EARTHQUAKE ENGINEERING & STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. (2012)

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eqe.2202

A model to simulate special concentrically braced frames beyond

brace fracture

Po-Chien Hsiao*,†, Dawn E. Lehman and Charles W. Roeder

Department of Civil Engineering, University of Washington, Seattle, WA 98195-2700, USA

SUMMARY
Special concentrically braced frames (SCBFs) are commonly used as the lateral-load resisting system in
buildings. SCBFs primarily sustain large deformation demands through inelastic action in the brace, including
compression buckling and tension yielding; secondary yielding may occur in the gusset plate and framing
elements. The preferred failure mode is brace fracture. Yielding, buckling, and fracture behavior results in
highly nonlinear behavior and accurate analytical modeling of these frames is required. Prior research has
shown that continuum models are capable of this level of simulation. However, those models are not suitable
for structural engineering practice. To enable the use of accurate yet practical nonlinear models, a research study
was undertaken to investigate modeling parameters for line-element models, which is a more practical modeling
approach. This portion of the study focused on methods to predict brace fracture. A fracture modeling approach
simulated the nonlinear, cyclic response of SCBFs by correlating onset of fracture to the maximum strain range
in the brace. The model accounts for important brace design parameters including slenderness, compactness,
and yield strength. Fracture data from over 40 tests was used to calibrate the model and included single-brace
component, single story frame, and full-scale multistory frame specimens. The proposed fracture model is more
accurate and simpler than other, previously proposed models. As a result, the proposed model is an ideal
candidate for practical performance simulation of SCBFs. Copyright © 2012 John Wiley & Sons, Ltd.

Received 21 August 2011; Revised 10 April 2012; Accepted 12 April 2012

KEY WORDS: braced frames; brace fracture; nonlinear analysis; performance-based engineering
seismic performance

1. INTRODUCTION

Special concentrically braced frames (SCBFs) are a commonly used seismic-resistant structural
system. SCBFs meet a wide range of performance levels. Braced frames can economically provide
lateral stiffness and strength during frequent and moderate events to control drift. In large
earthquakes, the bracing members may be subjected to large axial deformations resulting in severe
overall and local buckling deformations, and tensile yielding. This inelastic response permits SCBFs
to sustain large, cyclic drift demands. Eventually, failure of one of the components will occur.
Experimental research indicates that the local buckling deformations, typically at the center of the
brace, eventually lead to tearing of the brace. Although fracture is never desirable in a structure
system, a system will fail if the demand is large enough. To maintain gravity-load carrying capacity,
it is necessary to define a fracture mode that will not compromise the structure integrity. For braced
frames, loss of a column or connection would not meet this criterion. Therefore, brace fracture is the
preferred failure mode with other secondary failure modes suppressed, such as connection failure [1].

*Correspondence to: Po-Chien Hsiao, Department of Civil Engineering, University of Washington, Seattle, WA
98195-2700, USA.
†
E-mail: pchsiao@uw.edu

Copyright © 2012 John Wiley & Sons, Ltd.

 P.-C. HSIAO, D. E. LEHMAN AND C. W. ROEDER

The seismic response of steel bracing members dominates the seismic behavior of SCBFs. However,
the framing elements and the gusset plates are also important. They contribute to the nonlinear
response, and provide boundary conditions to the steel brace. Therefore, all members and their
connections must be accurately modeled. The authors have investigated and validated an accurate
method for the nonlinear analysis using line-element models [2]. The modeling approach is illustrated
in Figure 1. The example shows a single story specimen that was tested in the laboratory (Figure 1(a)).
The line-element modeling approach models all of the components, including the braces, beams,
columns, and connections, as indicated in Figure 1(b). Fiber-type beam–column elements are used to
model the beams, columns, and braces. Rigid links are used to simulate the rigidity of the gusset,
gusset-to-beam, and gusset-to-column connections. Finally, springs are used to simulate flexibility of
the gusset plates and shear plate connections, as shown in Figure 1.
Previous studies using a similar model [3, 4] have verified the benefits of using the discrete brace model
and modeling rigid end zones at the connections on the simulation of SCBFs, and mentioned the concept
of using rotational spring at the ends of the brace element. Typical modeling approaches for SCBFs use
either fully restrained or fully pinned models for the gusset plate connections. However, test results
show that the gusset plate connection is neither pinned nor fixed and its flexibility must be modeled
explicitly to capture the nonlinear response. A new approach was developed to simulate the behavior of
the gusset plate connections, as shown in Figure 1(c). The primary response is simulated using a
nonlinear, zero-length rotational spring, which has a spring stiffness based upon the gusset plate
material and geometry and located at the ends of the brace. This modeling approach simulates the
out-of-plane rotational behavior of the gusset plate connections and provided correct boundary
conditions of the brace members. Rigid end zones were used for the remainder of the connections
beyond the brace.
This modeling approach was verified using the experimental results, and a typical comparison is
shown in Figure 1(d). Relative to conventional rigid-ended or pin-ended nonlinear modeling
approaches that are commonly used for SCBF seismic analysis, the modeling approach illustrated in
Figure 1 provides greatly improved accuracy [2, 5]. The improved model successfully predicted the

Figure 1. Typical single-story (a) test specimen, (b) discrete line-element nonlinear model, (c) details of the
gusset plate modeling, and (d) comparison of measured and simulated results.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 A MODEL TO SIMULATE SCBFS BEYOND BRACE FRACTURE

buckling capacity, tensile yielding, distribution of yielding, and post-buckling behavior of the braces in
the structural systems [2].
Although the concept of performance-based seismic design is widely accepted, capacity evaluation
of SCBF structures including fracture and collapse is difficult to conduct in a practical manner. In
particular, it is difficult to predict brace fracture, which is the preferred failure mode of SCBFs, and
therefore its prediction is central to accurate performance-based seismic design of SCBFs. During a
test, the severe out-of-plane deformation because of the buckling behavior resulted in brace fracture.
This behavior is illustrated in Figure 2. The out-of-plane displacement of the brace increases with
the story drift of the frame, which results in a concentration of inelastic deformation at the midspan
of the brace and eventual formation of a plastic hinge, as shown in Figures 2(a) to (c). At the
location of the formed plastic hinge, the cupping behavior (Figure 2(d)) of the brace tube occurred
first, and then the crack initiation (Figure 2(e)) occurred at the cupped location after a few additional
cycles. The crack grew as the brace sustained tensile force again in the following cycles and
eventually fully fractures (Figure 2(f)).
This study was undertaken to develop a practical yet accurate model to predict the onset and impact
of brace fracture, and in turn, evaluate the impact of brace fracture on the system performance. The
model was developed for implementation in a nonlinear structural analysis program using the
modeling approach described previously.
The brace fracture model was based on past experimental studies. The test specimens varied widely
in their slenderness ratio, width–thickness ratio, and yield strength of the bracing members, and these
parameters were specifically included in the fracture model. Although different structural steel sections
may be used as bracing members (e.g., wide flanges, angles, channels, and tube sections), this study
focuses on square hollow structural sections (HSS) because they are economical and widely used in
practice. The resulting model was compared with other previous models

Figure 2. The global and local behavior causing brace fracture.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 P.-C. HSIAO, D. E. LEHMAN AND C. W. ROEDER

2. PRIOR BRACE FRACTURE SIMULATION MODELS

With proper design, brace fracture is the preferred failure mode of SCBFs [5]. Prior research has been
conducted to improve the understanding of brace fracture [6–19]. Several different approaches had
been developed to predict the fracture life of the braces under cyclic loading. In the case of
continuum analysis, the equivalent plastic strain has been calibrated to predict brace fracture, for
example, Yoo et al. [20] and Fell et al. [21]. In this study, the focus is on modeling approaches to
simulate brace fracture that are appropriate for use in nonlinear analyses using line-elements. Several
research studies have developed these types of models and a brief discussion of selected models
follows. Models that link brace fracture to the global response include those by Lee and Goel [18],
Shaback and Brown [16] and Tremblay et al. [19]. Uriz et al. [3] linked the local deformation
response of the brace to predict the onset fracture.
Lee and Goel tested 13 full-scale hollow and concreted-filled square tubular bracing members with
different effective slenderness and cross-sectional slenderness ratios [18]. Six specimens tested as part
of that study and three specimens from a prior study [22] that utilized HSS tubes were combined to
develop a model to predict the brace fracture life. The researchers postulated the cyclic deformation
demands on the brace resulted in low-cycle fatigue fracture. To simulate the observed response, the
cyclic axial deformation response of the HSS tube was used as the primary parameter in the brace
fracture model. Figure 3(a) shows a typical cyclic normalized axial force–deformation response
from a single cycle from a brace test. Two normalized axial deformations are defined. The
compressive portion of the normalized axial deformation Δ1 is the displacement that ranges from
the maximum displacement in the compressive loading direction to the displacement corresponding
to a tensile force of Py/3, as shown in the figure. The normalized axial deformation Δ2 is the
remaining portion of the displacement range. These two deformations are used to quantify the
defined nondimensional parameter of the fracture life, Δf, using the defined weights indicated in
Equation (1), where n is the number of cycles achieved before the brace fracture. Equation (2)
provides the fracture limit, which was based on the experiments.
X
n
Δf ¼ ð0:1Δ1 þ Δ2 Þ (1)
j¼1
 1:2  
315=Fy 4b=d þ 1
Δf;pred:1 ¼ Cs:1 (2)
½ðb  2tÞ=t1:6 5

In Equation (2), the following variables are used: b, d, and t are the width, depth, and thickness of
the HSS tube, respectively; Fy is the yield strength of the brace, in MPa; Cs.1 is an experimentally

Figure 3. (a) Definition of displacement components d1 and d2 from (b) comparison of predicted and
experimental measured fracture life [18].

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 A MODEL TO SIMULATE SCBFS BEYOND BRACE FRACTURE

determined nondimensional constant. The fracture life was related to three primary variables: Fy, b/d,
and (b–2t)/t, or yield strength, cross-sectional aspect ratio, and cross-sectional slenderness. The
experimental data set had effective slenderness ratios, KL/r, between 35 to 77 and cross-sectional
slenderness ratios, (b–2t)/t, between 14 to 30, where L and r are the length and the ratio of gyration
of the tube, and K is defined by Lee and Goel’s study between 0.50 and 0.85; all of the experiments
used ASTM-A500, Grade B structural tubes. According to the test results, the corresponding Cs.1
values varied from 1160 to 1508, and a mean value of 1335 was suggested. Figure 3(b) shows the
comparison between the fracture life measured during the tests, Δf,test, and the fracture life predicted
using Equation (2), Δf,pred.1.
Shaback and Brown also proposed an empirical equation to predict the fracture life of an HSS brace.
The method by Lee and Goel was adopted (Equation (1),) and different limits were developed. Instead
of Equation (2), a pair of equations was proposed to estimate the fracture life (Equations (3a) and (3b)),
differentiated on the slenderness ratio; all other variables were the same.

 3:5  
350=Fy 4b=d  0:5 0:55 KL
Δf;pred:2 ¼ Cs:2 ð70Þ2 for < 70 (3a)
½ðb  2t Þ=t 1:2 5 r

 3:5    
350=Fy 4b=d  0:5 0:55 KL 2 KL
Δf;pred:2 ¼ Cs:2 for ⩾70 (3b)
½ðb  2t Þ=t 1:2 5 r r

In the expressions, Fy is the yield strength of the brace, in MPa; Cs.2 is an experimentally determined
nondimensional constant, with a value of 0.065 suggested by the researchers. It is noted that this model
was based on different data, and the researchers used a combined data set with nine full-scale HSS-tube
braces tested by Shaback and Brown and brace tests conducted by Archambault [23]. The tests had
cross-sectional slenderness ratios, (b–2t)/t, between 12 and 18 and effective slenderness ratios, KL/r,
between 52 and 66, where K is determined by Shaback and Brown’s study to be between 0.66 and
0.77. Figure 4 shows the comparison between the fracture life measured during the tests, Δf,test, and
the fracture life predicted using Equations (3a) and (3b), Δf,pred.2.
Tremblay et al. used the brace end rotation to estimate the fracture of rectangular hollow sections
[19]. The researchers conducted a total of 24 cyclic tests, including full-scale X-bracing and single
diagonal bracing systems in 2003. A combined data set was developed and included 22 tests with
brace fracture from their test program and 24 tests from other studies on single bracing members

Figure 4. Comparison of predicted and experimental measured fracture life [16].

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 P.-C. HSIAO, D. E. LEHMAN AND C. W. ROEDER

[16, 18, 22, 24, 25]. The resulting variation in the brace design parameters was as follows: the effective
slenderness ratios, KL/r, varied from 60 to 143 according to the study and the width–thickness ratios,
bo/t and do/t, varied from 8 to 13 and 7 to 23, respectively, where bo and do are the flat width and depth
of the cross-section. The end rotation is related to the total ductility of the bracing member by
Equations (4) and (5), where LH is the brace length using the assumption that the ends are simply
supported as illustrated in Figure 5(a). The axial movement of the brace under compression, dc,
(shown in Figure 5(a)) is defined by Equation (5), where mc and mt are the peak compression and
tensile ductility values and dy is the axial yield deformation of the bracing member.
sffiffiffiffiffiffiffi
2dc
θf ¼ (4)
LH

dc ¼ ðmc þ mt  1Þdy (5)

The maximum end rotations corresponding to brace fracture, θf,pred, was derived from the database
of cyclic tests and are given by Equation (6).

 0:1  0:3
bo do KL
θf;pred ¼ 0:091 (6)
t t r

Figure 5(b) shows the comparison of the experimental and their predicted end rotations
corresponding to the onset of brace fracture.
Uriz [14] developed a prediction expression based on low-cycle fatigue, which used an accumulative
strain limit. The model was implemented in the OpenSees structural analysis software platform (Frank
Mckenna, Berkeley, CA). The low cycle fatigue modeling approach was based on Miner’s rule
[26, 27], which is expressed as the damage index, DI, give in Equation (7). In the equation, the damage
for each amplitude of cycling is estimated by dividing the number of cycles at that amplitude, n(ei), by
the predicted number of constant amplitude cycles at that amplitude necessary to cause failure, Nf(ei).
When DI = 0, there is no damage. A value of DI = 1 indicates failure. Because the cyclic history is not
necessarily symmetric, the amplitude of each cycle of the strain history was determined using a
modified rainflow cycle counting method. The Coffin–Manson relationship [26, 27] expressed in
Equation (8), was used to determine the strain amplitude, in which Nf is the number of cycles at strain
amplitude ei resulting in failure, and the parameters of m and e0 are empirical constants, where e0 is the
failure strain for a single reversal, while m is known as the fatigue ductility exponent.

Figure 5. (a) Simplified deformed shapes for prediction of out-of-plane deformations for single bracing and
(b) experimental versus predicted values. [19].

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 A MODEL TO SIMULATE SCBFS BEYOND BRACE FRACTURE

n 
X 
nðei Þ
DI ¼ (7)
j¼1
Nf ðei Þ j

e i ¼ e0 ðN f Þm (8)

Uriz used experiments performed by Yang on full-scale bracing members to validate the modeling
approach and calibrate necessary parameters, for which m = 0.5 and e0 = 0.095 [14]. Figure 6 shows
the comparison between the predicted and measured maximum axial displacement before brace fracture.

3. EXPERIMENTAL DATA SET

Experimental data from prior studies were gathered to provide the basis for the proposed fracture
model [6–19]. The following criteria were used: (1) the test must use square HSS tubes for the
bracing members, (2) the test must exhibit brace fracture, and (3) sufficient information regarding
the geometry, material properties and test loading must be included to permit simulation. Prior
studies suggest that the fracture life of the HSS bracing members depends on the slenderness ratio
of the bracing members, the width-to-thickness ratio of the cross-section, the yield strength of the
steel and the imposed displacement history. Synthesis of the experimental studies indicates that
smaller width-to-thickness ratios [12, 15–19] and larger slenderness ratio [12, 15, 16, 19] increase
brace ductility. As much as possible, a dataset was developed that included a wide range of these
important design variables.
Table I provides the details of the 44 specimens included in the data set, including size, geometry, and
material. The tests include a range of configurations including: (1) component tests consisting of only the
brace and gusset plate connections [12, 13, 15–19], (2) single-story and multistory frame tests including
the beams, columns, braces, and gusset plate connections simulating a lateral frame configuration
[6–11, 14]. An example of a full-scale, single-story one-bay braced frame is shown in Figure 7(a); these
tests were conducted at the University of Washington [6–9]. Examples of full-scale two-story and
three-story frames are shown in Figures 7(b) and (c); these frames were tested at the National Center
for Research on Earthquake Engineering (NCREE). A different two-story frame test is shown in
Figure 7(d); this example frame was tested at University of California (UC) Berkeley [14]. Figure 7(e),
shows an example isolated brace (or component) test; this type of test was conducted at UC Davis [12],
UC Berkeley [13, 14], University of Dublin in the United Kingdom [15] and University of Calgary in

Figure 6. Comparison of predicted and experimental displacement range before brace fracture [14].

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 P.-C. HSIAO, D. E. LEHMAN AND C. W. ROEDER

Table I. Characteristics of the specimens and the corresponding maximum strain ranges.
Ag Max.
Study Specimens Brace shape Steel (mm2) w/t KL/r E/Fy erange, simul.

[6] HSS2a HSS553/8 A500, Gr. B 3987 11.3 77.6 414.3 0.0443
HSS3a HSS553/8 A500, Gr. B 3987 11.3 77.2 414.3 0.0493
HSS4a HSS553/8 A500, Gr. B 3987 11.3 77.7 396.2 0.0520
HSS5a HSS553/8 A500, Gr. B 3987 11.3 81.0 396.2 0.0605
[7] HSS6a HSS553/8 A500, Gr. B 3987 11.3 81.2 448.2 0.0467
HSS7a HSS553/8 A500, Gr. B 3987 11.3 66.1 448.2 0.0606
HSS8a HSS553/8 A500, Gr. B 3987 11.3 75.8 448.2 0.0512
HSS9a HSS553/8 A500, Gr. B 3987 11.3 77.1 448.2 0.0469
HSS10a HSS553/8 A500, Gr. B 3987 11.3 77.8 440.7 0.0534
[8] HSS11a HSS553/8 A500, Gr. B 3987 11.3 64.8 440.7 0.0438
HSS12a HSS553/8 A500, Gr. B 3987 11.3 70.1 440.7 0.0492
HSS13a HSS553/8 A500, Gr. B 3987 11.3 81.8 440.7 0.0480
HSS14a HSS553/8 A500, Gr. B 3987 11.3 81.0 440.7 0.0460
HSS15a HSS553/8 A500, Gr. B 3987 11.3 82.7 440.7 0.0426
[9] HSS17a HSS553/8 A500, Gr. B 3987 11.3 81.7 440.7 0.0486
HSS24a HSS553/8 A500, Gr. B 3987 11.3 81.0 448.2 0.0508
HSS25a HSS553/8 A500, Gr. B 3987 11.3 66.1 448.2 0.0535
[10] TCBF1-1 HSS1251259 A500, Gr. B 3967 11.3 56.4 450.0 0.0589
TCBF1-3 HSS1251259 A500, Gr. B 3967 11.3 69.6 448.8 0.0488
[11] TCBF2-1 HSS553/8 A500, Gr. B 3987 11.3 64.7 431.8 0.0512
[12] Kavinde-1 HSS441/4 A500, Gr. B 2174 14.2 60.2 630.4 0.0550
Kavinde-2a HSS441/4 A500, Gr. B 2174 14.2 60.2 630.4 0.0613
Kavinde-4 HSS443/8 A500, Gr. B 3084 8.5 67.4 630.4 0.0692
[13] Yang-4a HSS663/8 A500, Gr. B 5213 14.2 39.3 483.3 0.0607
Yang-5 HSS663/8 A500, Gr. B 5213 14.2 39.3 483.3 0.0647
[14] Patxi-SCBF-1 HSS663/8 A500, Gr. B 5213 14.2 45.8 478.5 0.0617
[15] Broderick-S1-40H 40402.5SHS S235JRH 375 13.1 35.8 741.2 0.0693
Broderick-S4-20H 20202.0SHS S235JEH 144 6.7 74.4 621.3 0.0577
[16] Shaback-1B RHS1271278.0 G40.21-350W 3620 12.9 53.9 453.7 0.0477
Shaback-2A RHS1521528.0 G40.21-350W 4430 16.0 53.3 457.0 0.0458
Shaback-2B RHS1521529.5 G40.21-350W 5210 13.5 52.4 443.4 0.0476
Shaback-3Ab RHS1271276.4 G40.21-350W 2960 16.7 64.8 425.2 0.0384
Shaback-3B RHS1271278.0 G40.21-350W 3620 13.6 65.8 453.7 0.0409
Shaback-3C RHS1271279.5 G40.21-350W 4240 10.5 61.6 438.2 0.0397
Shaback-4A RHS1521528.0 G40.21-350W 4430 15.8 63.5 457.0 0.0403
Shaback-4B RHS1521529.5 G40.21-350W 5210 13.6 59.7 436.7 0.0411
[17] Han-S77-28b HSS1001003.2 SPSR400 1239 28.3 76.9 497.5 0.0248
[18] Lee-1a RHS550.188 A500, Gr. B 2271 25.7 63.0 469.3 0.0320
Lee-2a RHS550.188 A500, Gr. B 2271 25.7 34.4 464.7 0.0345
Lee-4a RHS440.125 A500, Gr. B 1226 31.5 77.1 500.0 0.0452
Lee-5a RHS440.250 A500, Gr. B 2316 14.2 79.0 391.9 0.0555
Lee-6a RHS440.250 A500, Gr. B 2316 14.2 45.1 391.9 0.0698
Lee-7a RHS440.250 A500, Gr. B 2316 14.2 45.1 391.9 0.0593
[19] Tremblay-S3Ac RHS76764.8 G40.21-350W 1310 12.8 166.7 510.1 0.0490
a
Applied by asymmetrical loading history.
b
Width–thickness ratio does not satisfy the current AISC requirements.
c
Slenderness ratio does not satisfy the current AISC requirements.

Canada [16]. Figure 7(f) shows an example test of an inclined brace test; this type of test has been
conducted at Hanyang University in Korea [17], University of Michigan [18], and Ecole Polytechnique
in Canada [19].
The specimens were grouped and reference the name used in the experimental study in Table I. The
braces encompass a wide range of HSS sections, from SHS20202.0 (mm) to HSS663/8. These
sizes result in a wide range of width–thickness ratios, consistently defined as w/t = (b–3t)/t here
according to the definition in the American Institute of Steel Construction (AISC) manual of steel
construction [28], ranging from 7 to 28 for the test specimens included; where b and t are the width
and thickness of the steel tube. The effective slenderness ratios, KL/r, of the braces ranged from 34

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 A MODEL TO SIMULATE SCBFS BEYOND BRACE FRACTURE

Figure 7. Specimen configurations of the experimental studies (a) single-story braced frame, (b) multistory
x-braced frame, (c) three-story braced frame, (d) chevron braced frame, (e) brace tests, and (f) inclined brace test.

to 167. The effective length factor, K, of all specimens here was determined using the method proposed
by Jain et al. [29], in which b is obtained using the ratio of flexural stiffness of the bracing member and
gusset plates, bθ and bm using Equations (9a)–(9d).
 
EIg
bθ ¼ (9a)
lave gusset plate

 
EIb
bm ¼ (9b)
L bracing member

tanm b
¼ 2 m ð0:5p⩽m⩽pÞ (9c)
m bθ

p
K¼ (9d)
2m

In the expressions, E is modulus of elasticity, Ig and Ib are the moment inertia of the Whitmore
section for gusset plates and cross-section of the tube for the bracing members, respectively, lave is
average of the three buckling lengths, l1 to l3, as illustrated in Figure 8 modified from Thornton
method [30], and L is length of the brace tube. The variable m is between 0.5p and p. The measured
yield strength of the steel, Fy, is reported herein. Table II shows the mean and variations of the
design parameters computed using the 44 tests.
In addition to the geometric and material variations, the cyclic drift history can impact the fracture
life of the brace. A study was conducted to investigate the influence of the drift history by evaluating
seven parameters: (1) the maximum axial deformation range of the bracing member for the entire
history (also correlates to the maximum end rotation of the brace [19]); (2) the maximum axial
deformation range for a single cycle; (3) the maximum tensile axial deformation; (4) the maximum

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 P.-C. HSIAO, D. E. LEHMAN AND C. W. ROEDER

Figure 8. Gusset plate buckling illustration.

Table II. Variations of design parameters.

w/t KL/r E/Fy

Mean 13.7 67.6 468.3

Standard deviation 3.1 13.4 48.0
Standard deviation/mean 22.2% 19.8% 10.2%

compressive axial deformation; (5) the accumulated axial deformation counted using the counting
approach proposed by Lee and Goel [18], Equation (1); (6) the accumulated axial deformation
derived by summation of all deformations; and (7) the accumulated axial deformation counted using
the rain-flow counting method. Table III shows the results. The variables that were based solely on
drift had large variability, from 27% to 66%, which is larger than the variation in the design
parameters (w/t, KL/r, and E/Fy), shown in Table II. Therefore, using drift as the sole parameter was
deemed unreliable. Similarly, the axial deformation parameters were considered having a relatively
weak relationship with the brace fracture. This also illustrates possible limitations with several of the
past brace fracture prediction methods [16, 18, 19].

4. STRAIN-RANGE FRACTURE MODEL

As illustrated in Figure 2, fracture of an HSS brace results from the large strains and local deformation
demands at the middle of the brace during large cyclic demands in compression and tension. These
experimental observations were used as the basis for the proposed fracture model. Specifically, the
local strain demands at the midspan of the brace were used to predict the onset of brace fracture.
For each test specimen, the local strain demands corresponding to the drift (or other global
deformation measure) at the onset of fracture behavior was determined analytically. The analytical

Table III. Variations of several characteristics at drift level.

Max. axial def. range (%) Max. axial def. (%) Accumulated axial def. (%)

Entire history For a given Σ Δf

[19] cycle Tension Compression [18] Σ Δrainflow ΣΔ

Mean 2.3 2.0 1.3 1.0 51.2 21.9 41.9

Standard deviation 0.7 0.5 0.4 0.4 33.5 11.5 21.7
Standard deviation/mean 30.1% 27.4% 29.7% 44.3% 65.5% 52.4% 51.8%

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 A MODEL TO SIMULATE SCBFS BEYOND BRACE FRACTURE

model is an experimentally-validated nonlinear line-element modeling approach implemented in the

OPENSEES structural analysis software. Additional details of the model are provided elsewhere [2].
Nonlinear beam–column elements with fiber sections are used to simulate the braces, beams, and
columns. This modeling approach assumes a linear strain distribution across the section. Therefore,
the model can simulate plastic hinge behavior and the out-of-plane deflection at the midspan of the
brace length, which contributes to the onset of brace fracture. However, the line-element model does
not simulate local strain concentrations resulting from local deformations, such as cupping as
depicted in Figure 2(d). Therefore, the model cannot capture the full local brace strain. However, it
captures a significant portion of the local strain, and these computed strains were calibrated to brace
fracture. Each specimen was simulated through its entire deformation history. For each test, the
simulated strains at midspan of the brace were recorded until the deformation where brace fracture
occurred, and the simulated strain range was used as the indicator of brace fracture for that
specimen. The indicators for all test specimens were combined in a statistical analysis to develop the
proposed model, and the predicted deformations were compared with the actual deformations to
establish the accuracy of the model.
Prior to the investigation of the magnitude of the strains in individual fibers, the strain demand along
the entire brace length and the stability of the demand were investigated. Figure 9(a) shows
the simulated maximum strain ranges through the depth of the tubular cross-section at midspan of
the brace and at two integration points away from midspan of the brace. The results show that the
maximum strain demand always occurred at the plastic hinge associated with brace buckling at the
extreme fiber on the compression side of the hinge, as illustrated in Figure 9(a). Uriz’s study using
the similar brace model showed that the strain history of the brace fibers varied with the number of
the brace elements [14], therefore influence of the number of the brace segments on the maximum
strain range was investigated here. Figure 9(b) shows the relationship of the obtained maximum
strain range at the critical location and the number of elements used to model the brace. Using 16
elements along the brace length results in a tolerable error (less than 1.2% in the example shown
here), and this mesh size was used throughout the study. It should be noted that midspan buckling is
not guaranteed; the method shown in Figure 9 is streamlined for the similar boundary conditions at
each brace end. For the specimens with different gusset plate design at each end, such as Specimen
TCBF1-1 and TCBF2-1 in Table I, the maximum strain range at the developed plastic hinge has
been considered in the study.
The simulated strain history for Specimen HSS24 [9] is used as an example to illustrate typical brace
response and is shown in Figure 10(b). The plotted strain history indicates the strain on the tension and
compression sides of the plastic hinge at midspan of the brace. The compressive strains are larger than
the tensile strains, which agree with the test observation that the crack initiation occurred at the
compression side (see Figure 2(e)). Therefore, the maximum strain history computed at the midspan
brace for each test was used to evaluate the effectiveness of strain as a fracture indicator.
Four different demand parameters were developed and evaluated using the maximum simulated
strains as a basis, including: (1) maximum tensile strain, et,max; (2) maximum compression strain,

Figure 9. Maximum strain ranges as function of (a) location and (b) number of elements.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 P.-C. HSIAO, D. E. LEHMAN AND C. W. ROEDER

Figure 10. (a) Schematic of the midspan fiber section in brace model and (b) typical strain history of the
fibers at midspan of the brace (specimen HSS24 shown).

ec,max; (3) the maximum strain range for a given cycle; and (4) the maximum strain range for the entire
time history, where the maximum strain range is the summation of the absolute values of the maximum
tensile and the maximum compressive strains.
In addition to the maximum values, accumulated strain was also evaluated. Two expressions were
developed. The first calculated the accumulated strain using the rain-flow counting method, given by
Equations (7) and (8) with values for the parameters of m and e0 as determined by Uriz [14]. A
second accumulate strain model used summation of strain from every cycle with three exponential
powers of 1, 2, and 3 applied the strains before summation.
For each expression, the strain or damage index values were computed at the measured drift
corresponding to fracture. Table IV presents the mean and coefficient of variation values for each
expression for the 44 tests. Using the maximum strain range through the entire history has the
lowest coefficient of variation of 15.7% of all of the methods. The maximum strain and accumulated
strain have variations of more than 29% and 34%, respectively. In terms of the accumulated damage
approach using rainflow-counting method, the mean value of the damage index throughout the tests
was close to 1.0; however, the results had a large coefficient of variation of over 50%. Therefore,
the maximum strain range through the entire history was considered as the best and most reliable
variable for the prediction of brace fracture.
The simulated maximum strain range, Max. erange, simul, from the strain analysis described above is
listed for each specimen in Table I. For specimens with multiple braces, the maximum strain range of
the first brace to fracture was reported. A regression analysis of the data was developed for predicting
the limiting maximum strain range, Max. erange, pred., given by Equation (10). As discussed previously,
brace fracture depends on the width–thickness ratio of the cross-section (w/t), the slenderness ratio of
the bracing member (KL/r) and the ratio of the elastic modulus to the yield strength of the steel
(E/Fy). Figure 11 shows the comparison between the resulting simulated and calculated values of the
maximum strain ranges limit.

Table IV. Fracture models based on fiber strains.

Max. strain range Max. strain Accumulated strain

Entire For a
history given cycle Tension Compression Σerainflow ΣDIrainflow Σe Σe2 Σe3

Mean 0.050 0.041 0.019 0.031 0.467 1.052 0.873 0.0292 0.00115
Standard deviation 0.008 0.010 0.008 0.009 0.158 0.532 0.349 0.0121 0.00053
Standard deviation/mean 15.7% 24.4% 41.4% 29.4% 33.9% 50.6% 40.0% 41.4% 46.4%

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 A MODEL TO SIMULATE SCBFS BEYOND BRACE FRACTURE

Figure 11. (a) Calculated and numerical maximum strain ranges for all specimens; (b) relationship of the
experimental axial deformation ratios and KL/r; and (c) relationship of the maximum strain range and KL/r.

w0:4 KL0:3  E 0:2

Max:erange;pred: ¼ 0:1435 (10)
t r Fy

This calibration of the model equation included more tests (total of 44) than that used in prior
fracture models. This regressive result has negative power for the ratio w/t, which agrees with the
trend in the previous models (i.e., Equations (2), (3), and (6)). In addition, it has a negative power
for the ratio KL/r. On the surface, this appears to be contrary to the previous models, Equations
(3) and (6). However, the maximum strain range on the brace does not directly correlate with the
global deformation of the brace, which is the parameter used for the other models. Figures 11
(b) and (c) show these opposite trends. Figure 11(b) shows experimental axial deformation rises as a
function of KL/r. The trend line has a positive slope. Figure 11(c) shows the simulated maximum
strain range versus KL/r. This relation is negative.
Moreover, in Equation (10), the positive power of ratio E/Fy implies that stronger steel would lead
to earlier fracture and be less ductile, which agree with Lee and Goel’s model (Equation (2)); whereas
in Equations (3a) and (3b), it is raised to a negative power, which shows stronger steel will delay
fracture and be more ductile. This difference is likely caused by the related small variation in the
E/Fy ratios in past tests (as shown in Table II), and the fact that the two programs used different
tests to calibrate their models.

5. CONSTITUTIVE MODEL FOR BRACE FRACTURE SIMULATION

The fracture expression was implemented in the OPENSEES framework. A new fracture material model
was integrated with the Steel02 constitutive material model available in OPENSEES, and operates
independently on each fiber in the cross-section. The fracture model monitors the maximum strain
range of the fibers and reduces the stress of the fibers at the limiting value estimated by the
Equation (10) to predict fracture, as illustrated in Figure 12(a). For strains beyond the fracture strain,
an elastic constitutive model with a very small effective modulus (0.001ksi) is used to ensure
post-fracture convergence and allow post-fracture simulation, as shown in the figure.
A further limitation was placed on the fracture model to ensure that the fracture behavior of the model is
only permitted until the stress in the model is tensile, because experimental observations indicate that
initial fracture always occurs only when the brace resists a tensile force. The fracture model was applied
to all fibers through the entire length and cross-section of the brace member; however, simulated
fracture would be triggered at the layer of fibers having the largest maximum strain range on the brace,
which always occur at the extreme fiber on the compression side of the plastic hinge of the brace as
mentioned previously. The model simulates fracture through the cross-section, because fracture of the
initial fiber would sequentially increase the strain range in the adjacent fibers until the full rupture of the
cross-section. Therefore, the proposed fracture material model is capable of simulating initiation of
tearing and fracture and conducting the analysis beyond brace fracture.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 P.-C. HSIAO, D. E. LEHMAN AND C. W. ROEDER

Figure 12. (a) Schematic of the fracture material model, (b) relationship of the maximum strain range versus
story drift range (for specimen HSS13), and (c) the analytical and measured responses (for specimen HSS 13).

Figure 12(b) presents the relationship of the maximum strain range at the critical location of initial
fracture versus the story drift range, where Specimen HSS13 [8] is used as an example. The model
permits analysis beyond brace frame as demonstrated in Figure 12(c) for Specimen HSS13, for
which testing was continued for cycles beyond complete brace fracture. System stability, stiffness,
and strength after brace fracture depends on secondary capacity provided by moment resistance
developed through the braced-frame gusset plate connections. The model predicted the brace
fracture at the correct story drift and accurately represented the post-fracture behavior.

6. COMPARISONS WITH THE EXISTING FRACTURE MODELS

The 44 specimens of the experimental database (Table I) were simulated with the new fracture model
and the proposed fracture models discussed previously. Specifically, the accuracy of four other existing
brace fracture models was assessed: the accumulated axial-deformation model developed by Lee and
Goel [18] and used by Shaback and Brown [16], the maximum end-rotational model developed by
Tremblay et al.[19] and the strain fatigue model developed by Uriz [14].
Figure 13(a) shows the resulting values using the proposed fracture model and the results calculated by
Lee and Goel’s and Shaback and Brown’s models. These models are based on accumulated axial
deformations, Equations (2) and (3). Because a comparison requires that all of the models are expressed
using the same demand parameter, the accumulated axial deformations corresponding to predicted
fracture using the proposed fracture model was also calculated using Equation (1). The simulated
accumulated axial deformation, Δf,simul., corresponding to observed braced fracture in the test is
compared. Figure 13(a) shows that the proposed model accurately predicted the fracture life of the
braces through a wide range of ductility, while the models using Equations (2) and (3) significantly
underestimated the fracture life for the ductile specimens where the accumulated axial-deformation was
larger than 40.
The proposed fracture model was also compared with the maximum end-rotation model developed
by Tremblay et al. [19] and expressed by Equation (6). Figure 13(b) shows the maximum end-rotations

Figure 13. Comparison of the proposed and (a) the accumulated axial-deformation, (b) the maximum
end-rotational, and (c) the strain fatigue models.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 A MODEL TO SIMULATE SCBFS BEYOND BRACE FRACTURE

predicted by the proposed fracture model and the Tremblay model in comparison with the
end-rotations corresponding to observed brace fracture for the experimental results, θf,test. The
maximum end-rotations for the proposed model and the test results were estimated using Equations
(4) and (5) and the measured ductility of the experiments. The results show that the proposed model
is more accurate. The maximum end-rotational type model [19] underestimated the accumulated
axial deformation at brace fracture for the specimens for which the end-rotation exceeded 0.25 radians.
The proposed model was also compared with the strain fatigue model developed by Uriz [14]. It
should be noted that the physical test data of the tests from other researchers’ studies [12–19] were
unavailable. Comparison of the two models based on local strain deformation using energy
dissipated or the accumulated DI was not possible. Therefore, the models here compared using axial
brace deformation, which directly correlates to the experimental drift capacity. The predicted axial
deformation range was compared with the measured axial deformation range, as shown in Figure 13
(c). In the figure, Δrange,pred. is the maximum axial deformation range predicted by the models, and
Δrange,test is that from the test results. The Uriz model results were calculated using an available
fatigue material model in OPENSEES framework developed by the researcher, using parameters
values of 0.5 and 0.091 for m and e0, respectively [14]. The comparison shows that the proposed
model provides an improved prediction of brace fracture. The fatigue model had larger variation in
the deformation at fracture than the proposed model.
Figures 14(a), (b), and (c) show the accuracy of the proposed model as a function of the slenderness
ratio, width–thickness ratio, and yield strength of the brace, respectively. Because some of the
experimental specimens do not fall within the design specification, the limiting values of the
American Institute of Steel Construction requirements are also shown. For each point, the ordinate is
the ratio of the simulated maximum story drifts range prior the brace fracture relative to the
experimental values and the abscissa is the value of the design parameter for a test specimen.
Figure 14(b) shows the model accuracy relative to the w/t ratio and indicates that the accuracy is
reduced for width-to-thickness ratios exceeding 25. However, many of these specimens had
properties that do not meet the AISC seismic design requirements [31]. Considering only the
specimens that satisfy the AISC requirements, the proposed fracture in the study had highly accurate
prediction of brace fractures, while the other three type models in general underestimated the onset
of brace fracture. It should be noted that the study verified that the proposed fracture model based
on the maximum strain range of the brace has the best correlation to the rupture of braces because
of fatigue; however, the model might not be appropriate for the cases of lower strains, where
fracture can occur.

Figure 14. Accuracy of proposed as a function of (a) slenderness ratios, (b) width–thickness ratios, and (c)
yield strength.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 P.-C. HSIAO, D. E. LEHMAN AND C. W. ROEDER

7. IMPACT OF DESIGN PARAMETERS OF PREDICTED FRACTURE

To investigate the impact of the design parameters, including width–thickness and brace slenderness
ratios, w/t and KL/r, on the maximum strain range and fracture potential, a series of single-story,
single-bay braced frames were analyzed with various lengths of the brace and thicknesses of the
brace tubes as described below. Figure 15(a) shows an illustration of the frame model. A single
diagonal brace with 45 degree angle was consistently used. The analysis focused on the brace
response, thus the beam and column members were simulated as truss elements.
The brace was modeled using the modeling approach described previously. Identical gusset plate
design used in Specimen HSS10 using a 12-mm-thick tapered gusset plate was used for all models
here. The connection had a rotational stiffness of 224 kN-m/rad. and yield moment of 16.8 kN-m.
The analysis started from the story height, H, of 2.54 m with HSS553/8 brace.
Two groups of frames were investigated. One group had variation in w/t ratio by changing the tube
thickness, t (7.9 and 6.4 mm). This changed w/t and KL/r was nearly constant. The other increased the
brace length by changing the story height and span, H, to 3.81 and 5.08 m. This change varied the KL/r
ratios, because the effective length coefficient K was taken as 1.0 for all cases. It should be noted that
the radius of gyration, r, was approximately equal to 0.4b of the rectangular HSS cross-sections.
Varying the brace width would simultaneously change both w/t and KL/r. Therefore, the brace
width, b, remained constant in the analysis to isolate the impact of the two design parameters
studied here.
The analysis was performed using OPENSEES program without fracture modeling, and a single
symmetric cycle with various story drift ratios, SDR, of 1%, 2%, and 3%, was applied to each frame
in sequence. The maximum strain ranges at the extreme fiber on the compression side of the plastic
hinge at midspan were compared. Figures 15(b) and (c) show the results of the two frame groups. In
addition, the estimated limiting strain-range values, calculated by Equation (10), are also shown in
the figures. It is apparent that varying the brace thickness, t, (or w/t ratio) did not have significant
impact on the computed maximum strain range, while increasing the brace length, L, (or KL/r ratio)
significantly decreased the computed maximum strain range at each story drift level. As noted by
Figure 15(c), the proposed fracture model predicts more drift capacity for frames with smaller w/t
and larger KL/r ratios, which agreed with the trends observed in the past experiments and the prior
fracture models.

Figure 15. (a) The illustration of the single-story frames and the applied loading protocol, (b) the variation of
max. strain range with w/t ratios, and (c) KL/r ratios.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 A MODEL TO SIMULATE SCBFS BEYOND BRACE FRACTURE

8. SUMMARY AND CONCLUSIONS

Brace fracture is the preferred failure mode of SCBF systems. Therefore, a full nonlinear analysis of an
SCBF building must include an accurate fracture model. Development of this type of model for
implementation in a nonlinear line-element modeling environment was the primary objective of this
study. The model was developed to simulate the behavior of a brace frame prior to and beyond
fracture as observed in laboratory testing. The analysis is based upon the large local strains in the
middle of the brace resulting from out-of-plane movement of the brace and eventual tearing under
tensile loading. The brace is modeled using a beam column element with discretized fiber
cross-sections at the integration points. The maximum strain range in the extreme fiber was used to
predict the onset of brace fracture. The model was validated using 44 experimental research results,
which had a wide range of brace parameters and system configurations and resulted in brace
fracture. The fracture limit included the impact of the brace parameters influencing the onset of
fracture, including slenderness ratio, yield strength, and w/t ratio. Comparison with prior fracture
models indicates that the proposed model offers improved accuracy over a wide range of parameters
with sufficient simplicity to permit implementation in practical nonlinear structural analysis. The
fracture model was implemented in the OPENSEES framework to permit simulation of SCBF
systems. A limited analytical study on a single-bay, single-story braced frame was conducted to
evaluate the relationship between the story drift and the maximum strain range values and the
impact of the design parameters on the maximum strain range.
The conclusions of the study are as follows:
• The maximum strain range was found to be the best indicator of brace fracture and was used to
estimate the fracture life of the braces. A regression equation that had three dependent variables
based on the geometric and material properties of the brace tubes was used to estimate the limiting
maximum strain range corresponding to brace fracture observed in past tests.
• Other existing fracture models, based on accumulated axial deformation, end-rotation of
the braces, and the accumulated strain deformation, were evaluated. In general, these models
underestimated the fracture life of the brace and were not as accurate as the proposed model.
• The improved model with the fracture material model provided accurate simulations of not only
buckling capacity, tensile yielding, and post-buckling behavior of the braced frames, but also
the fracture and post-fracture behavior of the frames. It is useful for both evaluating seismic
demand and capacity of SCBF structures in research and for performance based seismic design
in structural engineering practice.
• A new fracture material model was implemented in the OPENSEES framework based on the results
of the study. The maximum strain-range limit calculated by the proposed equation was used to
trigger the fracture behavior in the analysis. The proposed fracture model, based on maximum
strain range, enabled prediction of brace fracture over a great number of experimental specimens.
• The study of single-bay, single-story frames showed that the maximum strain range varied with
KL/r. With larger slenderness and smaller width–thickness ratios, these frames have larger drift
capacities, which lowers their tendency for premature brace fracture.

ACKNOWLEDGEMENTS
This research was funded by the National Science Foundation under grants CMS-0619161, NEESR-SG
International Hybrid Simulation of Tomorrow’s Braced Frames. Supplemental funding was provided by
AISC. Advice and guidance was provided by Stephen Mahin, Keh-Chyuan Tsai, Tom Schlafly (AISC),
Tim Fraser, Walterio Lopez (Rutherford and Chekene), Larry Muir (CIVES) and Rafael Sabelli (Walter
P. Moore). The advice and financial support of these individuals and institutions are greatly appreciated.

REFERENCES
1. Roeder CW, Lumpkin EJ, Lehman. Balanced Design Procedure for Special Concentrically Braced Frame Connections.
Journal of Constructional Steel Research 2011; 67(11):1760–72.
2. Hsiao P-C, Lehman DE, Roeder CW. Improved Analytical Model for Special Concentrically Braced Frames. Journal
of Constructional Steel Research 2012; 73:80–94.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
 P.-C. HSIAO, D. E. LEHMAN AND C. W. ROEDER

3. Uriz P, Filippou FC, Mahin SA. Model for Cyclic Inelastic Buckling of Steel Braces. Journal of Structural Engineering
(ASCE) 2008; 134(4):619–628.
4. Yang TY. Performance evaluation of innovative steel braced frames. Doctoral dissertation, Graduate Division of the
University of California, Berkeley, 2006.
5. Lehman DE, Roeder CW. Improved seismic design of concentrically braced frames and gusset plate connections.
ASCE, 2008.
6. Johnson S. Improved seismic performance of special concentrically braced frames. A thesis submitted in partial
fulfillment of the MSCE, University of Washington, Seattle, 2005.
7. Herman D. Further improvements on and understanding of SCBF systems. A thesis submitted in partial fulfillment of
the MSCE, University of Washington, Seattle, 2006.
8. Kotulka BA. Analysis for a design guide on gusset plates used in special concentrically braced frames. A thesis
submitted in partial fulfillment of the MSCE, University of Washington, Seattle, 2007.
9. Powell J. Evaluation of special concentrically braced frames for improved seismic performance and constructability.
A thesis submitted in partial fulfillment of the MSCE, University of Washington, Seattle, 2009.
10. Clark KA. Experimental Performance of Multi-Story X-Brace Systems. A thesis submitted in partial fulfillment of the
MSCE, University of Washington, Seattle, 2009.
11. Lumpkin EJ. Enhanced seismic performance of multi-Story special concentrically braced frames using a balanced
design. A thesis submitted in partial fulfillment of the MSCE, University of Washington, Seattle, 2009.
12. Fell BV, Kanvinde AM, Deierlein GG, Myers AM, Fu X. Buckling and fracture of concentric braces under inelastic
cyclic loading. SteelTIPS, technical information and product service, Structural Steel Educational Council, Moraga,
CA, 2006.
13. Yang F, Mahin SA. Limiting net section failure in slotted HSS braces. Structural Steel Education Council, Moraga,
CA, 2005.
14. Uriz P. Towards earthquake resistant design of concentrically braced steel structures. Ph.D. Dissertation, Dept. of
Engineering-Civil and Environmental Engineering, University of California, Berkeley, CA, 2005.
15. Goggins J M, Broderick BM, Elghazouli AY. Experimental behavior of hollow filled RHS bracing members under
earthquake loading. Composite Construction in Steel and Concrete V. ASCE, 2006.
16. Shaback B, Brown T. Behavior of square HSS braces with end connections under reserved cyclic axial loading. A
thesis submitted in partial fulfillment of the MSCE, Department of Civil Engineering, University of Calgary, Calgary,
Canada, 2001.
17. Han SW, Kim WT, Foutch DA. Seismic behavior of HSS bracing members according to width-thickness ratio under
symmetric cyclic loading. Journal of Structural Engineering (ASCE) 2007; 133(2):264–273.
18. Lee S, Goel SC. Seismic behavior of hollow and concrete-filled square tubular bracing members. Research Report
UMCE 87–11, Department of Civil Engineering, University of Michigan, Ann Arbor, MI, 1987.
19. Tremblay R, Archambault M-H, Filiatrault A. Seismic response of concentrically braced steel frames made with
rectangular hollow bracing members. Journal of Structural Engineering (ASCE) 2003; 129(12):1626–1636.
20. Yoo JH, Roeder CW, Lehman DE. FEM Simulation and Failure Analysis of Special Concentrically Braced Frame
Tests. Journal of Structural Engineering (ASCE) 2008; 134(6):881–889, Reston, VA.
21. Fell BV, Kanvinde AM, Deierlein GG, Myers AT. Experimental Investigation of Inelastic Cyclic Buckling and
Fracture of Steel Braces. Journal of Structural Engineering (ASCE) 2009; 135(1):19–32.
22. Liu Z, Goel SC. Investigation of Concrete-filled Steel Tubes under Cyclic Bending and Buckling. Report No. UMCE
87–3, Department of Civil Engineering, University of Michigan, Ann Arbor, MI, 1987.
23. Archambault M-H. Etude du Comportement Seismique Des Contreventements Ductiles en X Avee Profiles Tubulaires en
Acier. Report EPM/GCS-1995-09, Department of Civil Engineering, Ecole Polytechnique, Montreal, Que, 1995.
24. Gugerli H. Inelastic cyclic behavior of steel bracing members. PhD thesis, University of Michigan, Ann Arbor, MI,
1982.
25. Walpole WR. Behaviour of cold-formed steel RHS members under cyclic loading. Research Report No. 96–4,
University of Canterbury, Christchurch, New Zealand, 1996.
26. ASTM. ASTM E 1049–85: Standard Practices for Cycle Counting in Fatigue Analysis. West Conshohocken, PA,
2003.
27. Fisher JW, Kulak GL, Smith IFC. A Fatigue Primer for Structural Engineers. Advanced Technologies for Large
Structural Systems (ATLSS), Lehigh University, Bethlehem, PA, 1997.
28. AISC Manual of Steel Construction: Load and Resistance Factor Design, 4th Edition, American Institute of Steel
Construction, Chicago, IL, 2011.
29. Jain AK, Hanson RD, Goel SC. Inelastic response of restrained steel tubes. Journal Structural Division 1978;
104(6):897–910.
30. Thornton WA. Bracing Connections for Heavy Construction. Engineering Journal (AISC) 1984; 21(3):139–148.
31. AISC. Seismic Provisions for Structural Steel Buildings. ANSI/AISC Standard 341–05, American Institute of Steel
Construction, Chicago, IL, 2005.

Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe