See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/333536009

Nonlinear modeling of concentrically braced frames

Article in Journal of Constructional Steel Research · June 2019

DOI: 10.1016/j.jcsr.2019.02.007

CITATIONS READS
34 4,292

4 authors, including:

Andrew Sen Charles W. Roeder

Marquette University University of Washington Seattle
16 PUBLICATIONS 241 CITATIONS 184 PUBLICATIONS 6,343 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Andrew Sen on 10 July 2019.

The user has requested enhancement of the downloaded file.

 1 Nonlinear Modeling of Concentrically Braced Frames

2 Andrew D. Sen1, Charles W. Roeder2, Dawn E. Lehman3, and Jeffrey W. Berman4

3 ABSTRACT

4 Concentrically braced frames (CBFs) comprise a large proportion of lateral-force-resisting

5 systems used in steel building infrastructure around the world. Many categories of CBFs exist,

6 including special, ordinary, and non-seismically detailed (i.e., current “R = 3” or pre-1988

7 construction) CBFs. Experimental testing of these different types of CBFs has shown that they

8 have widely varying nonlinear behavior depending on the relative strengths of their yielding

9 mechanisms and failure modes and level of ductile detailing. Numerical modeling of this range of

10 behavior types is necessary to evaluate the seismic performance, including to quantify potential

11 damage to special CBFs and the vulnerability of lower-ductility CBFs. Special CBFs have been

12 the focus of many previous nonlinear modeling recommendations, including simulation of brace

13 fracture, gusset-plate flexural strength and stiffness, and gusset-plate contribution to frame

14 stiffness. However, recommendations for lower-ductility CBFs have not been well established. To

15 provide important guidance for modeling these common systems, new recommendations based on

16 the large quantity of available experimental data are proposed for simulating fracture of rectangular

17 HSS braces with varying local slenderness, asymmetric load histories, and concrete in-fill; axial

1
Postdoctoral Research Associate, Department of Civil and Environmental Engineering, University of Washington,
Seattle, WA 98195 (corresponding author). E-mail: [email protected]
2
Professor, Department of Civil and Environmental Engineering, University of Washington, Seattle, WA 98195. E-
mail: [email protected]
3
Professor, Department of Civil and Environmental Engineering, University of Washington, Seattle, WA 98195. E-
mail: [email protected]
4
Professor, Department of Civil and Environmental Engineering, University of Washington, Seattle, WA 98195. E-
mail: [email protected]

1
18 yielding of gusset plates; fracture of brace-to-gusset-plate welds; fracture of gusset-plate interface

19 welds; post-fracture, secondary yielding mechanisms in gusset-plate interface connections;

20 yielding beams in the chevron configuration; and buckling columns. These recommendations are

21 validated using experiments of two-story chevron CBFs with yielding beams which simulated an

22 existing, pre-1988 CBF (i.e., a nonductile CBF) and a subsequent repair where the braces and

23 gusset plates were replaced.

24 KEYWORDS

25 Braced frames, nonlinear analysis, seismic effects, steel connections, retrofit

26 1. INTRODUCTION

27 Concentrically braced frames (CBFs) have been frequently used as lateral-force-resisting systems

28 in steel building infrastructure to resist earthquake-induced forces. CBFs are often idealized as

29 trusses in design, but their behavior is more complex as the connections provide significant

30 resistance and stiffness (i.e., the members are not pin-connected) and the braces may buckle and/or

31 yield in large earthquakes. While much of their ductility is associated with the braces, CBFs are

32 complex systems with many other components which contribute to the inelastic response,

33 including beams, columns, and connections (e.g., gusset plates). Consequently, seismic analysis

34 of these systems requires sophisticated nonlinear modeling approaches.

35 There are multiple categories of CBFs in use depending on local seismicity and

36 construction era. The most stringent detailing and design is specified for special CBFs (SCBFs),

37 which are used in regions with high seismic hazard and specified by the American Institute of Steel

38 Construction (AISC) Seismic Provisions for Structural Steel Buildings, ANSI/AISC 341-16 [1].

39 SCBFs are intended to maintain resistance at large inelastic deformations through a specific

2
40 yielding and failure hierarchy which, at a minimum, includes brace buckling in compression and

41 yielding in tension. In addition to these primary yielding mechanisms, SCBFs may exhibit

42 secondary yielding mechanisms within the beams, columns, panel zones, and gusset plates. The

43 expected and preferred failure mode is brace fracture after numerous, large cyclic deformation

44 demands.

45 Gusset plates are commonly used to connect braces to framing members and, in SCBFs,

46 must be designed to accommodate rotations associated with out-of-plane brace buckling; this is

47 often achieved through either a linear [2] or elliptical [3] clearance. As such, gusset-plate flexural

48 yielding often enters into the yielding and failure hierarchy. Roeder et al. [4] extended the SCBF

49 yielding hierarchy with a balanced design procedure (BDP) that encourages secondary yielding

50 mechanisms (e.g., gusset-plate yielding in tension) to increase system deformation capacity

51 without sacrificing strength or reparability. The desired behavior is achieved by suppressing

52 premature failure modes through a combination of capacity-based and ductile design provisions in

53 the Seismic Provisions [1].

54 SCBFs represent just one category of CBFs. Prior to the formal inclusion of SCBFs in the

55 1997 Seismic Provisions [5] and the earlier adoption of capacity-based design principles in the

56 1988 Uniform Building Code [6], CBFs were designed with limited consideration of ductility and

57 without a clear yielding and failure hierarchy. These vintage CBFs are classified as potentially

58 nonductile CBFs (NCBFs). NCBFs are common on the West Coast of the US and share

59 characteristics of non-seismically detailed “R = 3” CBFs used in regions with lower seismic hazard

60 and ordinary (R = 3.25) CBFs (OCBFs) used in regions with moderate and high seismicity (with

61 strict height limitations for regions with high seismicity). NCBFs have several important

62 differences with SCBFs: (1) the braces were not required to meet seismic compactness

3
63 requirements and therefore exhibit premature fracture when subjected to cyclic loading; (2) they

64 are not expected to be able to undergo significant inelastic deformations prior to serious strength

65 degradation, since the beams, columns, and connections were not necessarily designed to develop

66 the expected brace strengths in tension and compression; (3) their welded connections were formed

67 with electrodes with lower toughness than permitted today for demand-critical welds; and (4) the

68 gusset plates were not designed with clearance to accommodate rotation. To better understand and

69 quantify the characteristics of NCBFs, a limited building infrastructure review was conducted, and

70 the results demonstrated that these deficiencies are prevalent in NCBFs [7]. Subsequent

71 experimental research has demonstrated that several of these common failure modes can limit

72 ductility and potentially compromise stability [8].

73 Tools to evaluate the seismic performance of CBFs, including damage to SCBFs and the

74 vulnerability of existing and retrofitted NCBFs, are of great importance to the engineering

75 community. While the American Society of Civil Engineers (ASCE) Seismic Evaluation and

76 Retrofit of Existing Buildings, ASCE 41-17 [9] is intended to provide guidance for modeling

77 recommendations, performance acceptance criteria, and retrofit strategies for CBFs, it does not

78 address most of the issues affecting CBF performance listed above. A study was conducted to

79 improve nonlinear modeling tools for all categories of CBFs. Using experimental research as a

80 validation set, this paper reviews previously proposed modeling approaches for SCBFs and

81 introduces new improvements to simulate fracture and strength degradation for the range of

82 properties CBF components. Specifically, computational methods for modeling brace fracture,

83 connection yielding and fracture, beam yielding, and column buckling are described. With these

84 improvements taken as a whole, these models are a significant advancement over prior work

85 because they can be employed to simulate system behavior of NCBFs, “R = 3” CBFs, OCBFs, and

4
 86 SCBFs (where current models focus on performance modeling of SCBFs alone). The

87 recommendations for modeling are validated using two, two-story chevron CBFs to demonstrate

88 their adequacy.

89 2. BACKGROUND

90 Decades of experimental and computational research have advanced the understanding of the

91 seismic behavior of CBFs and driven the evolution of ductile design provisions [10]. The behavior

92 of CBFs with rectangular hollow structural section (HSS) brace is focused on here due to their

93 wide use in CBFs in existing buildings and new construction, the availability of a large

94 experimental data set, and their pervasiveness in prior nonlinear modeling recommendations. It is

95 noted that many researchers have investigated the cyclic behavior of rectangular HSS braces or

96 brace-gusset assemblies alone, including Black et al. [11], Lee [12], Liu and Goel [13], and

97 Tremblay [14]. This important literature helps establish brace component behavior but neglects

98 the influence of the adjacent components on the brace performance. In view of this, studies which

99 focus on evaluating system behavior through large-scale subassemblage testing and nonlinear

100 analysis will be discussed here.

101 2.1 Experimental Behavior of SCBFs

102 The behavior of SCBFs has been investigated in dozens of subassemblage experiments conducted

103 in the past 20 years, including single-bay tests of 1-, 2-, and 3-story frames under quasi-static,

104 pseudo-dynamic, and dynamic loading. These tests have investigated the effects of bracing

105 configuration, complex behavior at the brace-beam-column intersection, design of gusset plates,

106 and influence of frame action on system behavior. The response quantity “story-drift range” is

107 used below to describe damage states from experiments. Story-drift range refers to the sum of the

5
108 maximum absolute values of story drifts sustained in both directions of loading and is preferred

109 since the cyclic response of the brace depends on both compressive and tensile deformation

110 histories [15].

111 Uriz and Mahin [16] tested a two-story chevron SCBF compliant with the 1997 Seismic

112 Provisions [5]. The first-story braces buckled as intended and fractured after a total story-drift

113 range of approximately 3.7% was achieved. Flexural resistance of the frame (i.e., beams and

114 columns alone) contributed significantly to the system resistance. The beam was sized to resist the

115 unbalanced load at the brace intersection resulting from the difference in the compressive and

116 tensile resistances of the braces and promoting a “yielding-brace” plastic mechanism. However,

117 the braces did not yield in tension due to elastic, downward deflection of the beam. This

118 elongation-limiting effect of relatively strong chevron beams has been noted in one-story

119 experiments of CBFs subjected to dynamic loading as well [17]. After brace fracture, the columns

120 sustained significant damage adjacent to the beam-column connection, where cracking initiated in

121 the interior flanges and extended into the webs at a story-drift range of 5.7%. Subsequent damage

122 localized in the weakened first story of the frame.

123 Extensive testing of one-story, single-diagonal-configuration SCBFs was performed at the

124 University of Washington to investigate the impacts and design of gusset plate connections [3,4].

125 The research demonstrated limited drift capacity resulting from gusset-plate connections designed

126 using conventional methods, most importantly due to excessively strong gusset plates. This

127 research gave rise to the BDP, which enhances system deformation capacity by permitting

128 controlled tensile yielding of the gusset plate, utilizing the elliptical clearance model described by

129 Lehman et al. [3], and designing the interface welds to develop the tensile strength of the plate. A

130 smaller, thinner gusset plate results from this method, which also minimizes the required interface

6
131 weld size. Roeder et al. [18] and Lumpkin et al. [19] demonstrated the effectiveness of the BDP in

132 a series of two- and three-story, multi-story-X SCBF experiments where quasi-static loading was

133 applied at the top of the frame. These specimens had larger story-drift range capacities (between

134 4.3 and 5.8%) and less severe beam and column damage than reported in prior studies, due, in part,

135 to the improved gusset-plate design. In the multi-story-X brace configuration, the braces fully

136 yielded in tension and the desired yielding and failure hierarchy was achieved. Figure 1a illustrates

137 the extent of yielding and failure mode sustained by one of the three-story specimens (upper story

138 not shown), TCBF2-HSS, and the hysteretic response of its first story.

139 2.2 Experimental Behavior of Existing and Retrofitted NCBFs

140 Relative to SCBFs, experimental research on NCBFs has been limited. A recent study funded by

141 the National Science Foundation investigated the NCBF deficiencies revealed by the infrastructure

142 study discussed previously. Most importantly brace, beam, and gusset-plate connection

143 deficiencies were studied experimentally.

144 The earliest tests investigating the impact of beam yielding were large-scale tests

145 conducted in the 1980s as part of a US-Japan cooperative research program [20–22]; these are

146 today considered tests of NCBFs, though they represented the state of the practice at the time.

147 Compared to modern requirements, these experiments had braces with large local slenderness

148 ratios that do not meet the current limits for highly ductile members [1], fully restrained brace ends

149 or gusset plates lacking clearance to permit brace-end rotation, welds formed with lower-toughness

150 filler metal, and beams as part of the chevron configuration which were susceptible to yielding

151 after brace buckling. A major implication of this international collaboration was the introduction

152 of specific provisions to prevent beam yielding in favor of the yielding-brace mechanism [21].

7
153

154 FIGURE 1. Experimental behavior of first story of multi-story CBFs: (a) Specimen TCBF2-HSS,

155 (b) Specimen TNCBF1-N-HSS, and (c) Specimen TNCBF1-R-HSS.

156

157 Recent experiments have re-examined the yielding-beam mechanism in NCBFs [23,24].

158 The beams in these experiments were continuous, and Figures 1b and 1c show the locations of

159 yielding, failure modes, and hysteretic behavior for the first story of existing and retrofitted two-

160 story NCBFs loaded quasi-statically at the top of the frame [23]. The existing NCBF specimen,

161 TNCBF1-N-HSS (Fig. 1b), had premature brace fracture at an inter-story drift range of only 1.2%

162 due to the high local slenderness ratio of the braces (2.1 times larger than permitted). The frame

163 maintained significant resistance after brace fracture as the beam acted like a long link in an

164 eccentrically braced frame (EBF); similar effects have been observed in testing of an “R = 3” CBF

165 [25]. The brace-to-gusset connection fractured after a drift range of 2.0%. The system behavior

166 was clearly nonductile due to the premature brace and connection fractures. A second specimen,

8
167 TNCBF1-R-HSS (Fig. 1c), retained the weaker beam but replaced the first-story braces and gusset

168 plates to meet the recommended BDP requirements for SCBFs [4]. With the elimination of the

169 brace and gusset plate deficiencies, the yielding-beam mechanism formed and provided ductile

170 behavior. Delayed brace fracture occurred at an inter-story drift range of 4.8%, consistent with the

171 drift capacity observed for SCBFs even though the weaker beam did not meet the beam-strength

172 requirements of the Seismic Provisions [1].

173 A series of 18 single-story, single-diagonal-configuration tests of existing and retrofitted

174 NCBFs were conducted to examine brace and connection deficiencies (like those shown in Fig.

175 1b), which appear to be more critical than the moderate beam-strength deficiency (i.e., up to about

176 2.5 times weaker than permitted) in the chevron configuration (see Sen et al. [8] for full details).

177 The experimental program included a variety of gusset-plate connection configurations (welded

178 and bolted connections with shear plates, double angles, end plates, etc.) and retrofit strategies

179 (buckling or buckling-restrained brace replacement, brace concrete in-fill, weld reinforcement,

180 etc.). Figure 2 summarizes the four (4) main behavior types (Types A through D) observed in these

181 tests, which are distinguished by the severity and sequence of failure modes. Each column of Fig.

182 2 links backbone curve behavior (Fig. 2a) with failure mode sequence (Fig. 2b) and example

183 damage observations from the experimental program (Fig. 2c). The backbone curve behavior is

184 presented in terms of generalized lateral force and displacement quantities, Q and Δ, respectively.

185 The backbone curves are identical until yielding (Δ𝑦 , Qy), at which point the different failure

186 mechanisms develop at a deformation Δ𝑓𝑖 which may reduce strength to a residual force, Qri or

187 zero. This post-yielding response depends upon the failure-mode types described below. In

188 addition, it is noted that the specimen identifiers in the connection drawings of Fig. 2c correspond

189 to specimen names in Sen et al. [8].

9
190 Three (3) types of failure-mode limit states (L1, L2, and L3) are defined in Table 1 to

191 clarify the discussion. Table 1 also provides examples of each failure-mode type which are possible

192 in brace-gusset assemblies. The failure mode sequences in Fig. 2b describe the progression of

193 system behavior from an initial state (“0”) to a potential collapse due to an L3 failure mode. L3

194 failure modes were rarely observed in the experimental study, even at relatively large drifts

195 following an L2 failure mode. Stepping alphabetically from Type A to Type D, the behavior types

196 transition from highly nonductile to ductile, with Type D representing behavior of an SCBF.

197

198 TABLE 1. Failure-Mode Types.

Type Description Examples
L1 Brace partially disconnected from the frame such that  Fracture of a single gusset plate
it continues to develop axial forces below its full tensile interface connection
strength.
L2 Total loss of brace axial resistance. Residual resistance  Brace fracture
provided by frame action.  Brace-to-gusset plate connection
fracture (fracture of welds, bolts, net
section, etc.)
 Gusset plate Whitmore section or block
shear fracture
 Fracture of both gusset plate interface
connections
L3 Severe or total loss of lateral load resistance in the  Beam-to-column connection fracture
system, likely leading to collapse.
199

200 2.3 Previously Developed Modeling Approaches for Nonlinear Analysis

201 Modeling approaches for nonlinear analysis of CBFs are provided in ASCE/SEI 41-17 [9] but focus

202 almost entirely on the brace, which is modeled as a translational spring, as shown in Fig. 3a, with

203 simplified behavior like that of the Type-B backbone curve in Fig. 2a. In typical practice, the brace

204 and beam ends would be modeled as pins and flexural yielding of the columns would be simulated

205 using nonlinear rotational springs (again, with simplified behavior like that of Type B in Fig. 2a).

206 This approach is depicted in Fig. 3a and significantly underestimates the strength and stiffness of

10
207 the system [26] (Figs. 3b and 3c depict more sophisticated approaches described later). Further,

208 ASCE/SEI 41-17 requires brace connection limit states, which are known to be commonly deficient

209 in NCBFs, to be analyzed as force-controlled actions unless explicit modeling approaches are

210 implemented and the inelastic behavior is stable and ductile. Force-controlled actions have

211 unacceptable performance if the demand computed from analysis exceeds the lower-bound

212 strength of the component. However, “deficiencies” which are yielding mechanisms are not

213 usually detrimental to system behavior and the development of a localized failure mode in the

214 brace or connection does not constitute failure of the entire system. Thus, NCBF performance may

215 be judged too severely without explicit nonlinear modeling.

216

217

218 FIGURE 2. Behavior types for gusset plate connections of existing and retrofitted NCBFs.

219

11
220 More accurate nonlinear analysis modeling approaches have been well established for

221 SCBFs for research applications [26–30]. These approaches have been implemented in OpenSees

222 [31] and employ a combination of fiber-based beam-column elements and rotational springs to

223 simulate the intended SCBF yielding and failure hierarchy. The specific approach by Hsiao et al.

224 [26,28], which is extended to NCBFs, “R = 3” CBFs, and OCBFs in this paper, has the following

225 characteristics:

226  Beams and columns are modeled with a single, force-based nonlinear beam-column

227 element between work points with five (5) integration points per element;

228  Braces are discretized with 16 elements between work points with four (4) integration

229 points per element and given an initial sinusoidal imperfection with an apex of L/500,

230 which has been shown to provide compressive strength commensurate with the AISC

231 buckling curve [32,33];

232  Braces, beams, and columns, have rigid offsets approximated with stiff elastic beam-

233 column elements (e.g., elastic modulus 103 times that of the original material) or

234 analytical rigid links to account for connection geometry, including 75% of the gusset-

235 beam connection length and the full gusset-column connection length;

236  Gusset plates are modeled with nonlinear springs in the rotational degree of freedom

237 normal to the direction of buckling (usually out of plane) with rotational stiffness and

238 moment capacity calculated using the plate’s yield stress, thickness, Whitmore width,

239 and length; and

240  Beam-to-column connections, which are typically bolted shear plates, are modeled with

241 nonlinear rotational springs in the plane of the frame after Liu and Astaneh-Asl [34].

12
242 It is noted that these modeling decisions are not trivial and may impact system performance. For

243 example, D’Aniello et al. [35] demonstrated that the initial imperfection used to simulate brace

244 buckling can impact drift demand and collapse probability.

245 The response of each line element (braces, beams, and columns) is obtained using fiber

246 sections representing the initial cross-sectional shape of the member. Individual steel fibers are

247 modeled with the Steel02 (Giuffré-Menegotto-Pinto) material [36,37]. The Type D behavior

248 depicted in Fig. 2 can be simulated well for SCBFs, where the L2 failure mode is delayed brace

249 fracture (following the intended yielding and failure hierarchy). This is accomplished

250 computationally with material “wrappers” which monitor strain history in the underlying material

251 and initiate fracture of individual brace fibers based on either the maximum strain range (MSR)

252 between tension and compression (Hsiao et al. [28]) or a low-cycle fatigue (LCF) model (Uriz and

253 Mahin [27]). When the fracture criterion is reached, the strength and stiffness of the fiber is reduced

254 by the wrapper to nearly zero; conceptually, the fiber is removed from the section at that location,

255 and the progressive fracture of the fibers leads to complete brace fracture.

256 To illustrate the effectiveness of the approach by Hsiao et al. [26], two experiments of

257 retrofitted NCBFs, Specimens 4O-W and 8O-W, were numerically simulated (these specimens are

258 described in Sen et al. [8]). Each consisted of a single diagonal braced bay, as shown in the

259 drawing; the OpenSees schematic model of these systems are shown in Figs. 4c and 4d. The frames

260 were theoretically retrofitted using a two-part scheme where (1) the brace was replaced with one

261 that meets highly ductile local slenderness requirements and (2) the gusset-plate interface welds

262 were reinforced with fillet welds meeting demand critical requirements to obtain sufficient

263 combined thickness to develop the tensile capacity of the gusset plate. The failure mode in each

264 was delayed brace fracture, an L2 failure mode which is simulated using the maximum strain range

13
265 (MSR) approach by Hsiao et al. [28] using experimentally calibrated MSR values at fracture

266 (MSRf). As shown in Fig. 5, the models approximate the global behavior of the specimens well

267 both before and after brace fracture occurs.

268 While SCBFs and NCBFs with Type-D behavior can be simulated using existing modeling

269 approaches, research on advanced computational modeling of NCBFs with other behavior types

270 has been comparatively sparse. Following ASCE/SEI 41-17 [9], practicing engineers are permitted

271 to consider the effects of NCBF deficiencies in post-analysis evaluation rather than through

272 explicit modeling; researchers have employed this approach to evaluate fracture of HSS braces

273 with and without concrete infill [24] and brace connections [38,39]. Others have modeled NCBF

274 deficiencies and retrofit strategies to better understand their implications on system behavior. For

275 example, Rai and Goel [40] analyzed a four-story NCBF damaged in the 1994 Northridge

276 Earthquake using SNAP-2DX [41] and simulated fracture of existing HSS braces with and without

277 concrete infill. The improved fracture life resulting from the concrete fill was modeled using an

278 equivalent (reduced) local slenderness ratio [12] as an input parameter in a low-cycle fatigue model

279 calibrated for hollow HSS braces [42]. More recently, Hsiao et al. [43] and Sizemore et al. [44]

280 modeled CBFs in OpenSees with brace-to-gusset weld fracture using a nonlinear translational

281 spring that fractures when the weld strength is exceeded. These explicit modeling approaches

282 provide a basis for simulating NCBFs, but given the variety in both possible deficiencies and

283 behavior types, there is a clear need for new modeling approaches which can be used to evaluate

284 seismic performance beyond initial failure modes; such approaches will be addressed in this paper.

285 Specifically, the modeling approaches include: (1) an improved brace fracture model to simulate

286 a wider range of brace geometries and load histories, (2) connection models to simulate the

287 common failure-mode types and sequences shown in Fig. 2, and (3) more accurate models for the

14
288 beams and columns. The following sections summarize the validated modeling approaches for

289 each of these aspects of the nonlinear behavior and failure models.

290

291 FIGURE 3. Nonlinear analysis approaches for CBFs.

15
292

293 FIGURE 4. Experimental setup and corresponding OpenSees modeling approach for various

294 experimental research programs.

295

296 FIGURE 5. Comparison of experimental and numerical global behavior.

16
297 3. BRACE FRACTURE SIMULATION

298 Brace fracture is an L2 failure-mode type and the anticipated failure mode for SCBFs and NCBFs

299 which do not have severe connection and system deficiencies. Under reversed cyclic loading

300 characteristic of far-field earthquakes, brace fracture is driven by (1) local deformation of the cross

301 section (local cupping in HSS braces) associated with incompatibility between the undeformed

302 cross-sectional shape and the large curvature demands at the plastic hinge location in compression,

303 and (2) subsequent axial deformation demand in tension. Slender cross sections (high b/t in HSS)

304 develop this local post-buckling deformation at smaller deformations, resulting in reduced fracture

305 life. SCBFs delay this behavior by employing stockier cross sections which meet highly ductile

306 member requirements of the Seismic Provisions [1]. Fiber-based beam-column elements do not

307 simulate this local behavior, and hence calibration of the fracture criterion is required. Previously

308 proposed brace-fracture models (e.g., Hsiao et al. [28], Karamanci and Lignos [29], and Tirca and

309 Chen [30]) are calibrated from nondimensionalized geometric and material parameters: local

310 slenderness ratio (b/t), global slenderness ratio (Lc/r = KL/r), and elastic-modulus-to-yield-stress

311 ratio (E/Fy). Recall that the present work is limited to rectangular HSS; for research on round HSS

312 and wide-flange sections the reader is referred to Karamanci and Lignos [29], which uses the

313 approach proposed by Uriz and Mahin [27].

314 Hsiao et al. [28] propose the following MSRf prediction equation for rectangular HSS:
0.2
𝑏 −0.4 𝐿c −0.3 𝐸
𝑀𝑆𝑅f,force = 0.1435 ( ) ( ) ( ) (1)
𝑡 𝑟 𝐹y

315 The value predicted by Eq. 1 is denoted MSRf,force because it was calibrated using force-based

316 beam-column elements (FBEs). While Eq. 1 works reasonably well (R2 of 0.41), particularly for

17
317 SCBFs, several factors necessitated improvement of the MSR-based fracture criteria to simulate

318 all CBFs:

319  Displacement-based beam-column elements (DBEs) may be preferred for brace modeling

320 since large, sudden changes in strength and stiffness due to buckling and fracture can be

321 solved at the global level; in OpenSees, this allows for the use of robust solution algorithms

322 in comparison to FBEs, which were used in the prior work by Hsaio et al. [28].

323  Recent experimental research on chevron-configuration CBFs shows that brace fracture

324 life is longer under reversed cyclic loading when the brace axial deformation demand in

325 tension is limited, even if axial deformation in compression is large [45]. In chevron-

326 configuration CBFs, this asymmetry in demand is attributed to beam deflection resulting

327 from the unbalanced brace loads in tension versus compression, but similar effects exist in

328 other bracing configurations when drift demand is asymmetric. This load-direction effect

329 is not accounted for by existing modeling approaches and is incorporated into the MSRf

330 expression with a brace axial deformation ratio term.

331  Additional experimental data is available for calibration that was not used in prior work,

332 including data collected from tests with high-b/t (much larger than the current requirement

333 for highly ductile members [1]) rectangular HSS braces common to NCBFs and concrete-

334 filled braces rectangular HSS braces which may be used in retrofit.

335 These issues are addressed through new brace fracture models based on the MSR approach by

336 Hsiao et al. [28]. Note that the brace fracture model in Eq. 1 and others proposed in this paper are

337 sensitive to element type (DBE or FBE), fiber discretization, element discretization, and steel

338 constitutive modeling parameters and hysteretic rules. These properties are held constant in this

339 study, where the element type is specified for each prediction equation, the brace section is

18
340 discretized with four (4) fiber layers through the wall thickness and four (4) fiber layers along the

341 wall flats and corners (128 total fibers), the brace length is discretized with 16 elements with four

342 (4) integration points per element, and the Steel02 material model in OpenSees is used with 1%

343 kinematic strain hardening (R0 of 15, cR1 of 0.925, and cR2 of 0.15). Using these models with a

344 different element type, discretization schemes, or material properties will not result in the same

345 simulated brace fracture life and requires recalibration of the MSRf. Appendix A provides data

346 used to calibrate the brace fracture models presented here.

347 3.1 Maximum Strain Range and Load History Effects

348 The experiments from the numerous test programs referenced in Table A.1 were simulated in

349 OpenSees using the above modeling approach. Equation 2 was developed to provide a new MSRf

350 prediction equation and is termed MSRf,disp since it is calibrated from and intended for use with

351 DBEs. This prediction equation is calibrated from an expanded data set, including 17 additional

352 tests (59 total tests) of rectangular HSS braces from Lee [12], Liu and Goel [13], Sen et al. [8], and

353 Richard [46]. The constant and variable coefficients (exponents in Eq. 2) were determined through

354 linear regression of the data in logarithmic space (R2 of 0.71). This data set includes a higher

355 proportion of test data from braces with b/t ratios which exceed the highly ductile local slenderness

356 ratio limit, λhd, than included in Hsiao et al. [28] (37% versus 26%).
0.21 0.068
𝑏 −0.75 𝐿c −0.47 𝐸 𝛿c,max
𝑀𝑆𝑅f,disp = 0.554 ( ) ( ) ( ) ( ) (2)
𝑡 𝑟 𝐹y 𝛿t,max

357 Equation 2 also contains an additional term, δc,max/δt,max, which quantifies the ratio of the

358 axial deformations to capture the direction bias based on the maximum global brace deformations

359 recorded in compression and tension (δc,max and δt,max, respectively). To compute the prediction

360 equation, δc,max/δt,max is calculated for each experiment from the simulation (as opposed to the

19
361 measured value) when fracture occurs. Strain (and hence MSRf) is a local measure, but axial brace

362 deformation is considered here to avoid increasing the importance of the computed strain from

363 fiber-based line-element models, where important local effects such as cupping are not simulated.

364 In Eq. 2, a larger δc,max/δt,max ratio results in a larger MSRf value. Figure 6 illustrates this

365 load-history effect using Specimens Lee-6 and Lee-7 from Lee [12], which were brace-only tests

366 as shown in Figs. 4a and 4b. These specimens were nominally identical with HSS4×4×1/4 braces

367 with a 3.3-m end-to-end brace length loaded with different displacement histories characteristic of

368 chevron-configuration braced frames with yielding beams. Deflection of the beam in this

369 configuration limits tensile elongation and increases compressive shortening of the brace, which

370 results in δc,max/δt,max ratios greater than unity, even if drift demand is symmetric. Specimens Lee-

371 6 and Lee-7 had δc,max/δt,max ratios of 12.4 and 4.92, respectively (see Fig. 6c). Specimen Lee-6

372 achieved a larger axial deformation range prior to fracture, corresponding to a larger MSRf value

373 (0.0550 versus 0.0479).

374 Unlike other parameters of MSRf,disp, δc,max/δt,max is not known a priori; instead δc,max/δt,max

375 varies with time under reversed cyclic loading, as shown in Fig. 6c for Specimens Lee-6 and Lee-

376 7. Consequently, MSRf,disp also varies with time and thus requires a specific computational

377 implementation. In OpenSees, this is accomplished with the MaxStrainRange material wrapper.

378 Tags of the brace end nodes can be passed to the MaxStrainRange material, allowing the material

379 to compute global deformation from the nodal coordinates and update δc,max/δt,max and MSRf at each

380 step of the analysis.

381

20
382

383 FIGURE 6. Experimental and numerical brace hysteretic response of specimens with (a) high

384 δc,max/δt,max ratio loading, (b) moderate δc,max/δt,max ratio loading; (c) evolution of δc,max/δt,max for

385 each specimen (adapted from Lee [12]).

386

387 3.2 Concrete-Filled Braces

388 Filling locally slender HSS braces with concrete is an effective method for improving brace

389 fracture life and presents an attractive seismic retrofit scheme for NCBFs [8,13]. To be effective,

390 the concrete should delay strain concentration in the plastic hinge region of the brace in

391 compression by preventing inward cupping of the brace wall, leading to a less severe outward

392 deformation mode. This beneficial phenomenon has been observed in experiments of HSS braces

393 designed to act either as a composite or noncomposite member. In composite HSS braces, the

394 concrete fills the full length of the brace and increases its critical buckling load. If the concrete is

395 blocked out just before the gusset-plate region, the concrete does not stiffen or strengthen the

396 gusset plate; this is advantageous in seismic retrofit since the brace demands on the connections,

397 which are often vulnerable, would not be increased.

398 Steel-concrete composite action can be numerically simulated by including concrete fibers

399 in the interior of the steel tube. In this work, a 4-by-4 grid of concrete fibers is used as in Fig. 7c,

21
400 matching the mesh length for the elements of the steel tube. Using this formulation, the concrete

401 is perfectly composite (perfect bond and no relative movement) with the steel. The concrete is

402 modeled using Concrete02 with very small tensile strength and stiffness to aid convergence.

403 Unlike recommendations for modeling concrete-filled tubes subjected to lateral loads (e.g.,

404 Stephens et al. [47]), the concrete strength deteriorates in compression after the peak compressive

405 stress as shown in Fig. 7c. Total strength degradation occurs at 10εco, where εco is the strain at the

406 peak compressive stress defined as (f’c)1/4/1150 for f’c in MPa [48]. This strength degradation is

407 necessary to increase the tensile reloading stiffness of the brace after buckling; without

408 degradation, the concrete will retain significant compressive stress until the brace has fully

409 straightened. Using the concrete model in Fig. 7c, the concrete fill primarily contributes to the

410 initial buckling load of the brace but does not significantly increase the post-buckling strength,

411 which is consistent with experimental results.

412 Non-composite braces require no alteration to the fiber section. In both composite and non-

413 composite cases, the improved fracture life attained with the concrete fill is accounted for with the

414 MSRf prediction equation given by Eq. 3, which is intended for use with displacement-based

415 elements and denoted MSRf,conc. This constant-MSRf,conc model was calibrated using numerical

416 simulations of ten experiments with concrete-filled braces described in Table A.1, and the constant

417 value is the geometric mean of the calibrated MSRf,conc values. Linear models were considered

418 using combinations of the same nondimensionalized parameters as in Eqs. 1 and 2, but correlation

419 between the parameters and calibrated MSRf,conc values was not statistically significant. Note that

420 the constant MSRf,conc value in Eq. 3 must be larger than the corresponding MSRf,disp for the bare

421 steel tube, since a decrease in MSRf due to concrete in-fill is not logical.

𝑀𝑆𝑅f,conc = 0.0505 ≥ 𝑀𝑆𝑅f,disp (3)

22
422 Figures 7a and 7b show the effects of concrete in-fill using Specimens Liu-T633H and Liu-

423 T633C6 from Liu and Goel [13], which were brace-only experiments, as depicted in Fig. 4a. The

424 specimens were nominally identical with HSS6×3×3/16 braces and the same load histories prior

425 to fracture of the hollow brace specimen. The concrete-filled brace was designed to be composite

426 using concrete with a specified compressive strength of 41.4 MPa (6 ksi). Comparing both Figs.

427 7a and 7b, the compressive strength, deformation range, and calibrated MSRf,conc of the composite

428 brace are larger than those of the hollow brace. The additional compressive strength due to

429 composite action is also simulated using the fiber model approach described above.

430

431

432 FIGURE 7. Experimental and numerical brace hysteretic response of specimens with (a) hollow

433 and (b) concrete-filled braces; (c) modeling of composite fiber section (not to scale) and concrete

434 material (adapted from Liu and Goel [49]).

435 4. GUSSET-PLATE CONNECTION SIMULATION

436 Most of the observed variation in the behavior of NCBF and retrofitted NCBF specimens is

437 attributed to yielding mechanisms and failure modes associated with the gusset-plate connection.

438 NCBFs are known to have a wide range of gusset-plate connection configurations, particularly at

439 the brace-beam-column intersection (i.e., corner gusset plates) [7]. For example, Fig. 2c shows

23
440 four (4) corner gusset-plate connections, each eliciting different behaviors defined by the sequence

441 of failure-mode types (L1-L3). The connection configuration itself does not determine the behavior

442 type but it affects which limit states are present and hence which behavior types are possible. For

443 instance, a connection without a yielding mechanism on the vertical gusset plate interface would

444 not be able to develop Type-C behavior if the horizontal gusset-plate interface connection fractured

445 (see Specimen 8O in Fig. 2c). On the other hand, if the vertical interface was a bolted connection

446 with the potential for bolt-hole elongation, a secondary yielding mechanism could develop after

447 gusset-to-beam weld fracture. This secondary yielding mechanism provides post-fracture integrity,

448 resulting in Type-C behavior (see Specimen 6O in Fig. 2c). The modeling approach is therefore

449 dependent upon an intact load path and the yielding mechanisms and failure modes that control

450 connection behavior.

451 This section provides a framework for modeling gusset plate connections with the

452 capability of producing any of the behavior types shown in Fig. 2. These behavior types are

453 simulated here through specific consideration of the gusset-plate axial yielding, brace-to-gusset

454 weld fracture, and gusset-plate interface weld fracture limit states, but it is possible to extend the

455 framework to model other limit states if their component response can be rationally predicted.

456 4.1 Gusset-Plate Yielding

457 Gusset-plate axial yielding in tension is generally a beneficial yielding mechanism which can

458 increase the drift capacity of CBFs. If a gusset plate yields in tension, inelastic deformation demand

459 is shared between the brace and plate, reducing demands on the brace and delaying brace fracture.

460 Encouraging this mechanism just after brace tensile yielding in SCBFs is one of the central

461 components of the BDP, which recommends the use of the expected yield stress of the plate and a

462 balance factor (analogous to a resistance factor) of one for evaluating yielding of the Whitmore

24
463 section [4]. As indicated previously, tensile yielding of the gusset plate is simulated using an axial

464 spring in the direction of the brace (Fig. 3c). This section describes how to determine the strength,

465 stiffness and fracture of this axial spring model.

466 Using the BDP, the demand-to-capacity ratio (DCR) for gusset-plate yielding would be

467 equal to one (1) using Eq. 4. In NCBFs, gusset-plate yielding may also occur since connections

468 were not required to be designed for the expected strength of the brace in tension, the numerator

469 in Eq. 4. If this DCR is greater than one, gusset-plate yielding controls the strength and the capacity

470 of the brace in tension cannot be developed, reducing the overall strength of the system.

(𝑅y 𝐹y ) 𝐴g
br (4)
𝐷𝐶𝑅gpy =
(𝑅y 𝐹y )p 𝐵w 𝑡p

471 The axial behavior of the gusset plate can be approximated using the Whitmore section and

472 length, where the strength and stiffness is approximated using Equations 5 and 6. Note that this

473 geometry, shown in Fig. 8b, is used by Hsiao et al. [26] to model the flexural strength and stiffness

474 of the plate. The Whitmore length, Lavg, is the average of L1, L2, and L3, where any of the individual

475 lengths are negative if the Whitmore width extends beyond the gusset plate (e.g., L3 in Fig. 8b

476 would be a negative value).

𝑃y = (𝑅y 𝐹y ) 𝐵w 𝑡p (5)
p

𝐸𝐵w 𝑡p
𝐾= (6)
𝐿avg

477 This model for gusset-plate axial yielding was implemented in OpenSees using an axial

478 spring in line with the brace (see Fig. 3c) with force-displacement behavior defined by the Steel02

479 uniaxial material model (using R0 of 15, cR1 of 0.925, and cR2 of 0.15). A kinematic strain

480 hardening ratio of 0.2% is recommended because it gave the most accurate computational

25
481 simulation of the available experimental specimens. This translational spring can be used within

482 the same zero-length element as the gusset plate’s flexural spring.

483

484

485 FIGURE 8. Gusset plate (a) weld details, (b) Whitmore width and length geometry, and (c)

486 elliptical clearance geometry.

487

488 Figure 9 shows an implementation of this model using Specimen 2O from Sen et al. [8],

489 where the gusset plate had a Whitmore yielding DCR of 1.3. Specimen 2O was a one-story, one-

490 bay NCBF with a single diagonal brace as shown in Fig. 4c and modeled in Fig. 4d. In Fig. 9a, the

491 gusset plate yielding model is not implemented in the numerical simulation, and the lateral

492 resistance of the system is overestimated. Results from the numerical simulation with the proposed

493 model are shown in Fig. 9b, and Fig. 9c indicates that the spring yielded and elongated to

494 approximately 7 mm, a large but reasonable deformation for the gusset plate. The failure mode of

495 Specimen 1E is gusset plate interface weld fracture, and simulation of this failure mode is

496 described below.

497

26
498

499 FIGURE 9. Experimental and numerical frame response of Specimen 2O (a) without and (b) with

500 the gusset plate yielding model; (c) response of corresponding gusset plate translational spring.

501

502 4.2 Brace-to-Gusset Plate Weld Fracture

503 Fracture of the brace-to-gusset plate welded joint is considered an L2 failure-mode type since the

504 brace load path is lost. Using HSS braces, the brace is most often slotted and connected to the

505 gusset plate with four longitudinal fillet welds of leg size wbr and length Lw (see Fig. 8a).

506 Consequently, the DCR for brace-to-gusset weld fracture is given by Eq. 7. A DCR equal to or

507 below the AISC resistance factor, ϕ = 0.75, would be acceptable for modern design.

(𝑅y 𝐹y ) 𝐴g
br (7)
𝐷𝐶𝑅brw =
1.2√2𝐹EXX 𝑤br 𝐿w

508 Several researchers have previously simulated the behavior of HSS brace-to-gusset plate

509 welds, including fracture. Hsiao et al. [43] implemented a translational spring at the brace end

510 (similar to the spring for gusset plate yielding discussed above). The spring had a linear-elastic

511 force-displacement behavior and a fracture displacement calibrated from the experimental

512 response of a single NCBF specimen (Specimen 1E as described by Sen et al. [8]). Sizemore et al.

513 [44] followed a similar approach for modeling brace-to-gusset weld fracture observed in

27
514 experiments of an “R = 3” CBF and an OCBF. In this alternative model, fracture was initiated

515 based upon a force limit and post-fracture contact between the brace and gusset plate was simulated

516 with a gap-contact element in parallel with the brace-to-gusset weld spring. This contact behavior

517 has also been observed in experiments of NCBFs [23,43] but is difficult to predict, as the brace

518 potentially bears on the gusset plate and/or adjacent concrete slab, which can in turn deform.

519 Therefore, this behavior is not considered in the modeling approach here.

520 A similar approach for modeling brace-to-gusset weld behavior is proposed here using

521 existing understanding of weld nonlinear behavior. The AISC Steel Construction Manual [50]

522 defines the normalized stress-deformation behavior of welds as Eq. 8 for use with the instantaneous

523 center of rotation method when determining the weld-group strength. In Eq. 8, p is defined as the

524 ratio of the weld deformation, Δ, to the deformation at ultimate load, Δm. This relationship, based

525 on work by Lesik and Kennedy [51], is proposed for use in modeling brace-to-gusset weld fracture

526 using a translational spring in the direction of the brace. Equation 9 utilizes Eq. 8 to develop a

527 force-deformation relationship for a group of four concentrically loaded longitudinal fillet welds,

528 as would be common for an HSS brace-to-gusset plate weld. Equation 9 uses Δm of 0.167wbr, also

529 based on findings from Lesik and Kennedy [51].

𝐹nw (𝑝) = 0.60𝐹EXX (1.0 + 0.50 sin1.5 𝜃𝑖 )[𝑝(1.9 − 0.9𝑝)]0.3 (8)

Δ Δ 0.3
𝑃brw (Δ) = 4.80𝐹EXX 𝑤br 𝐿𝑤 [ (0.35 − )] (9)
𝑤br 𝑤br

530 The force-deformation behavior of the brace-to-gusset weld can be modeled in OpenSees

531 using a translational spring at the brace end, identical to the approach for the simulation of gusset

532 plate axial yielding. The generic MultiLinear uniaxial material can be used to model the force-

533 deformation relationship given by Eq. 9; note that this material has a fully hysteretic response.

534 Fracture can be modeled using the MaxStrainRange material wrapper with a unidirectional

28
535 deformation limit of Δu (0.339wbr for longitudinally loaded welds). Note that if both gusset-plate

536 axial yielding and brace-to-gusset weld fracture need to be modeled, these materials can be placed

537 in series in OpenSees using the Series uniaxial material.

538 Figure 10 shows the dual implementation of gusset-plate axial yielding and brace-to-gusset

539 weld fracture using the Series material for Specimen 1E in Sen et al. [8], where the Whitmore

540 yielding DCR was 1.0, the brace-to-gusset weld fracture DCR was 1.7, and the failure mode was

541 brace-to-gusset weld fracture. In Fig. 10a, a base model is shown in which only the gusset-plate

542 axial yielding model is implemented. Equation 2 was used to predict the MSRf,disp value for the

543 locally slender HSS6×6×1/4 brace. Since the Whitmore yielding DCR was not excessive, the brace

544 yielded and fractured in tension (BRF) in the numerical simulation. Both the strength and

545 deformation capacity prior to fracture are overpredicted compared to the experimental results.

546 Figure 10b shows the response with a similar model which includes the proposed brace-to-gusset

547 weld fracture (BRW) model (depicted in Fig. 10c) in series. The lateral resistance when the brace

548 is in tension (positive drift) and the fracture behavior is predicted well using a wbr of 1.25 times

549 the specified weld leg size, which may be attributed to oversizing during fabrication or

550 overstrength of the filler metal. In both the base model and proposed model results shown in Figs.

551 10a and 10b, the post-fracture resistance when the brace is in compression (negative drift) is

552 underpredicted by the numerical simulation since post-fracture contact is not considered.

553

29
554 FIGURE 10. Experimental and numerical frame response of Specimen 1E (a) without and (b)

555 with the brace-to-gusset weld fracture model; (c) response of corresponding gusset plate

556 translational spring.

557

558 4.3 Gusset-Plate Interface Weld Fracture

559 One of the most common deficiencies reported in NCBFs is the inadequacy of the welds which

560 connect the gusset plate to adjacent beams, columns, and other connecting elements (e.g., shear

561 plates) [7]. These welds are termed gusset-plate interface welds here, as indicated in Fig. 8a for a

562 gusset plate welded directly to the beam and column. Such connections are common in CBFs,

563 since one interface of the gusset plate can be shop-welded to the beam or column to facilitate

564 erection. The remaining gusset-plate interface is either welded or bolted in the field.

565 Gusset-plate interface welds are vulnerable components of the connection since they are

566 subjected to considerable in- and out-of-plane deformation demands due to opening and closing

567 rotations of the beam-column joint and bending of the gusset plate in out-of-plane buckling

568 configurations. Interface-weld tearing generally initiates after buckling when the brace is in

569 compression, the beam-to-column connection is opening, and the plate is bent out of plane. This

570 failure mode is avoided in SCBFs by:

571  Sizing the weld to develop the plastic capacity of the gusset plate;

572  Providing adequate clearance to accommodate gusset-plate rotation when the brace

573 buckles, usually with either the 2tp linear [2] or 8tp elliptical [3] clearance model (see Fig.

574 8c); and

30
575  Using filler metal which satisfies demand critical weld toughness requirements [1] to allow

576 ductile tearing of the welds without complete fracture.

577 In the Seismic Provisions [1], the weld strength criterion is met by designing the weld to

578 resist either 0.6RyFytp per unit length of shear (simplified approach) or the assumed edge forces in

579 combination with the reduced weak-axis flexural strength of the plate [52]. The BDP provides a

580 more stringent but well validated alternative approach based on the tensile capacity of the plate

581 and weld [4]. The BDP expression for the interface weld DCR is given by Eq. 10. For design

582 purposes, DCRgpw must be less than the corresponding resistance factor, ϕ = 0.75. Note that for

583 ASTM A572 Grade 50 plate and E70 electrodes, the required leg size in design is 0.82tp for the

584 BDP and 0.74tp for the simplified AISC approach.

(𝑅y 𝐹y )p 𝑡p
𝐷𝐶𝑅gpw = (10)
0.9√2𝐹EXX 𝑤p

585 NCBFs are not expected to meet any of the above criteria to ensure ductile tearing of the

586 gusset plate interface welds, and recent experimental research suggests that these welds are prone

587 to premature, brittle fracture after brace buckling has occurred [8]. Table 2 summarizes relevant

588 data for seven tests of NCBFs in which gusset-plate interface weld fracture was the first failure

589 mode. Equation 11 was calibrated from numerical simulation of these experiments, where the

590 analytical fracture rotation, θf, is the gusset-plate spring rotation corresponding to gusset-plate

591 interface weld fracture in the test. For practical purposes, gusset-plate rotation can be converted to

592 brace compressive deformation using the expression Lθ2/2, which is based upon small-angle

593 approximations and an idealized triangular brace deflected shape. An upper-bound fracture

594 rotation of 0.257 rad (or 3.3% brace compressive deformation) is recommended, which

595 corresponds to the fracture rotation prediction for 8tp elliptical clearance and DCRgpw of 0.75. Note

596 that this equation is calibrated from and intended for use with gusset plate interface welds which

31
597 do not meet demand critical weld toughness requirements. Higher fracture rotations would be

598 expected in SCBFs with adequate clearance and weld strength.

0.33
𝐿clear
𝜃f = 0.11 ( ) 𝐷𝐶𝑅gpw −0.57 ≤ 0.257 rad (11)
𝑡p

599

600 TABLE 2. Gusset plate interface weld fracture data from Sen et al.[8]
Specimen Connection DCRgpw Lclear/tp Analysis Failure mode Post-weld-fracture description
configuration fracture and
rotation, θf behavior type
(rad)
2O CSP-W 1.4 3.8 0.153 L3 / Type A Near total loss of lateral resistance
2C-B CSP-W+ 1.4 3.8 0.143 L1 / Type C Bolt-hole elongation → Brace fracture (L2)
3E SSP-B 1.6 1.2 0.0915 L2 / Type B Simultaneous fracture of shear plate bolts
4O S2L-B 1.0 2.2 0.155 L1 / Type C Angle yielding → Brace fracture (L2)
6O CSP-B 1.0 2.2 0.142 L1 / Type C Bolt-hole elongation → Bolt-hole tearout
(L2)
7O SSP-B 1.2 2.3 0.140 L1 / Type C Bolt-hole elongation → Bolt fracture (L2)
8O END 1.6 4.1 0.141 L2 / Type B Weld fracture on both gusset interfaces
CSP = Continuous Shear Plate, SSP = Split Shear Plate, S2L = Split Double Angle, END = End Plate, W = Welded to Gusset and
Beam, B = Bolted to Gusset and Beam
+
Connection reinforced with bolts along shear plate and brace filled with concrete as a retrofit
601

602 The behavior type for CBFs with gusset-plate interface weld fracture as the initial failure

603 mode depends upon the presence of yielding mechanisms in the remaining connection

604 components. The post-weld-fracture behavior of specimens with gusset-plate interface weld

605 fracture is described in Table 2. With the exception of Specimen 2O, the gusset-to-beam weld

606 fractured first in the specimens and thus the characteristics of the gusset-to-column connection

607 determined the ultimate behavior. The welded gusset-to-end-plate connection tested (Specimen

608 8O) did not have yielding mechanisms in the adjoining end plates, and this weld fractured

609 essentially simultaneously with the gusset-to-beam weld (Type B). Connections with bolted shear

610 plates (Specimens 2C-B, 3E, 6O, and 7O) either had premature bolt fracture nearly simultaneously

611 with the gusset-to-beam weld (Type B) or ductile bolt-hole elongation that led to a delayed failure

612 mode (Type C), depending on the relative strength of the bolt bearing versus bolt shear limit sates.

32
613 A similar phenomenon is possible with bolted double-angle connections (Specimen 4O), where

614 yielding of the angles can also provide significant deformation capacity if they are thin enough to

615 induce prying action in the bolts and preclude bolt fracture (Type C).

616 Gusset-plate interface weld fracture results in reduced or total loss of load in the brace (L1

617 or L2 failure mode, respectively). Both conditions can be simulated in OpenSees using the full

618 combination of springs in Fig. 3c. Weld fracture can be triggered using the MaxStrainRange

619 material wrapper on the rotational spring used to simulate flexural yielding of the gusset plate;

620 similar to the brace, fracture is not triggered until the load direction reverses. Note that the rotation

621 range is approximately equal to the rotation in one direction since rotation under brace tension is

622 near zero. Using a rotation range criterion here is advantageous since the rotation direction does

623 not need to be specified. A pair of translational springs is used to control post-fracture response.

624 Upon weld fracture, the rigid spring element is removed by the MaxStrainRange material. The

625 severity of weld fracture is a modeling decision determined by the presence of secondary yielding

626 mechanisms possible in the connection. If weld fracture results in an L2 failure-mode type, the

627 remaining spring is elastic and flexible, allowing essentially free translation of the brace end and

628 no force development. If weld fracture results in an L1 failure-mode type, the remaining spring is

629 nonlinear with force-deformation behavior representative of the remaining connection

630 components. An analogous pair of rotational springs are used at the beam-column connection to

631 simulate the fixity lost if the gusset plate disconnects from the beam. This rigid spring is also

632 removed by the MaxStrainRange material, and an appropriate moment-rotation behavior can be

633 defined for the remaining spring.

634 Figure 11 shows how the model can be implemented to simulate gusset-plate interface weld

635 fracture exhibiting L2-type failure mode, which was observed in Specimen 8O, as described in

33
636 Sen et al. [8] and in Table 2. In the experiment, both gusset-plate interface welds fractured

637 simultaneously, completely disconnecting the gusset plate from the frame. The base model in Fig.

638 11a does not implement a fracture criterion based upon gusset plate rotation, and the deformation

639 capacity is overpredicted since the failure mode is delayed brace fracture. When the proposed

640 model is implemented with a θf of 0.141 radians and no secondary yielding mechanism, the

641 behavior of the frame both before and after gusset plate interface weld fracture (GPW) is well

642 predicted (see Fig. 11b). Figure 11c shows the moment-rotation behavior of the gusset plate, which

643 confirms that the gusset plate rotation is near zero in one direction and therefore the

644 MaxStrainRange modeled can be used.

645 Use of a post-fracture translational spring to simulate the force-deformation behavior of a

646 bolted-bolted double-angle connection on the vertical gusset plate interface is shown in Fig. 12

647 using Specimen 4O from Sen et al. [8]. The gusset-to-beam interface weld fractured in this

648 experiment but local yielding of the gusset-to-column angles provided a secondary load path,

649 resulting in significantly increased strength and deformation capacity. The secondary yielding

650 translational spring was calibrated with the Steel02 material model using a yield force of 710 kN,

651 elastic stiffness of 370 kN/mm, kinematic hardening ratio of 3.0%, R0 of 15, cR1 of 0.925, and cR2

652 of 0.15, and its force-deformation behavior is shown in Fig. 12c. A base model which includes

653 gusset-plate interface weld fracture but no secondary yielding mechanism is shown in Fig. 12a.

654 Here, weld fracture is incorrectly modeled as an L2 failure-mode type, and the post-fracture

655 resistance is significantly lower than in the experiment. When the nonlinear spring behavior in Fig.

656 12c is introduced, the response in Fig. 12b is obtained, where the larger post-fracture resistance

657 matches the experiment well and the L2 failure mode is delayed brace fracture.

658

34
659

660 FIGURE 11. Experimental and numerical frame response of Specimen 8O (a) without and (b)

661 with the gusset plate interface weld fracture model; (c) response of corresponding gusset plate

662 rotational spring.

663

664

665 FIGURE 12. Experimental and numerical frame response of Specimen 4O (a) without and (b)

666 with the gusset plate post-fracture model; (c) response of corresponding gusset plate post-fracture

667 translational spring.

668 5. FRAMING MEMBER SIMULATION

669 At large lateral deformations, the beams, columns, and/or beam-column connections in CBFs are

670 expected to yield, and consideration of the strength and stiffness provided by these components

671 can have a large effect on seismic response [53]. In US construction, beams and columns are most

35
672 often wide flange sections and beam-to-column connections that are not adjacent to gusset plates

673 are often bolted shear plates. In SCBFs, ductility of the beams and columns is ensured by limiting

674 cross-sectional width-to-thickness ratios (b/t and h/t) to meet the highly ductile criteria in the

675 Seismic Provisions [1]. Stability of these members is ensured through capacity design based on

676 the expected brace forces and lateral bracing requirements for beams in chevron-configuration

677 frames. NCBFs, in comparison, may have beams and columns which are more susceptible to local

678 buckling and columns which may buckle globally prior to brace yielding. The following

679 recommendations address modeling to simulate yielding and buckling of these components.

680 5.1 Beam-to-Column Connections

681 Beam-to-column connection stiffness and strength are often neglected when modeling of CBFs in

682 design practice, but it plays an important role in developing lateral resistance through frame action.

683 Hsiao et al. [26] investigated different modeling approaches for frames, including models with

684 pinned, fully restrained, and realistic connections at the brace-beam-column intersection. The latter

685 model is shown in Fig. 3b, where the gusset plate is modeled using a nonlinear rotational spring

686 and the brace, beam, and column are offset from the work point to account for the stiffening

687 resulting from the gusset plate. The beam-to-column connection is fully restrained in this scenario

688 because the gusset plate acts like a haunch to restrain relative rotation of the beam and column.

689 This modeling approach was developed for SCBFs and remains applicable for NCBFs

690 unless gusset-plate interface weld fracture occurs (either L1 or L2 failure-mode types). For

691 example, in Specimen 8O in Fig. 11b, the gusset-plate interface weld fractures completely,

692 disconnecting the plate from the frame. The gusset plate therefore no longer restrains rotation of

693 the beam-to-column connection, though the end plate still provides some rotational restraint. A

694 similar effect occurs in Specimen 4O shown in Fig. 12b, but the retained beam-to-column

36
695 connection is a bolted-bolted double angle, which provides considerably less restraint. The

696 modeling approach in Fig. 3c can be employed to model these post-fracture scenarios using stiff

697 elastic and flexible nonlinear elements in parallel. Upon gusset-plate interface weld fracture, the

698 very stiff elements are removed from the analysis but the nonlinear elements, which simulate the

699 post-fracture condition of the frame, are retained. This element removal procedure was

700 accomplished using the MaxStrainRange material by inputting the tags of elements which are

701 marked for removal when the fracture criterion is met.

702 The behavior of beam-to-column connections that are not adjacent to gusset plates,

703 including connections part of the gravity framing in buildings, also greatly contributes to the lateral

704 resistance and can help prevent collapse once the braces or brace connections have fractured (i.e.,

705 an L2 failure-mode type has occurred) [53]. The flexural and rotational capacity of these

706 connections can be considered with nonlinear rotational springs. Liu and Astaneh-Asl [34] provide

707 guidance for such a modeling approach for bolted shear plates which are applicable for both new

708 and NCBF-era construction.

709 5.2 Yielding-Beam Mechanism in the Chevron-Configured CBFs

710 Recent experimental research on SCBFs [45], existing and retrofitted NCBFs [23,24] and “R = 3”

711 CBFs used in regions with low seismic hazard [25] has demonstrated the potential benefits of a

712 plastic mechanism in which beam yielding precludes brace yielding in the chevron configuration.

713 These tests show that beam yielding does not compromise and may even enhance CBF ductility.

714 Further, after the first brace or brace connection fractures, the yielding beam mechanism can

715 provide significant reserve capacity as the beam acts as long link in an EBF. This beam yielding

716 mechanism is not currently permitted in SCBF design but could be leveraged in retrofit of NCBFs.

717 The proposed recommendations from Hsaio et al. [26] for modeling beams in SCBFs (Fig. 3b)

37
718 were employed to simulate this behavior, where the beam is modeled with one force-based beam-

719 column element (five integration points per element) between work points as in Fig. 4f. Since the

720 beam is modeled as a fiber section, yielding can occur under the combined flexural and axial load

721 induced by the braces.

722 5.3 Column Buckling

723 Columns in CBFs significantly contribute to system seismic response through frame action. In

724 SCBFs, the columns are anticipated to yield under combined axial-flexural demands to

725 accommodate drift demands and are designed to prevent buckling induced by the accumulated

726 brace demands along the building height. SCBF columns are not designed for drift-induced

727 flexural demands but are required to satisfy highly ductile local slenderness requirements to ensure

728 ductility. Since buckling is not a concern with SCBF columns, they can be simulated using a single

729 force-based nonlinear beam-column element at each story with at least five integration points along

730 the height. Fiber-based approaches are recommended in favor of spring-based lumped plasticity

731 formulations to account for axial-flexural interaction. While spring-based approaches can be used

732 to simulate cyclic deterioration effects [29,54], SCBF columns meet highly ductile requirements

733 and thus are expected to have limited deterioration.

734 When the SCBF column design criteria are not met, as with NCBF columns, alternative

735 modeling approaches must be considered. NCBF column design was based upon axial loads

736 induced from the vertical gravity and lateral seismic forces. NCBF columns were not designed to

737 resist expected brace strengths and are potentially susceptible to column buckling if the brace axial

738 force exceeds its design force. A modeling approach similar to that used for brace buckling is

739 recommended for columns as shown in Fig. 13a. Here, the column is discretized with 12, equally

740 spaced nonlinear beam-column elements (either DBEs or FBEs can be used since fracture is not

38
741 modeled) with an initial imperfection of L/1000, which is consistent with the maximum out-of-

742 straightness allowed by the AISC Code of Standard Practice for Steel Buildings and Bridges [55].

743 This initial imperfection varies from that of the brace, L/500, where the objective was to match

744 strength to the AISC buckling curve. However, imperfection of this magnitude in a column would

745 be undesirable, since this would amplify axial shortening deformations imposed by gravity loads.

746 Each beam-column element has a fiber-based cross section at each of their 5 integration points. As

747 with other steel components considered in this paper, the Steel02 material model is used.

748 The recommended number of elements was determined from a study using Test 2 from

749 Lamarche and Tremblay [56], which investigated the compressive response of a W12×87 column

750 under cyclic loading. The response numerical response of the column using 12 force-based beam-

751 column elements is shown in Figs. 13b and 13c. An initial imperfection of L/2000 was used to

752 match these test results, but such small imperfection should not necessarily be assumed in practice.

753 Figure 14 shows the error in axial force and midheight deflection of this column using different

754 element types and discretizations. The error plotted in Fig. 14 is the sum of the errors at each load

755 reversal point in Figs. 13b and 13c. It can be seen that using more than 12 elements for either the

756 displacement-based or force-based formulations provides only a marginal increase in accuracy

757 based upon the measured response. If fewer than 12 elements are desired to reduce computational

758 cost, FBEs are recommended for use in OpenSees.

759

39
760

761 FIGURE 13. (a) Schematic of OpenSees column model and validation of numerically simulated

762 (b) axial force-axial deformation and (c) axial force-midheight deflection responses.

763

764

765 FIGURE 14. Convergence of (a) axial force error and (b) midheight deflection error in numerical

766 simulation of Test 2 from Lamarche and Tremblay [56].

767 6. COMPUTATIONAL MODEL VALIDATION

768 The proposed computational modeling recommendations are validated using two experiments of

769 two-story chevron CBFs tested at the National Center for Research on Earthquake Engineering in

770 Taiwan as part of a research program investigating existing and retrofitted NCBF behavior [23].

40
771 These specimens, TNCBF1-N-HSS and TNCBF1-R-HSS, were introduced in Figs. 1b and 1c and

772 have been selected since they demonstrated complex nonlinear behavior that would be atypical for

773 SCBFs. Their experimental setup is shown in Fig. 4e. Further details of the specimens are reported

774 in Sen et al. [23].

775 TNCBF1-N-HSS had HSS7×7×1/4 braces on both stories, which a local slenderness ratio

776 which is twice that allowed for SCBFs in the Seismic Provisions [1]. Thus, premature brace

777 fracture was an expected failure mode. The brace ends were slotted and welded to the gusset plates

778 and had gusset plate yielding and brace-to-gusset weld fracture, and brace net-section rupture

779 DCRs of 0.95, 0.96, and 1.3, respectively, for the first-story components. The beam was a W16×45

780 section with welded shear plate beam-to-column connections. Based upon the development of Pt

781 and 0.3Pc in the first-story braces (the expected brace forces in tension and compression,

782 respectively) required by the Seismic Provisions [1] and assuming simply supported beam ends,

783 the beam had an axial-flexural interaction value of 2.1. The retrofitted specimen, TNCBF1-R-HSS,

784 had SCBF-compliant HSS5×5×3/8 braces on the first story, a gusset-plate yielding DCR of 0.88,

785 and a beam axial-flexural interaction value of 2.3. The brace-to-gusset weld and net-section

786 fracture vulnerabilities were eliminated in the repair, since these connections were replaced. Both

787 specimens had concrete slabs on metal decks but did not have shear connectors to promote

788 composite action. The second story of each specimen was designed to be intentionally strong in

789 order to deliver the actuator load into the frame; consequently, the braces did not buckle on the

790 second story of these tests.

791 The specimens were computationally simulated following the proposed modeling

792 approaches, with the exception that the second-story braces were not given initial imperfection in

793 order to prevent buckling and facilitate comparison of the results. In the test frames, the upper slab

41
794 and beams were stiffer and stronger in the second story to permit load transfer from the actuators

795 to the frame without yielding or damage. Figure 4f shows OpenSees schematic models for the

796 frames. The gusset-plate axial yielding and brace-to-gusset weld fracture mechanisms were

797 simulated with translational springs in series at the brace ends. Brace fracture was simulated using

798 the MSR method with the MSRf prediction model given in Eq. 2 which accounts for load history

799 effects known to be important for braces in the chevron configuration. The first-story beam ends

800 were connected to the columns with welded shear plates, which were modeled as nonlinear

801 rotational springs with the Steel02 material behavior. The yield moment and elastic stiffness of the

802 springs were based on the beam web properties. The frames were loaded to larger deformations in

803 the simulations than in the tests to attempt to observe the experimental specimens’ failure modes.

804 Figures 15a through 15c plot the second- and first-story hysteretic responses and the first-

805 story beam deflection for the specimen. Figure 16a compares the experimental and numerical

806 sequence of yielding and failure for TNCBF1-N-HSS based on first-story drift range. The dashed

807 lines connect common events which occurred in both the experiment and numerical simulation.

808 The proposed modeling recommendations result in good agreement between the experimental and

809 numerical responses, but several discrepancies remain. The story-drift ranges at buckling and

810 fracture of the first brace are very well predicted and result in the long-link EBF mechanism

811 observed in the experiment. However, the rapid degradation of the brace resistance in compression

812 does not occur in OpenSees since the fiber section does not simulate local deformation. This

813 phenomenon is especially apparent in braces with high local slenderness ratios, like the

814 HSS7×7×1/4 in the specimen. However, it is important to note that local cupping of braces with

815 high b/t leads to brace fracture very quickly, which is achieved with the proposed MSRf, disp fracture

816 model. Thus, not simulating the immediate deterioration following buckling due to local cupping

42
817 is unlikely to significantly change the predicted NCBF performance using the proposed modeling

818 framework. Comparing the beam displacement-story drift plot in Fig. 15c, the downward beam

819 deflection is overestimated in the numerical model, which indicates that some amount of

820 composite action was developed in the beam in the experiment. Finally, the brace-to-gusset

821 connection failure mode (BRW in Fig. 15c) did not occur in the numerical model, as the beam was

822 too weak to develop demand in the brace-to-gusset weld translational spring necessary to induce

823 fracture.

824 Despite minor discrepancies, the above indicates that the proposed modeling is appropriate

825 for NCBFs and retrofitted NCBFs and is capable of reproducing the key behavior modes observed

826 in large-scale experiments.

827 7. SUMMARY AND CONCLUSIONS

828 New SCBFs and vintage NCBFs are prevalent in regions with high seismic hazard in the US.

829 Experimental research conducted since the early 2000s has significantly advanced the

830 understanding of their seismic behavior and spurred the development of sophisticated modeling

831 approaches for SCBFs in OpenSees. However, many NCBFs remain in service and may require

832 seismic retrofit to ensure occupant safety and mitigate potential economic losses for these

833 structures. These older CBFs have complex yielding and failure hierarchies which differ from

834 SCBFs due to a myriad of component and system deficiencies. Other CBF variants, including

835 OCBFs and “R = 3” CBFs are more common in regions with lower seismicity but, similar to

836 NCBFs, may have premature brace and connection failure modes. To further investigate the effects

837 of these component behaviors and provide a framework to study system seismic performance,

838 modeling approaches have been developed in OpenSees to simulate:

839  Fracture of rectangular HSS braces considering load direction bias and concrete in-fill,

43
840  Axial yielding of gusset plates,

841  Fracture of brace-to-gusset plate welds, and

842  Fracture of gusset plate interface welds with and without secondary yielding mechanisms.

843 Fracture phenomena are modeled in OpenSees using the MaxStrainRange material wrapper

844 that is revised from work by Hsiao et al. [28]. The material wrapper provides optional support for:

845  Tracking global deformation history between two nodes for consideration of load history

846 effects in braces (Eq. 2),

847  Specifying a lower-bound MSRf,conc for modeling fracture of concrete-filled braces (Eq.

848 3), and

849  Removing elements from the analysis when the fracture criterion is reached for use with

850 gusset plate interface weld fracture.

851 The MaxStrainRange can be considered a general-purpose material wrapper for simulating fracture

852 in general, not just fracture of a brace.

853 The efficacy of the modeling recommendations was demonstrated using two multi-story

854 NCBF experiments. The models employed the recommendations in this paper for the lower story

855 where yielding, damage, and component fracture occured. The global force-deflection behavior

856 and sequence of yielding and failure were predicted with reasonable accuracy, validating the

857 modeling approaches for existing and retrofitted NCBFs. However, the complex brace-to-gusset

858 connection failure mode of the existing NCBF specimen was not fully realized in the simulation.

859 Further, compressive resistance degradation of the brace due to local cupping deformation was not

860 simulated, which is a well-known limitation of fiber-based beam-column element formulations.

861 Modeling of these phenomena is a subject for potential future work, though it is noted that local

44
862 cupping in the brace leads quickly to brace fracture and thus may not have a significant impact on

863 seismic performance.

864

865

866 FIGURE 15. Comparison of experimental and numerical responses of existing and retrofitted

867 NCBF specimens.

45
868

869 FIGURE 16. Sequence of yielding and failure observed in experiments and numerical simulations

870 for first-story of (a) existing and (b) retrofitted NCBF specimens.

871

872 The modeling approaches presented here enable system-level analysis of existing and

873 retrofitted NCBFs but are also applicable for SCBFs, OCBFs, and “R = 3” CBFs. Further research

874 by the authors will utilize the modeling approaches proposed in this paper to investigate the

875 system-level seismic performance implications for existing and retrofitted NCBF building suites,

876 provide retrofit guidance for practicing engineers, and develop alternative simplified modeling

877 approaches suitable for practicing engineers.

878 Other modeling issues remain for concentrically braced frames and present opportunities

879 for future advancements. This research has developed models appropriate for simulating failure

880 modes commonly observed in large-scale experiments and in post-earthquake reconnaissance;

881 still, others have been documented in buildings which have sustained earthquake damage (e.g.,

882 column base uplift or fracture [57]). Modeling of steel-concrete composite beams also introduces

46
883 further computational complexity beyond the scope of the present research, and the additional

884 initial strength and subsequent cyclic deterioration of composite action may affect system

885 performance.

886 8. ACKNOWLEDGMENTS

887 This material is based on work supported by the National Science Foundation (NSF) Network for

888 Earthquake Engineering Simulation under Grant Nos. CMMI-0619161 and CMMI-1208002, the

889 NSF Graduate Research Fellowship under Grant No. DGE-1256082, and the NSF East Asia and

890 Pacific Summer Institutes Fellowship under Grant No. OISE-1614277. The generous support of

891 the American Society of Civil Engineers through the O. H. Ammann Research Fellowship is also

892 appreciated. In addition, the authors acknowledge Prof. Dimitrios Lignos of École Polytechnique

893 Fédérale de Lausanne, who provided experimental data from several tests utilized in this work.

894 Any opinion, findings, and conclusions or recommendations expressed in this material are those

895 of the authors and do not necessarily reflect the views of the sponsoring agencies.

896 9. NOTATION

897 The following symbols are used in this paper:

898 Ag = gross area;

899 Bw = Whitmore width;

900 DCR = demand-to-capacity ratio;

901 E = elastic modulus;

902 FEXX = weld filler metal classification strength;

903 Fy = yield stress;

904 K = effective length factor;

47
905 L = end-to-end length;

906 Lavg = average gusset-plate length;

907 Lc = effective length;

908 Lclear = gusset-plate elliptical clearance;

909 Lw = gusset-to-brace weld length

910 MSRf = maximum strain range at fracture;

911 Py = yield force;

912 R = response modification factor;

913 Ry = ratio of expected-to-nominal yield stress;

914 b = HSS width, the clear distance between the webs less the inside corner radius on each side;

915 f’c = concrete compressive strength;

916 p = ratio of weld deformation to deformation at maximum stress;

917 r = radius of gyration;

918 t = HSS wall thickness;

919 tp = plate thickness;

920 wbr = gusset-to-brace weld leg size;

921 wp = gusset-plate interface weld leg size;

922 Δ = weld deformation;

923 Δm = weld deformation at maximum stress;

924 Δu = weld deformation at ultimate stress;

925 δc,max = absolute maximum compressive deformation;

926 δt,max = absolute maximum tensile deformation;

927 εco = concrete strain at peak compressive strength;

48
928 θ = gusset-plate rotation;

929 θf = gusset-plate rotation at interface-weld fracture;

930 θi = weld load angle.

931 REFERENCES

932 [1] AISC, Seismic Provisions for Structural Steel Buildings, in: ANSI/AISC 341-16, American Institute

933 of Steel Construction, Chicago, IL, 2017.

934 [2] A. Astaneh-Asl, S.C. Goel, R.D. Hanson, Earthquake-Resistant Design of Double-Angle Bracings,

935 AISC Eng. J. (1986) 133–147.

936 [3] D.E. Lehman, C.W. Roeder, D. Herman, S. Johnson, B. Kotulka, Improved Seismic Performance

937 of Gusset Plate Connections, J. Struct. Eng. 134 (2008) 890–901. doi:10.1061/(ASCE)0733-

938 9445(2008)134:6(890).

939 [4] C.W. Roeder, E.J. Lumpkin, D.E. Lehman, A balanced design procedure for special concentrically

940 braced frame connections, J. Constr. Steel Res. 67 (2011) 1760–1772.

941 doi:10.1016/j.jcsr.2011.04.016.

942 [5] AISC, Seismic Provisions for Structural Steel Buildings, in: ANSI/AISC 341-97, American Institute

943 of Steel Construction, Chicago, IL, 1997.

944 [6] ICBO, Uniform Building Code, International Conference of Building Officials, Whittier, CA, 1988.

945 [7] A.D. Sen, D. Sloat, R. Ballard, M.M. Johnson, C.W. Roeder, D.E. Lehman, J.W. Berman,

946 Experimental Evaluation of the Seismic Vulnerability of Braces and Connections in Older

947 Concentrically Braced Frames, J. Struct. Eng. 142 (2016) 04016052. doi:10.1061/(ASCE)ST.1943-

948 541X.0001507.

949 [8] A.D. Sen, M.A. Swatosh, R. Ballard, D. Sloat, M.M. Johnson, C.W. Roeder, D.E. Lehman, J.W.

950 Berman, Development and Evaluation of Seismic Retrofit Alternatives for Older Concentrically

951 Braced Frames, J. Struct. Eng. 143 (2017) 04016232. doi:10.1061/(ASCE)ST.1943-541X.0001738.

49
952 [9] ASCE, Seismic Evaluation and Retrofit of Existing Buildings, in: ASCE/SEI 41-17, American

953 Society of Civil Engineers, Reston, VA, 2017.

954 [10] C.-M. Uang, M. Bruneau, State-of-the-Art Review on Seismic Design of Steel Structures, J. Struct.

955 Eng. 144 (2018) 03118002. doi:10.1061/(ASCE)ST.1943-541X.0001973.

956 [11] R.G. Black, W.A.B. Wenger, E.P. Popov, Inelastic Buckling of Steel Struts under Cyclic Load

957 Reversals, in: UCB/EERC-80/40, Earthquake Engineering Reserach Center, Berkeley, CA, 1980.

958 [12] S. Lee, Seismic Behavior of Hollow and Concrete-Filled Square Tubular Bracing Members, in: Rep.

959 UMCE 87-11, Ann Arbor, MI, 1988.

960 [13] Z. Liu, S.C. Goel, Cyclic Load Behavior of Concrete‐Filled Tubular Braces, J. Struct. Eng. 114

961 (1988) 1488–1506. doi:10.1061/(ASCE)0733-9445(1988)114:7(1488).

962 [14] R. Tremblay, Inelastic seismic response of steel bracing members, J. Constr. Steel Res. 58 (2002)

963 665–701. doi:10.1016/S0143-974X(01)00104-3.

964 [15] C.W. Roeder, E.J. Lumpkin, D.E. Lehman, Seismic Performance Assessment of Concentrically

965 Braced Steel Frames, Earthq. Spectra. 28 (2012) 709–727. doi:10.1193/1.4000006.

966 [16] P. Uriz, S.A. Mahin, Toward Earthquake-Resistant Design of Concentrically Braced Steel-Frame

967 Structures, in: PEER 2008/08, Pacific Earthquake Engineering Research Center, Berkeley, CA,

968 2008.

969 [17] T. Okazaki, D.G. Lignos, T. Hikino, K. Kajiwara, Dynamic Response of a Chevron Concentrically

970 Braced Frame, J. Struct. Eng. 139 (2013) 515–525. doi:10.1061/(ASCE)ST.1943-541X.0000679.

971 [18] C.W. Roeder, D.E. Lehman, K. Clark, J. Powell, J.-H. Yoo, K.-C. Tsai, C.-H. Lin, C.-Y. Wei,

972 Influence of gusset plate connections and braces on the seismic performance of X-braced frames,

973 Earthq. Eng. Struct. Dyn. 40 (2011) 355–374. doi:10.1002/eqe.1024.

974 [19] E.J. Lumpkin, P.-C. Hsiao, C.W. Roeder, D.E. Lehman, C.-Y. Tsai, A.-C. Wu, C.-Y. Wei, K.-C.

975 Tsai, Investigation of the seismic response of three-story special concentrically braced frames, J.

976 Constr. Steel Res. 77 (2012) 131–144. doi:10.1016/j.jcsr.2012.04.003.

977 [20] D.A. Foutch, S.C. Goel, C.W. Roeder, Seismic Testing of Full‐Scale Steel Building—Part I, J.

50
 978 Struct. Eng. 113 (1987) 2111–2129. doi:10.1061/(ASCE)0733-9445(1987)113:11(2111).

979 [21] C.W. Roeder, Seismic Behavior of Concentrically Braced Frame, J. Struct. Eng. 115 (1989) 1837–

980 1856. doi:10.1061/(ASCE)0733-9445(1989)115:8(1837).

981 [22] T. Fukuta, I. Nishiyama, H. Yamanouchi, B. Kato, Seismic Performance of Steel Frames with

982 Inverted V Braces, J. Struct. Eng. 115 (1989) 2016–2028. doi:10.1061/(ASCE)0733-

983 9445(1989)115:8(2016).

984 [23] A.D. Sen, C.W. Roeder, J.W. Berman, D.E. Lehman, C.-H. Li, A.-C. Wu, K.-C. Tsai, Experimental

985 Investigation of Chevron Concentrically Braced Frames with Yielding Beams, J. Struct. Eng. 142

986 (2016) 04016123. doi:10.1061/(ASCE)ST.1943-541X.0001597.

987 [24] B.G. Simpson, S.A. Mahin, Experimental and Numerical Evaluation of Older Chevron

988 Concentrically Braced Frames with Hollow and Concrete-Filled Braces, J. Struct. Eng. 144 (2018)

989 04018007. doi:10.1061/(ASCE)ST.1943-541X.0001988.

990 [25] C.R. Bradley, L.A. Fahnestock, E.M. Hines, J.G. Sizemore, Full-Scale Cyclic Testing of Low-

991 Ductility Concentrically Braced Frames, J. Struct. Eng. 143 (2017) 04017029.

992 doi:10.1061/(ASCE)ST.1943-541X.0001760.

993 [26] P.-C. Hsiao, D.E. Lehman, C.W. Roeder, Improved analytical model for special concentrically

994 braced frames, J. Constr. Steel Res. 73 (2012) 80–94. doi:10.1016/j.jcsr.2012.01.010.

995 [27] P. Uriz, S.A. Mahin, Seismic Vulnerability Assessment of Concentrically Braced Steel Frames,

996 Steel Struct. 4 (2004) 239–248.

997 [28] P.-C. Hsiao, D.E. Lehman, C.W. Roeder, A model to simulate special concentrically braced frames

998 beyond brace fracture, Earthq. Eng. Struct. Dyn. 42 (2013) 183–200. doi:10.1002/eqe.2202.

999 [29] E. Karamanci, D.G. Lignos, Computational Approach for Collapse Assessment of Concentrically

1000 Braced Frames in Seismic Regions, J. Struct. Eng. 140 (2014) A4014019.

1001 doi:10.1061/(ASCE)ST.1943-541X.0001011.

1002 [30] L. Tirca, L. Chen, Numerical Simulation of Inelastic Cyclic Response of HSS Braces upon Fracture,

1003 Adv. Steel Constr. 10 (2014) 442–462.

51
1004 [31] F. McKenna, M.H. Scott, G.L. Fenves, Nonlinear Finite-Element Analysis Software Architecture

1005 Using Object Composition, J. Comput. Civ. Eng. 24 (2010) 95–107. doi:10.1061/(ASCE)CP.1943-

1006 5487.0000002.

1007 [32] I.R. Gunnarsson, Numerical performance evaluation of braced frame systems, University of

1008 Washington, 2004.

1009 [33] AISC, Specification for Structural Steel Buildings, in: ANSI/AISC 360-16, American Institute of

1010 Steel Construction, Chicago, IL, 2017.

1011 [34] J. Liu, A. Astaneh-Asl, Moment–Rotation Parameters for Composite Shear Tab Connections, J.

1012 Struct. Eng. 130 (2004) 1371–1380. doi:10.1061/(ASCE)0733-9445(2004)130:9(1371).

1013 [35] M. D’Aniello, G. La Manna Ambrosino, F. Portioli, R. Landolfo, The influence of out-of-

1014 straightness imperfection in physical theory models of bracing members on seismic performance

1015 assessment of concentric braced structures, Struct. Des. Tall Spec. Build. 24 (2015) 176–197.

1016 doi:10.1002/tal.1160.

1017 [36] M. Menegotto, P.E. Pinto, Method of Analysis for Cyclically Loaded R. C. Plane Frames Including

1018 Changes in Geometry and Non-Elastic Behavior of Elements under Combined Normal Force and

1019 Bending, in: Proc. IABSE Symp. Resist. Ultim. Deform. Struct. Acted by Well Defin. Loads,

1020 Lisbon, 1973: pp. 15–22. doi:http://dx.doi.org/10.5169/seals-13741.

1021 [37] F.C. Filippou, E.P. Popov, V. V. Bertero, Effects of Bond Deterioration on Hysteretic Behavior of

1022 Reinforced Concrete Joints, Berkeley, CA, 1983.

1023 [38] Y. Balazadeh-Minouei, S. Koboevic, R. Tremblay, Seismic evaluation of a steel braced frame using

1024 NBCC and ASCE 41, J. Constr. Steel Res. 135 (2017) 110–124. doi:10.1016/j.jcsr.2017.03.017.

1025 [39] Y. Balazadeh-Minouei, S. Koboevic, R. Tremblay, Seismic Assessment of Existing Steel Chevron

1026 Braced Frames, J. Struct. Eng. 144 (2018) 04018046. doi:10.1061/(ASCE)ST.1943-541X.0002037.

1027 [40] D.C. Rai, S.C. Goel, Seismic evaluation and upgrading of chevron braced frames, J. Constr. Steel

1028 Res. 59 (2003) 971–994. doi:10.1016/S0143-974X(03)00006-3.

1029 [41] D.C. Rai, S.C. Goel, J. Firmansjah, A General Purpose Computer Program for Nonlinear Structural

52
1030 Analysis, in: Rep. UMCEE 96-21, University of Michigan, Ann Arbor, MI, 1996.

1031 [42] X. Tang, S.C. Goel, Seismic Analysis and Design Considerations of Braced Steel Structures, in:

1032 Rep. UMCE 87-4, University of Michigan, Ann Arbor, MI, 1987.

1033 [43] P.-C. Hsiao, D.E. Lehman, J.W. Berman, C.W. Roeder, J. Powell, Seismic Vulnerability of Older

1034 Braced Frames, J. Perform. Constr. Facil. 28 (2014) 108–120. doi:10.1061/(ASCE)CF.1943-

1035 5509.0000394.

1036 [44] J.G. Sizemore, L.A. Fahnestock, E.M. Hines, C.R. Bradley, Parametric Study of Low-Ductility

1037 Concentrically Braced Frames under Cyclic Static Loading, J. Struct. Eng. 143 (2017) 04017032.

1038 doi:10.1061/(ASCE)ST.1943-541X.0001761.

1039 [45] C. Terpstra, Impact of beam strength on seismic performance of chevron concentrically braced

1040 frames, University of Washington, 2017. http://hdl.handle.net/1773/40001.

1041 [46] J. Richard, Étude du Comportemenet Sismique de Bâtiments Industriels avec Systèmes de

1042 Contreventement en Acier de Faible Ductilitè, Ècole Polytechnique de Montrèal, 2009.

1043 [47] M.T. Stephens, D.E. Lehman, C.W. Roeder, Seismic performance modeling of concrete-filled steel

1044 tube bridges: Tools and case study, Eng. Struct. 165 (2018) 88–105.

1045 doi:10.1016/j.engstruct.2018.03.019.

1046 [48] G.A. Chang, J.B. Mander, Seismic energy based fatigue damage analysis of bridge columns: Part I-

1047 -Evaluation of seismic capacity, in: Rep. NCEER-94-0006, National Center for Earthquake

1048 Engineering Research, Buffalo, NY, 1994.

1049 [49] Z. Liu, S.C. Goel, Investigation of Concrete-Filled Steel Tubes under Cyclic Bending and Buckling,

1050 Ann Arbor, MI, 1987.

1051 [50] AISC, Steel Construction Manual, American Institute of Steel Construction, Chicago, IL, 2017.

1052 [51] D.F. Lesik, D.J.L. Kennedy, Ultimate strength of fillet welded connections loaded in plane, Can. J.

1053 Civ. Eng. 17 (1990) 55–67. doi:10.1139/l90-008.

1054 [52] C.J. Carter, L.S. Muir, B. Dowswell, Establishing and Developing the Weak-Axis Strength of Plates

1055 Subjected to Applied Loads, AISC Eng. J. (2016) 147–157.

53
1056 [53] P.-C. Hsiao, D.E. Lehman, C.W. Roeder, Evaluation of the response modification coefficient and

1057 collapse potential of special concentrically braced frames, Earthq. Eng. Struct. Dyn. 42 (2013)

1058 1547–1564. doi:10.1002/eqe.2286.

1059 [54] D.G. Lignos, H. Krawinkler, Deterioration Modeling of Steel Components in Support of Collapse

1060 Prediction of Steel Moment Frames under Earthquake Loading, J. Struct. Eng. 137 (2011) 1291–

1061 1302. doi:10.1061/(ASCE)ST.1943-541X.0000376.

1062 [55] AISC, Code of Standard Practice for Steel Buildings and Bridges, in: ANSI/AISC 303-16, Chicago,

1063 IL, 2016.

1064 [56] C.-P. Lamarche, R. Tremblay, Seismically induced cyclic buckling of steel columns including

1065 residual-stress and strain-rate effects, J. Constr. Steel Res. 67 (2011) 1401–1410.

1066 doi:10.1016/j.jcsr.2010.10.008.

1067 [57] R. Tremblay, A. Filiatrault, M. Bruneau, M. Nakashima, H.G.L. Prion, R. DeVall, Seismic design

1068 of steel buildings: lessons from the 1995 Hyogo-ken Nanbu earthquake, Can. J. Civ. Eng. 23 (1996)

1069 727–756. doi:10.1139/l96-885.

1070 [58] S.M. Johnson, Improved seismic performance of special concentrically braced frames, University

1071 of Washington, 2005.

1072 [59] D.J. Herman, Further improvements on an understanding of special concentrically braced frame

1073 systems, University of Washington, 2007.

1074 [60] B.A. Kotulka, Analysis for a design guide on gusset plates used in special concentrically braced

1075 frames, University of Washington, 2007.

1076 [61] J.A. Powell, Evaluation of special concentrically braced frames for improved seismic performance

1077 and constructability, University of Washington, 2010.

1078 [62] K.A. Clark, Experimental performance of multi-story X-braced SCBF systems, University of

1079 Washington, 2009.

1080 [63] E.J. Lumpkin, Enhanced seismic performance of multi-story special concentrically brace frames

1081 using a balanced design procedure, University of Washington, 2009.

54
1082 [64] B. V. Fell, A.M. Kanvinde, G.G. Deierlein, A.T. Myers, Experimental Investigation of Inelastic

1083 Cyclic Buckling and Fracture of Steel Braces, J. Struct. Eng. 135 (2009) 19–32.

1084 doi:10.1061/(ASCE)0733-9445(2009)135:1(19).

1085 [65] F. Yang, S.A. Mahin, Limiting net section failure in slotted HSS braces, Moraga, CA, 2005.

1086 [66] B. Shaback, T. Brown, Behaviour of square hollow structural steel braces with end connections

1087 under reversed cyclic axial loading, Can. J. Civ. Eng. 30 (2003) 745–753. doi:10.1139/l03-028.

1088 [67] S.-W. Han, W.T. Kim, D.A. Foutch, Seismic Behavior of HSS Bracing Members according to

1089 Width–Thickness Ratio under Symmetric Cyclic Loading, J. Struct. Eng. 133 (2007) 264–273.

1090 doi:10.1061/(ASCE)0733-9445(2007)133:2(264).

1091 [68] R. Tremblay, M.-H. Archambault, A. Filiatrault, Seismic Response of Concentrically Braced Steel

1092 Frames Made with Rectangular Hollow Bracing Members, J. Struct. Eng. 129 (2003) 1626–1636.

1093 doi:10.1061/(ASCE)0733-9445(2003)129:12(1626).

1094 [69] D.A. Sloat, Evaluation and retrofit of non-capacity designed braced frames, University of

1095 Washington, 2014.

1096 [70] M.M. Johnson, Seismic evaluation of bolted connections in non-seismic concentrically braced

1097 frames, University of Washington, 2014.

1098 [71] R. Ballard, Impact of connection type on performance of non-seismic concentrically braced frames,

1099 University of Washington, 2015.

1100 [72] M.A. Swatosh, Seismic evaluation and retrofit of concentrically braced frames, University of

1101 Washington, 2016.

1102 [73] A.D. Sen, Seismic performance of chevron concentrically braced frames with weak beams,

1103 University of Washington, 2014.

1104 [74] S.M. Ibarra, Experimental Investigation of Chevron Special Concentrically Braced Frames with a

1105 Yielding Beam Plastic Mechanism, University of Washington, 2018.

1106

55
1107 APPENDIX A: BRACE FRACTURE SIMULATION DATA

1108 TABLE A.1. Brace Fracture Model Calibration Data.

Reference Specimen Brace Grade b/t Lc/r E/Fy δc,max/δt,max MSRf,disp

Johnson [58] HSS2 HSS5X5X3/8 A500 11.3 79 414 1.26 0.0387
HSS3 HSS5X5X3/8 A500 11.3 82 414 1.84 0.0412
HSS4 HSS5X5X3/8 A500 11.3 81 396 1.58 0.0425
HSS5 HSS5X5X3/8 A500 11.3 82 396 1.80 0.0442
Herman [59] HSS6 HSS5X5X3/8 A500 11.3 81 448 1.65 0.0421
HSS7 HSS5X5X3/8 A500 11.3 72 448 1.99 0.0438
HSS8 HSS5X5X3/8 A500 11.3 83 448 1.06 0.0422
HSS9 HSS5X5X3/8 A500 11.3 79 448 1.79 0.0350
HSS10 HSS5X5X3/8 A500 11.3 79 441 1.25 0.0399
HSS11 HSS5X5X3/8 A500 11.3 69 441 1.32 0.0324
Kotulka [60] HSS12 HSS5X5X3/8 A500 11.3 71 441 2.15 0.0421
HSS13 HSS5X5X3/8 A500 11.3 79 441 1.03 0.0387
HSS14 HSS5X5X3/8 A500 11.3 82 441 0.88 0.0410
HSS15 HSS5X5X3/8 A500 11.3 84 441 1.08 0.0380
Powell [61] HSS17 HSS5X5X3/8 A500 11.3 82 441 1.20 0.0414
HSS24 HSS5X5X3/8 A500 11.3 82 448 1.18 0.0377
HSS25 HSS5X5X3/8 A500 11.3 69 448 2.72 0.0401
Clark [62] TCBF1-1 HSS125X125X9 A500 10.9 55 451 0.88 0.0683
TCBF1-3 HSS125X125X9 A500 10.9 70 451 0.88 0.0556
Lumpkin [63] TCBF2-1 HSS5X5X3/8 A500 11.3 66 435 1.66 0.0553
Fell et al. [64] Kanvinde-1 HSS4X4X1/4 A500 14.2 62 433 0.98 0.0509
Kanvinde-2 HSS4X4X1/4 A500 14.2 62 433 3.01 0.0555
Kanvinde-4 HSS4X4X3/8 A500 8.5 64 403 1.00 0.0694
Yang and Mahin [65] Yang-4 HSS6X6X3/8 A500 14.2 39 479 0.34 0.0527
Yang-5 HSS6X6X3/8 A500 14.2 39 479 0.96 0.0560
Uriz and Mahin [16] Patxi-SCBF-1 HSS6X6X3/8 A500 14.2 46 479 1.34 0.0590
Shaback and Brown [66] Shaback-1B RHS127X127X8.0 G40.21-350W 12.9 52 475 1.94 0.0552
Shaback-2A RHS152X152X8.0 G40.21-350W 16.0 52 452 1.56 0.0458
Shaback-2B RHS152X152X9.5 G40.21-350W 13.0 51 452 1.94 0.0491
Shaback-3A RHS127X127X6.4 G40.21-350W 16.8 63 434 2.89 0.0437
Shaback-3B RHS127X127X8.0 G40.21-350W 12.9 63 475 1.73 0.0410
Shaback-3C RHS127X127X9.5 G40.21-350W 10.4 60 434 2.79 0.0492
Shaback-4A RHS152X152X8.0 G40.21-350W 16.0 62 452 2.57 0.0407
Shaback-4B RHS152X152X9.5 G40.21-350W 13.0 58 452 3.02 0.0463
Han et al. [67] Han-S77-28 HSS100X100X3.2 SPSR400 28.3 71 507 0.98 0.0213
Lee [12] Lee-1 RHS5X5X0.188 A500 25.7 64 469 2.43 0.0283
Lee-2 RHS5X5X0.188 A500 25.7 32 465 2.53 0.0304
Lee-4 RHS4X4X0.125 A500 31.5 78 500 5.44 0.0330
Lee-5 RHS4X4X0.250 A500 14.2 80 392 11.73 0.0477
Lee-6 RHS4X4X0.250 A500 14.2 43 392 12.44 0.0550
Lee-7 RHS4X4X0.250 A500 14.2 43 392 4.92 0.0479
Lee-8a RHS5X5X0.188 A500 25.7 32 465 5.34 0.0440
Lee-9a RHS4X4X0.125 A500 31.5 41 500 7.67 0.0505
Lee-10a RHS4X4X0.125 A500 31.5 41 500 2.85 0.0500
Lee-11a RHS4X4X0.125 A500 31.5 41 509 13.93 0.0445
Lee-12a RHS4X4X0.250 A500 14.2 43 392 7.94 0.0680
Lee-13a RHS4X4X0.250 A500 14.2 43 509 7.67 0.0535
Liu [13] Liu-T633H HSS6X3X3/16 A500 31.5 48 537 1.70 0.0302
Liu-T633C6a HSS6X3X3/16 A500 31.5 48 537 4.11 0.0536
Liu-T633C4a HSS6X3X3/16 A500 31.5 48 483 2.79 0.0525
Liu-T424H HSS4X2X1/4 A500 14.2 80 537 3.11 0.0464
Liu-T424C6a HSS4X2X1/4 A500 14.2 80 537 3.04 0.0485
Liu-T422H HSS4X2X1/8 A500 31.5 75 537 3.50 0.0240
Tremblay [68] Tremblay-S3A RHS76X76X4.8 G40.21-350W 12.8 147 514 1.00 0.0333
Sloat [69] NCBF-2E HSS7X7X1/4 A500 27.0 59 450 2.16 0.0208
NCBF-2C-Ba HSS7X7X1/4 A500 27.0 59 450 1.86 0.0436
Johnson [70] NCBF-3E HSS6X6X1/4 A500 22.8 56 494 1.25 0.0237
NCBF-5O HSS5X5X3/8 A500 11.3 71 469 1.51 0.0423
Ballard [71] NCBF-8I HSS6X4X3/8 A500 14.2 73 459 1.27 0.0388
Swatosh [72] NCBF-4O-W HSS5X5X3/8 A500 11.3 81 503 1.19 0.0406
NCBF-8O-W HSS5X5X3/8 A500 11.3 80 430 1.63 0.0410
Sen [73] TNCBF1-N-HSS HSS7X7X1/4 A500 27.0 52 451 1.46 0.0325

56
 Reference Specimen Brace Grade b/t Lc/r E/Fy δc,max/δt,max MSRf,disp
TNCBF1-R-HSS HSS5X5X3/8 A500 11.3 73 456 8.51 0.0485
TNCBF2-D-HSS HSS5X5X3/8 A500 11.3 73 456 5.48 0.0484
Ibarra [74] Chevron-5 HSS4X4X5/16 A500 10.7 84 453 11.99 0.0462
Richard [46] RHS4 HSS10X10X1/2 A500 18.5 43 504 0.99 0.0389
RHS10 HSS10X10X3/8 A500 25.7 60 476 1.66 0.0234
RHS12 HSS10X10X3/8 A500 25.7 43 466 1.00 0.0265
RHS13 HSS10X10X5/16 A500 31.4 43 585 1.00 0.0245
aConcrete-filled brace
1109

57

View publication stats