimental Validation
of a Buckling-Restrained Braced Frame System,” Engineering
Journal, Vol. 41, No. 4, pp. 177–186.
Nippon (2002), Seismic Behavior of Typical Steel (Unbonded
Brace)—Tensile Test Report for Steel Material of LYP-
225, Report: Nippon Steel Corporation, Japan.
Sabelli, R. (2001), Research on Improving the Design and
Analysis of Earthquake-Resistant Steel-Braced Frames,
Technical Report: PF2000-9–FEMA/EERI 2000 Professional
Fellowship, Oakland, CA.
Sabelli, R. (2004), “Recommended Provisions for Buckling-
Restrained Braced Frames,” Engineering Journal, Vol. 41,
No. 4, pp. 155–175.
SEAONC (2003), Recommended Provisions for Buckling-
Restrained Braced Frames, Structural Engineering Association
of Northern California, Sacramento, CA, July.
Shinozuka, M.(ed.), Ballantyne, D., Borcherdt, R., Buckle,
I., O’Rourke, T., and Schiff, A. (1995), The Hanshin-Awaji
Earthquake of January 17, 1995 Performance of Lifelines,
Technical Report: NCEER-95-0015, National Centre for
Earthquake Engineering Research, Buffalo, NY.
Wada, A., Saeki, E., Takeuchi, T., and Watanabe, A. (1989),
“Development of Unbonded Brace,” Column (A Nippon
Steel Publication), No 115, p. 12.
Wen, Y.-K. (1976), “Method for Random Vibration of Hysteretic
Systems,” Journal of Engineering Mechanics, Vol.
102(EM2), pp. 249–263.
Zahrai, S.M., and Bruneau, M. (1998), “Impact of Diaphragms
on Seismic Response of Straight Slab-on-Girder
Steel Bridges,” Journal of Structural Engineering, Vol.
124, No. 8, pp. 938–947.
Zahrai, S.M. and Bruneau, M. (1999), “Cyclic Testing of
Ductile End-Diaphragms for Slab-on-Girder Steel Bridges,”
Journal of Structural Engineering, Vol. 125, No. 9,
pp. 987–996.