Version of Record: https://www.sciencedirect.

com/science/article/pii/S0141029623012610
Manuscript_f5ed786ae0d795130583a3ad222aca86

1 NUMERICAL ASSESSMENT OF IN-PLANE BEHAVIOR OF MULTI-

2 PANEL CLT SHEAR WALLS FOR MODULAR STRUCTURES
3 Sujit Bhandari1, Erica C. Fischer2, Mariapaola Riggio3, Lech Muszynski4

4 Abstract

5 Ponderosa pine (Pinus ponderosa) harvested from restoration forests in the Pacific Northwest

6 region of the United States is being considered for use in cross laminated timber (CLT). A

7 prototype modular building was designed and constructed to access the feasibility of using such

8 CLT panels in emergency housing that can be rapidly assembled, disassembled, and reused. The

9 modular building uses narrow-width CLT panels connected with intra-modular connections (butt

10 joints with inclined screws) that develop the wall modules, which are connected to each other at

11 the corners with point-type inter-modular connections. To accurately predict the behavior of this

12 modular structure under seismic loading, it is necessary to be able to simulate the connections and

13 wall modules under lateral loading. These selected intra- and inter-modular connections in PP CLT

14 have been mechanically characterized in previous experiments. This paper presents the

15 methodology and results from benchmarking the selected in-plane CLT connections against the

16 experimental data in OpenSees. Further, the impacts of the selected connections on the in-plane

17 behavior of multi-panel CLT shear wall modules are also presented. The results show that the

18 Pinching4 material model in OpenSees, when benchmarked using the simulated annealing

19 algorithm, was able to successfully simulate the selected connections. Simulations of multi-panel

1
Graduate Research Assistant, Department of Wood Science and Technology, Oregon State University;
Corresponding Author: [email protected]
2
Assistant Professor, School of Civil and Construction Engineering, Oregon State University,
[email protected]
3
Associate Professor, Department of Wood Science and Technology, Oregon State University,
[email protected]
4
Professor, Department of Wood Science and Technology, Oregon State University,
[email protected]

© 2023 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
20 wall modules indicate that modules with closely spaced screws in the butt joint behave as a single

21 wall panel, while those with widely spaced screws behave as coupled walls.

22 1. Introduction

23 Cross Laminated Timber (CLT) is an engineered wood product constructed by gluing together

24 layers of solid-sawn lumber, with each layer oriented at right angles to the adjacent layer. CLT

25 was originally developed with the goal of using sideboards in high value applications [1,2].

26 However, commercially produced CLT is often constructed from structural grade feedstock, such

27 as Norway spruce (Picea abies), Douglas fir (Pseudotsuga menziesii), Scots Pine (Pinus sylvestris)

28 and European larch (Larix decidua) [3] which does not meet the goal of using sideboards in high

29 value applications. Therefore, attention has been given to using alternative, underutilized lumber

30 such as small-diameter Eucalyptus (Eucalyptus urophylla x E. grandis) [4], low-grade yellow-

31 poplar (Liriodendron tulipifera) [5], and small diameter logs of Norway spruce (Picea abies) and

32 Scots pine (Pinus sylvestris) [6]. Small diameter logs of Ponderosa pine (Pinus ponderosa, to be

33 referred to as PP), one of the underutilized resources, is produced in large quantities during forest

34 restoration projects in the western US produce aimed at mitigating wildfires. Jadedi et al. [7]

35 demonstrated a potential for the use of PP in cross laminated timber (CLT) even though the

36 mechanical characteristics of CLT fabricated from low grade PP lamstock are lower than E3 basic

37 CLT grade in PRG320 [8], the North American product standard for structural CLT. While this

38 low-grade CLT might not be suitable for the most demanding applications like tall buildings, it

39 has been proposed for low-raise rapidly deployable modular emergency housing units designed

40 for easy assembly, disassembly, and circular use in widespread crisis or disaster areas [9]. A

41 prototype modular building was developed to demonstrate the feasibility of multi-panel wall

42 modules consisting of narrow (1.2 m) PP CLT panels produced by a local manufacturer, butt joint
43 with inclined screws acting as the intra-modular connections [9]. These CLT panels joined together

44 with intra-modular connections are the standardized repetitive components known as modules.

45 The modules are connected using a point-type connection [10] installed at the corners of the wall

46 modules.

47 Two dimensional (2D) wall modules are formed by joining adjacent CLT panels in-plane usually

48 along their major strength axis [11]. These in-plane connections transfer shear loads between the

49 panels and provide structural integrity. Common CLT in-plane connections include spline

50 connections, half-lapped joints and butt joints with self-tapping screws [11,12]. In the prototype

51 modular building developed by Bhandari [9], the butt joint connections with inclined screws were

52 selected for the in-plane intra-modular connections. Consequently, the research presented in this

53 paper will be focused on this type of intra-modular connection.

54 The prototype building developed by Bhandari [9] used a proprietary point-type connection system

55 for the inter-modular and module-to-foundation connections, known as XRAD [13]. The system

56 installed at the corners of 2D modules has varying mechanical properties depending on the

57 direction of the loading. The capacity domain of the connection based on the previous experiments

58 [13,14] in eight major directions of loading is presented in Figure 1.

59

60 Figure 1: Capacity of the point-type connection in major loading directions based on experimental
61 results from [13,14]
62

63 On the scale of the building, the stiffness and ductility of the intra- and inter-modular connections

64 affect the stiffness, ductility, and damage mechanisms of the whole system in response to high

65 wind or seismic events by concentrating damage in either the connections or the CLT panels

66 [15,16]. In buildings designed for rapid deployment, disassembly, and circular use, concentrating

67 the damage in the connections rather than in the panels would be most desired. Understanding the

68 behavior of the multi-panel wall modules would be beneficial in analyzing and designing the

69 modular structure for rapid deployment and reuse.

70 1.1. Multi-panel wall behavior

71 A multi-panel CLT wall module subjected to lateral loading can behave in one of the following

72 ways: 1) as a series of coupled wall panels, 2) as a wall showing an intermediate behavior, or 3)

73 as a single wall segment as illustrated in Figure 2 [16,17]. That behavior is controlled by 1) the

74 stiffness ratio of the in-plane intra-modular connections to the corner connections, 2) the ratio of

75 horizontal to vertical load, and 3) the aspect ratio of the component panels. Relatively higher

76 stiffness of in-plane connections compared to the stiffness of the connections at the corners would

77 result in the multi-panel wall behaving as a single wall (Figure 2c). While relatively lower stiffness

78 would result in the multi-panel wall behaving as a coupled wall (Figure 2a).

79

80

81 Figure 2: Rocking behavior of multi-panel CLT wall modules subjected to lateral load: (a) coupled
82 panel (b) intermediate and (c) single wall [adapted from [17]]
83 Lukacs et al. [18] compared ten different simple analytical methods to calculate the strength and

84 stiffness of single-panel CLT shear walls connected at the base with hold-downs and angle

85 brackets to a set of experimental data. Such methods cannot be directly applied to multi-panel 2D

86 wall modules with in-plane intra-module connections which contribute both to the stiffness and

87 capacity of the multi-panel wall modules.

88 Three different analytical approaches have been proposed to predict the behavior of multi-panel

89 CLT wall modules: 1) linear model proposed by Casagrande et al. [17], 2) a bilinear elastic-
 90 perfectly plastic model proposed by Nolet et al. [16], and 3) five distinct models proposed by

91 Gavric et al. [19] based on trilinear force-deflection relationship of CLT connectors.

92 The linear approach proposed by Casagrande et al. [17] is based on the minimum total potential

93 energy principle and cannot predict the nonlinear behavior of multi-panel wall modules. In the

94 bilinear approach proposed by Nolet [16], the use of elastic-perfectly plastic behavior for the

95 connections neglects any post-peak behavior of the walls. The models by Casagrande et al. [17]

96 and Nolet et al. [16] assume the CLT component panels to be rigid elements and attribute all

97 internal deformations of multi-panel modules to connections. Both models neglected the sliding

98 effects of the wall. Gavric et al. [19] proposed a simplified trilinear force-deflection relationship

99 of connections to simulate the in-plane behavior of multi-panel walls with angle brackets and hold-

100 downs. The CLT panels were assumed to be elastic. This model was used to perform a pushover

101 analysis of multi-panel walls and compare the results to experimental data. The results of this

102 research suggested that the in-plane deformation in the CLT panels was almost negligible and that

103 the segmented CLT wall had a lower stiffness and strength capacity but a substantially higher

104 deformation capacity than a non-segmented wall with the same dimensions [19]. While these

105 analytical models can be used to predict the approximate behavior of a single multi-panel wall

106 module, they may not be suitable to capture the spatial interactions between wall modules in a

107 structure. Finite element models, unlike analytical models, can be scalable to predict the behavior

108 of three-dimensional structures using these multi-panel wall modules.

109 In addition to these simplified analytical formulations, numerical FE analyses have been used to

110 predict the cyclic behavior of multi-panel shear walls with hold-downs and angle brackets at the

111 base [20,21]. Izzi et al. [21] simulated CLT as 3D solid elements and the connections as two-node

112 hysteretic springs represented by piecewise linear force-displacement relationships in Abaqus.

113 Four different methods of analysis with the connections resisting shear and/or tension components

114 were investigated that provided varying levels of accuracy [21]. In the research by Shahnewaz et

115 al. [20], the component wall segments were modeled in OpenSees using orthotropic linear shell

116 elements for the CLT panels, and Pinching4 material for angle brackets, hold-down, half-lap, and

117 spline connections. The results of that FE analysis were compared to experimental data and

118 demonstrated that the model was able to simulate the hysteresis behavior of CLT shear walls [20].

119 However, in both of these papers, the hold-downs and angle brackets are represented by either

120 single or orthogonal spring elements to simulate the sliding and rocking behaviors. This modeling

121 approach using two spring elements does not sufficiently capture the distinct force-displacement

122 behavior of the point-type connector used in the prototype modular structure that has varying

123 mechanical behavior depending on the direction of the loading [22,23].

124 While previous researchers have proposed methods to assess the performance of multi-panel CLT

125 shear walls, there is still a methodological gap in simulation approaches to predict the behavior of

126 the multi-panel CLT wall modules consisting of point-type connections as inter-modular

127 connections as in the modular structure proposed in the current study. There is a need of examining

128 the impact of intra- and inter-modular CLT connections on the in-plane behavior of multi-panel

129 CLT shear wall modules for rapidly deployable and reusable modular systems using point-type

130 inter-modular connections.

131 1.2. Simulating Connections

132 Intra-modular in-plane connections, such as butt joints with screws, are often modeled as a pair of

133 orthogonal elements that resist tension and shear forces independently [20]. The behavior of inter-

134 modular point-type connections under monotonic loading has been successfully simulated either

135 as two (vertical and horizontal) linear link elements or as three (vertical, horizontal, and diagonal)
136 link elements with linear and multilinear characteristics [22,23]. The three-link model better

137 reflects the effect of loading direction on connection stiffness and better predicts the stiffness and

138 capacity of the connection [22]. However, there is limited research on the applicability of

139 multilinear links for the evaluation of the cyclic response of these connections.

140 Many finite element (FE) software packages allow defining of nonlinear force-displacement

141 behavior in material models for the in-plane connection elements. Two of the most utilized

142 material models for elements simulating timber connections are the SAWS [24] and Pinching4

143 [25] models within the OpenSees FEA package [26]. Both material models are capable of

144 simulating the pinching behavior and strength degradation often observed in the experimental data

145 from cyclic loading tests. The main difference between them is that the SAWS model simulates

146 the strength degradation with a combination of exponential function and a linear descent line [24]

147 while the Pinching4 model consists of four piecewise linear curves representing the envelope of

148 the force-displacement curves [25]. One important advantage of Pinching4 over SAWS is that it

149 allows additional parameters to reflect the intricacies of damage and failure mechanisms in the

150 connection which must consider both the damage in the base material (e.g. CLT) and yielding or

151 damage within the connection components (e.g. screws, plates) [27,28]. Another advantage is that

152 the Pinching4 model can simulate the asymmetrical behavior of connections [25] which is suitable

153 for simulating the behavior of inter-modular point-type connections used in the prototype modular

154 structure [9].

155 1.3. Benchmarking Connections

156 The ductility of structural components and systems and the ability of the structure to dissipate

157 energy during extreme loading events such as an earthquake can be assessed through nonlinear

158 structural analysis, which can be facilitated by FE simulations. However, to be reliable, these FE
159 methods require experimental data to benchmark the models against. The lack of suitable

160 experimental data on the performance of the selected intra- and inter-modular connections in PP

161 CLT has been limiting the development of numerical models for modular structures using PP CLT.

162 This gap has been addressed by a set of experimental data generated by the authors in a parallel

163 project [9] clearing the way for developing FE models for the analysis of the structural response

164 of modular structures incorporating multi-panel CLT wall modules.

165 Three benchmarking methodologies have been used to simulate the behavior of CLT connections:

166 1) reverse calibration, 2) least squares fitting, and 3) simulated annealing method. In the reverse

167 calibration procedure, outlined in the CLT Handbook [29], the parameters are iteratively adjusted

168 until an acceptable fit to the experimental data is achieved. This type of methodology has been

169 used to benchmark numerical models that simulate hold-down connections using SAWS and

170 Pinching4 material models in OpenSees [27]. Fitting the model parameters to experimental force-

171 displacement data using the least-squares method was successfully employed to benchmark SAWS

172 and Pinching4 models in OpenSees to simulate CLT wall angle brackets [28]. For benchmarking

173 the force-displacement behavior, the simulated annealing (SA) algorithm [30] can be used to

174 minimize the difference in energy or force between the results from numerical models and the

175 experiments. The SA algorithm was used with acceptable accuracy for benchmarking the

176 parameters of SAWS [31] and Pinching4 [32] models for angle brackets. Of these three methods,

177 the SA algorithm can optimize multiple parameters given the initial parameters, lower and upper

178 bounds, and a single optimization function. This can help find the best fit of multiple SAWS or

179 Pinching4 parameters at once with the least human intervention. Although the SA algorithm has

180 been used to benchmark single elements acting independently in shear and tension, their use in
181 three-link elements used for modeling point-type connection used as an inter-modular connection

182 in the modular structure has not been established before.

183 1.4. Scope and Objectives

184 This paper presents the numerical modeling of connections and PP CLT wall modules that will be

185 crucial in analyzing rapidly deployable modular structures. The aim of this study was to develop

186 a methodology to simulate multi-panel CLT wall modules under lateral loading using the

187 experimental data of the connections. The specific objectives of this paper are 1) to evaluate the

188 viability of using simulated annealing to benchmark the mechanical behavior of selected intra-

189 modular and inter-modular PP CLT connections, 2) to benchmark the mechanical behavior of

190 selected intra-modular and inter-modular PP CLT connections against experimental data, and 3)

191 to evaluate the impact of selected intra-modular (butt joints) and inter-modular (point-type

192 connections) CLT connections on the in-plane behavior of multi-panel CLT shear wall modules.

193 2. Materials and Methods

194 To accomplish these objectives, a system level FE model of multi-panel CLT shear wall modules

195 was developed with intra-module connections simulated by Pinching4 material models executed

196 in OpenSees FE software [26]. The connection models were benchmarked against experimental

197 data generated in a parallel study [9].

198 2.1. Reference experimental tests

199 Both the intra-modular and inter-modular connections are benchmarked against experimental

200 results of monotonic and cyclic in-plane tests generated by the authors [9]. A summary of that

201 work is provided here for reference. Details on the test setup and results are presented in [9].

202 Monotonic and cyclic shear and tension characteristics for PP CLT intra-modular butt joints with

203 45° inclined screws (Figure 3) were determined experimentally. The project included one
204 monotonic test in shear and tension, five cyclic tests in shear, and three cyclic tests in tension.

205 Tests were conducted on specimens consisting of two 100 mm thick, 3-ply PP CLT panels butt

206 jointed with VGS 11x125 screws (Figure 3). The maximum connection capacity in monotonic

207 tension and compression was 10.04 kN and 10.26 kN, respectively. Conversely, the maximum

208 capacities of the connections determined in cyclic tension and compression tests were 9.22 ± 0.45

209 kN and 10.64 ± 1.18 kN, respectively. The results from the monotonic and cyclic tests are

210 summarized in Figure 4.

211

212

213 Figure 3: Test setup for the intra-modular connection: a) front view of tension test, b) front view
214 of shear test and c) cross section of butt joint with 45° screws [9]
215
216 (a) (b)

217 Figure 4: Monotonic curve, backbone curves, and typical hysteretic curves of tests in intra-
218 modular connections: a) tension tests, and b) shear tests [9]
219

220 The monotonic and cyclic characteristics of the proprietary intra-modular point type connection

221 system (XRAD) installed on the corner of a 3-ply PP CLT specimen were determined with loads

222 applied in tension (45°) and compression (225°) (Figure 5a), tension-shear (0°/90°), and

223 compression-shear (180°/270°) configuration (Figure 5b). The load-displacement curves from

224 these monotonic and cyclic tests are presented in Figure 6a and Figure 6b respectively. The

225 maximum capacity of the connection was 142 kN in tension (45°), 109 kN in tension-shear

226 (0°/90°), and 210 kN (180°/270°) directions. The maximum capacity of the connection in

227 compression (225°) direction could not be reached. Detailed results from the tests are presented in

228 [9].
229

230 Figure 5: Test setup for experimental tests in inter-modular connection: testing at a) tension (45°)
231 – compression (225°) and (b) tension-shear (0°/90°) and compression-shear (180°/270°) [9]

232 (a) (b)

233 Figure 6: Monotonic force-displacement curve and hysteretic curves of inter-modular connection
234 in a) 45° tension/ 225° compression direction, and b) 0°/90° tension-shear/ 180°/270°
235 compression-shear direction [9] )
236 2.2. Simulating Connections

237 The behavior of the intra-modular connection was simulated as a uniaxial zero-length Pinching4

238 element. The benchmarking process was implemented separately for tension and shear loading

239 behavior. As the average connection stiffnesses and maximum capacities in the positive and

240 negative directions differed by less than 10%, the symmetric behavior of shear connections was

241 assumed for assembly level simulations.

242 The behavior of inter-modular connections was simulated as three link elements, shown in Figure

243 7, following the methodology developed by Polastri et al. [22,23]. Each link was assigned a set of

244 Pinching4 material parameters along its axis to simulate the nonlinear cyclic response of the

245 connection. The vertical and horizontal links were assigned identical parameters to enforce the

246 assumed symmetry of behavior in 0° or 90°, 135° or 315°, and 180° or 270° loading directions.

247 The material model was benchmarked against a set of reference backbone curves in pure tension

248 (45°) and tension-shear (0° or 90°) for the positive backbone and another set of reference backbone

249 curves in pure compression (225°) and compression-shear (180° or 270°) for the negative

250 backbone parameters. In addition, the three-link configuration was benchmarked for Pinching4

251 pinching and degradation parameters against 12 pairs of tests, each pair consisting of one of three

252 cyclic tests in tension and compression (45°, 225°) and one of four cyclic tests in tension-shear

253 and compression-shear (0°/90°, 180°/270°). Each of these 12 pairs was benchmarked for the

254 Pinching4 parameters each in positive backbone, negative backbone, and hysteresis, following the

255 process shown in Figure 8.

256

257 Figure 7: Three links used to simulate the XRAD connection – horizontal and vertical links have
258 identical properties [adapted from [22,23]]

259 2.3. Benchmarking the hysteresis behavior

260 Optimization of input parameters to benchmark both the intra-modular and inter-modular

261 connections against the experimental results [9] was performed using the SA algorithm [30]

262 implemented in MATLAB R2021a [33]. This provided an autonomous approach to finding

263 multiple best fit parameters using a single objective function at a time. The first objective function

264 (Equation 1) was used to estimate the best fit Pinching4 envelope parameters based on the

265 experimental backbone curves, and the estimates for the Pinching4 pinching and degradation

266 parameters based on experimental hysteresis curves. The goal of this objective function was to

267 reduce the error between the experimental force-displacement data and the numerical estimation.

268 This function was evaluated as the sum of absolute differences as shown in Equation 1, where

269 is the area under the experimental hysteretic curve and is the area under

270 the hysteretic curve predicted by numerical simulation at equivalent points at an interval of 0.1

271 mm displacement.

272 1
273 The second objective function (Equation 2) was used to estimate Pinching4 pinching and

274 degradation parameters based on experimental hysteresis curves. The goal of this function was to

275 minimize the difference in total energy under the experimental and experimental force-

276 displacement curves. This objective function was used to estimate the best fit curves for the

277 Pinching4 pinching and degradation parameters.

278 2

279 The process of benchmarking was carried out in four major steps as shown in Figure 8. The first

280 step of the benchmarking process consisted of extracting the envelope curve from the hysteresis

281 curve and determining the best fit Pinching4 envelope parameters via the SA algorithm with

282 objective function 1 on the extracted envelope curves or pair of curves. This step ensured that the

283 envelope curve of the numerical model match closely with the experimental envelope curve. The

284 starting values for the envelope parameters were estimated manually based on the experimental

285 envelope curves. In the second step, to ensure that the hysteretic curves were matched closely, best

286 fit Pinching4 pinching and degradation parameters were determined using the SA algorithm with

287 objective function 1. The envelope parameters from the first step were used along with the starting

288 values for the pinching and degradation parameters based on values on timber connection from the

289 literature. The third step was a repetition of the second step, except for objective function 2 used

290 as the objective function for the SA algorithm with pinching and degradation parameters from the

291 second step used as a starting point for the. In all these three steps, the SA algorithm was stopped

292 when the respective objective function was less than 5% or if the value of the objective function

293 did not decrease over 1,000 iterations. In the fourth step, the benchmarked Pinching4 parameters

294 for the connection were obtained as an average of the parameters benchmarked for each test or set

295 of tests.
 Extract envelope curve from
hysteresis
Determine best fit values for Determine best fit values for
Pinching4 pinching and Pinching4 pinching and
degradation parameters degradation parameters Determine average
Determine best fit values for using SA algorithm with using SA algorithm with Pinching4 parameters
Pinching4 envelope objective function 1 on objective function 2 on
parameters using SA hysteresis hysteresis.
algorithm with objective
function 1 on envelope
curve

296

297 Figure 8: Benchmarking process for connections

298 2.4. Multi-panel wall simulations

299 A two-dimensional (2D) wall module representing the prototype modular structure [9] that

300 consisted of two 1220 mm x 2440 mm (width x height) CLT panels with its intra- and inter-

301 modular connections was simulated using the Python version of OpenSees, OpenSeesPy [34].

302 Benchmarked Pinching4 connection models were used to simulate the in-plane, intra-modular

303 connections, and point-type inter-modular connections. CLT was simulated as a single layer 105

304 mm thick orthotropic shell (MITC4) divided into a rectangular mesh as shown in Figure 9. The

305 effective properties of the PP CLT estimated from the shear analogy method based on properties

306 of constituent lamellas [35] are listed in Table 1. Since the lamellas were not glued on their edges,

307 Poisson’s ratio for all directions was taken as zero [36].

308 Table 1: Material Properties for CLT

# $ G G# G $ Specific Poisson’s
/"" /"" /"" kN/mm kN/mm kN/mm Gravity ratio
5.992 0.226 0.207 0.047 0.047 0.039 0.43 0

309 E: elastic modulus, G: shear modulus; subscripts 0, 90 and perp respectively represent the major, minor,
310 and perpendicular directions of CLT.
311
312

313 Figure 9: Finite element model of the two-panel wall module

314 To simulate the foundation, the corners of the XRAD link elements were fixed in all directions at

315 the base. In addition, the bottom nodes of the CLT wall were connected to dummy nodes fixed in

316 all directions with high-stiffness, zero-length compression-only elements (Figure 9) to allow for

317 simulated uplift of the CLT walls.

318 The nodes at the top of the wall elements were constrained via rigid diaphragm constraint to

319 simulate a stiff CLT floor above. Dead loads from that floor were simulated by a vertical force

320 equivalent to a line load of 2.5 kN/m applied to the top of the wall. The wall’s lateral displacement

321 was induced through a displacement-controlled ramp loading applied to the top nodes of the wall

322 module.
323 3. Results

324 This section presents results from the benchmarking of intra- and inter-modular connections and

325 the numerical simulations of multi-panel wall modules.

326 3.1. Benchmarking of intra-modular connection

327 The averaged parameters obtained for the Pinching4 material model fitted to empirical data are

328 presented in Table 2. The behavior of the simulated connection as compared with the experimental

329 data are presented in Figures 10 and 11 for tension and shear, respectively. The agreement between

330 the numerical simulation and the experimental data was quantified as a ratio of the average

331 difference in total energy between the experimental and benchmarked curve to the experimental

332 value (Figures 10b and 11b). The Pinching4 material model was able to predict the force-

333 displacement behavior (Figures 10a and 11a) throughout the cyclic tension testing such that the

334 total difference in energy between the simulation and the experimental data was 1.8 ±1.1% for

335 tension tests and 1.8 ±1.0% for shear tests. For the tension tests (Figure 10b), the simulation was

336 less accurate at the higher displacements, where the relative difference in energy between the

337 simulation and the experimental data increased up to 13%. This increase can be attributed to the

338 model’s inability to simulate the pulling apart of wood fibers in CLT. For shear testing (Figure

339 11b), the difference in energy between the simulation and experimental data was below 5% for

340 each loading cycle.

341 (a) (b)

342 Figure 10: Experimental and numerical results of a tension cyclic test: (a) Force-displacement
343 curve and (b) energy dissipated (the cumulative energy is the area underneath the force-
344 displacement curve)
345

346 (a) (b)

347 Figure 11: Experimental and numerical results of a shear cyclic test: (a) Force-displacement
348 curve and (b) cumulative energy
349 Table 2: Pinching4 parameters for the in-plane connection
Intra-module Connection Inter-module Connection
Pinching4
Parameters Horizontal &
Parameters Shear Tension Diagonal
Vertical
ePf1 (kN) 4.25 4.03 4.74 5.06
ePd1(mm) 1.00 0.44 3.19 3.25
Positive
Envelope ePf2 (kN) 9.75 8.36 26.0 17.4
ePd2 (mm) 7.70 1.77 5.80 5.23
ePf3 (kN) 10.8 9.25 89.8 63.7
 ePd3 (mm) 19.9 3.11 26.0 22.0
ePf4 (kN) 6.67 3.27 22.7 11.0
ePd4 (mm) 36.8 9.41 34.0 27.6
eNf1 (kN) -8.34 -4.71
eNd1 (mm) -2.55 -3.02
eNf2 (kN) -68.8 -38.1
Negative eNd2 (mm) -7.67 -5.28
Envelope eNf3 (kN) -147 -135
eNd3 (mm) -18.4 -13.3
eNf4 (kN) -5.10 -111
eNd4 (mm) -31.6 -14.5
rDispP 0.625 0.523 0.357 0.260
rForceP 0.290 0.482 0.493 0.248
uForceP -0.142 -0.313 -0.253 -0.553
Pinching
rDispN 0.679 0.248
rForceN 0.461 0.381
uForceN -0.358 -0.401
gK1 0.109 0.176 0.589 0.329
Unloading gK2 0.406 0.153 0.614 0.345
stiffness gK3 0.104 0.330 0.717 0.316
degradation gK4 0.235 0.221 0.702 0.391
gkLim 0.150 0.177 0.352 0.259
gD1 0.164 0.106 0.639 0.192
Reloading gD2 0.251 0.144 0.264 0.577
stiffness gD3 0.419 0.293 0.643 0.882
degradation gD4 0.240 0.090 0.475 0.585
gDLim 0.149 0.003 0.301 0.296
gF1 0.000 0.000 0.000 0.000
gF2 0.000 0.000 0.000 0.000
Strength
gF3 0.000 0.000 0.000 0.000
degradation
gF4 0.000 0.000 0.000 0.000
gFLim 0.000 0.000 0.000 0.000
Energy
1.00 1.00 1.00 1.00
degradation gE
Damage type dmgType Cycle Cycle Cycle Cycle
Average % difference in energy* 1.8±1.1% 1.8±1.0% 2.1±1.6%
350 * represents the arithmetic mean of the relative differences in total predicted and measured energy dissipated for the
351 tests in referred directions
352 3.2. Benchmarking of inter-modular connection

353 The three-link connection modeling configuration (Figure 7) was benchmarked against 12 pairs of

354 experimental data. The Pinching4 parameters for the three-link elements are presented in Table 2.

355 One of these pairs of the benchmarked force-displacement curves along with the respective

356 experimentally measured tension-shear and compression-shear load-displacement data is

357 presented in Figure 12a. A plot of the cumulative displacement energy throughout the test for the

358 numerical simulation and experimental data is presented in Figure 12b. The average relative

359 difference in the total energy between the predicted and experimental results was 2.1 ±1.6%, with

360 larger differences in post-peak cycles. This three-link setup (Figure 7) was able to simulate the

361 cyclic force-displacement response of the point-type connection in both tension and compression

362 (45°, 225°), and in tension-shear and compression-shear (0°/90°, 180°/270°) directions such that

363 the difference in total energy between the experimental and predicted results were less than 5% of

364 the total energy from experiment for each test. While the relative difference in the total energy for

365 each cycle of loading for tension-shear and compression-shear (0°/90°, 180°/270°), and tension

366 and compression (45°, 225°), was less than 5%, at large displacement cycles, the difference

367 increased up to 20%.

368 (a) (b)

369 Figure 12: Hysteresis behavior of the point-type connection – experimental and analytical in
370 OpenSees for tension-shear and compression-shear loads: (a) load-displacement curves and (b)
371 energy dissipated

372 3.3. Multi-panel wall behavior

373 The benchmarked models of the connections were used to perform a pushover analysis of a single-

374 panel wall and a two-panel wall module using butt joints for the intra-modular connection. Both
375 walls were connected to the foundation with point-type connections at the bottom corners (Figure

376 9). The screw spacing of the butt joints for the intra-modular connections was varied between 100

377 mm and 350 mm.

378

379 Figure 13: Simulated pushover curves for a two-panels wall with point-type connection modeled
380 as a three-link Pinching4 element and intra-module butt-joint between the panels with 10 different
381 screw spacing compared to a single panel wall segment (dashed black line)
382 The load-displacement simulation curves show lower effective stiffness for the two-panel modules

383 compared to the single-panel wall, with abrupt post-peak failure in the point-type connection in

384 both single- and two-panel walls (Figure 12). The single- and two-panel walls reached their

385 maximum capacity when the point-type inter-modular connection was at its maximum capacity in

386 tension. The effective stiffness of the two-panel modules varied with the screw spacing in the butt

387 joint, with decreasing values of effective stiffness for an increase in screw spacing (Table 3). For

388 screw spacing less than or equal to 180 mm, the predicted drop in the effective stiffness compared
389 to a single-panel wall was within 5%, but without a measurable drop in maximum capacity (Figure

390 13, Table 3). For screw spacing greater than or equal to 200 mm, the predicted stiffness and

391 maximum capacity of the two-panel wall modules decreased with the increasing spacing of the

392 screws (Figure 13, Table 3). The shape of the simulated force-displacement curves for multi-panel

393 walls with screw spacing less than 150 mm is similar to a single panel wall, albeit with a slight

394 (less than 2.0%) decrease in the initial stiffness (Figure 13).

395 As the spacing of inclined screws increases, the displacement at the predicted failure of the walls

396 increases (Figure 13, Table 3). Compared to the single-panel wall, the predicted failure in modules

397 with screw spacing up to 170 mm occurs at the same maximum capacity but with gradually

398 increasing displacements. For screw spacing greater than or equal to 180 mm the failure is

399 predicted at lower maximum capacity and higher displacements (Figure 13).

400 The model predicted that the screw spacing in multi-panel walls should influence the relative slip

401 between the two panels, which for panels with screws spaced less than or equal to 150 mm, would

402 be within 35% of the displacement of the screw at maximum shear capacity (7 mm compared to

403 )*+ 20 "" ) (Figure 14), a behavior close to that of a single-panel wall. For screws spaced

404 more than 180 mm, the relative slip between the panels at maximum capacity exceeds the expected

405 withdrawal displacement )*+ of screws at the maximum shear capacity, indicating the shear

406 fracture of screws indicating the coupled panel behavior.

407
408

409 Figure 14: Initial stiffness and the relative slip between the panels of a two-panel wall for varied
410 spacing of screws of the intra-modular connections

411 4. Discussion

412 4.1. Intra- and inter-modular connection modeling

413 While the Pinching4 model was able to simulate the shear behavior of intra-modular connection

414 within 5% for hysteresis behavior, the difference increased to 13% for tension tests at higher

415 displacements. These errors suggest that while the model was able to simulate the yielding of

416 screws and crushing of the wood as observed during the shear tests, the model was not able to

417 simulate the pulling apart of wooden fibers during the tension loading at higher displacements.

418 However, the model predicted the force-displacement behavior such that the average difference in

419 total energy is around 1%. This small error implies the model could be satisfactorily used for

420 simulations of the selected intra-modular connection for structural-level modeling.

421 The error of simulating the inter-modular connection using the three-link connection was higher

422 than the intra-modular connection with the difference in energy reaching as high as 20% for large

423 displacements. This could be attributed be due to the model’s inability to simulate tensile yielding

424 and shear fracture of screws, crushing of the LVL insert, and yielding of the steel plate at the bolt

425 holes. However, the model was able to simulate many of the other failure mechanisms and the

426 total energy of the connection model is within 2.1 ± 1.6%, of the experimental data, therefore,

427 this modeling methodology could be used for structure level simulations. While it is assumed that

428 the behavior in shear (135°/315°) will be the interpolated behavior between the tension-shear

429 (0°/90°) and compression-shear (180°/270°) as expected from the previous tests (refer to Figure 1

430 for directions and the capacity envelope), there is still a need to verify the behavior of the

431 connection in shear direction with the experimental data.

432 4.2. Two-panel wall behavior

433 For screw spacing less than or equal to 150 mm, the behavior of the multi-panel wall is expected

434 to be controlled by the behavior of the point-type corner connections only because the force-

435 displacement of these walls closely matches that of the single panel wall. Two-panel wall modules

436 with screw spacing of less than or equal to 150 mm are expected to behave similarly to a single-

437 panel wall with a relative slip between the panels of less than 7 mm, which is considered small

438 compared to the displacement of the intra-modular connection at maximum capacity ()*+

439 20 "") (Figure 14). For screw spacing greater than 150 mm, the relative slip between the

440 component panels is expected to be substantially larger and cannot be neglected (Figure 14). Based

441 on these findings, modeling as a single panel could overestimate the capacity and stiffness for two-

442 panel wall modules with screw spacing greater than 150 mm.
443 In two-panel wall modules with screw spacing greater than 170 mm, at the maximum capacity of

444 the wall, both the in-plane intra-module connections and point-type inter-module connections are

445 expected to exceed the displacements at their maximum capacity ()*+ ). Although these two-

446 panel wall modules are expected to have a lower capacity than a single-panel wall module (Figure

447 13), the wall systems are projected to be capable of more lateral deformation prior to reaching

448 maximum capacity, which makes these wall modules more suited for designs requiring ductility

449 (e.g., in seismic regions). Another benefit of using a larger spacing of screws is construction cost

450 in terms of material, labor, and construction time.

451 The simulation suggests that the two-panel wall modules with screw spacing of less than or equal

452 to 150 mm could be modeled as single-panel walls (i.e., without modeling the in-plane

453 connections). For screw spacing between 150 -170 mm, the two-panel wall modules may be

454 substituted with a single-panel wall to simulate the maximum capacity, however, the displacement

455 at maximum capacity will be underestimated by 15 – 35%. For screw spacing above 180 mm, the

456 model has to reflect the response of the intra-module joint.

457 4.3. Use of simplified methods to predict the capacity of multi-panel walls

458 The maximum capacities of the two-panel wall modules predicted through the numerical

459 simulations conducted in this study were compared with predictions of the simplified model

460 proposed by Casagrande et al. [17] (Table 3). The maximum capacity of the multi-panel walls

461 predicted with the simplified model, which does not consider the sliding effect, is lower compared

462 to the results of the numerical simulation by 15 – 60%. The maximum capacity estimated by the

463 simplified methods is lower than those predicted by the numerical models. Therefore, they can be

464 used for initial design capacity estimates for multi-panel walls with closely spaced screws, but not

465 for the analysis of these walls.

466 Table 3: Maximum capacity of the two-panel wall system – estimates based on [17] and numerical
467 model
Predicted Maximum Capacity (kN)
Casagrande et al. [17] Numerical Model % Difference
100 115 136 15%
125 92 137 32%
150 77 138 44%
160 72 138 48%
170 68 138 51%
180 64 133 52%
200 58 124 54%
250 46 109 58%
300 41 98 59%
350 38 94 60%
468

469 5. Conclusions

470 This paper presents a methodology and results from benchmarking selected intra- and inter-

471 modular connections in PP CLT against experimental force-displacement data. These

472 benchmarked models of connections were then used to assess the influence of the geometry of

473 these connections on multi-panel wall modules.

474 The methodology with simulated annealing that used objective functions based on energies of the

475 experimental and simulated curves was able to benchmark the cyclic force-displacement behavior

476 of selected intra- and inter-modular PP CLT connections. The Pinching4 material model was

477 successfully used to simulate the behavior of these selected intra- and inter-modular PP CLT

478 connections that could be used for further modeling of these connections in structural-level

479 numerical models in OpenSees.

480 Numerical simulations using the benchmarked model suggest that the maximum capacity and the

481 lateral displacements of the two-panel wall module are affected by the screw spacing of the intra-

482 modular connections. The numerical simulations also suggest that two-panel wall modules with
483 closely spaced screws (less than 150 mm) can be used for designs where inter-story drift should

484 be limited. At spacings at or below 150 mm, these two-panel wall modules can also be estimated

485 as a single wall module and their capacities are estimated conservatively using the simplified

486 methods.

487 However, where deformation energy dissipation is required (as in seismic design), the multi-panel

488 wall modules with larger spacing between the screws (greater than 180 mm) are expected to be a

489 more suitable choice. In those instances, the numerical simulations of two-panel wall modules

490 have to reflect the response of the intra-module joint and so none of these simplifications are

491 possible.

492 Further research is needed to incorporate the modeling methodologies presented within this paper

493 into simulations of wall modules with more than two panels, to other types of connections, and to

494 complete modular buildings. This will be essential for predicting the seismic performance of

495 rapidly deployable modular CLT structures similar to that referenced in this paper. The results of

496 this study may be used by engineers and researchers to analyze and evaluate the behavior of

497 modular structures in high seismic regions to design more robust and resilient structures for a

498 variety of occupancies and usages.

499 6. Acknowledgment

500 This research is supported by USDA Forest Service (Wood Innovation Grants program 17-DG-

501 1162765-742 and 18-DG-11062765-738). Ian Morrel is gratefully acknowledged for his support

502 during the benchmarking process.

503 7. Author’s Contribution

504 SB, EF, and MR conceived the study. SB conducted the numerical analysis with input from EF.

505 SB performed the data analysis and interpretation of results with the guidance of EF and MR. SB
506 led the drafting of the article with critical input from EF, MR, and LM. All authors have reviewed

507 and agreed to the final version of the article.

508 8. Conflict of Interest

509 The authors declare that they have no conflict of interest.

510 9. Data Availability

511 All or parts of the data used for the paper can be obtained from the corresponding author on

512 reasonable request

513 10. References

514 [1] Guttmann E. Brettsperrholz: Ein Produktporträt (in German) [Cross laminated timber: a

515 product profile]. Zuschnitt 2008;31:12–4.

516 [2] Brandner R, Flatscher G, Ringhofer A, Schickhofer G, Thiel A. Cross laminated timber

517 (CLT): overview and development. Eur J Wood Wood Prod 2016;74:331–51.

518 https://doi.org/10.1007/s00107-015-0999-5.

519 [3] Brandner R. Production and Technology of Cross Laminated Timber ( CLT ): A state-of-

520 the-art Report. Focus Solid Timber Solut. - Eur. Conf. Cross Laminated Timber, Graz,

521 Austria: University of Bath; 2013, p. 3–36.

522 [4] Lu Z, Zhou H, Liao Y, Hu C. Effects of surface treatment and adhesives on bond

523 performance and mechanical properties of cross-laminated timber (CLT) made from small

524 diameter Eucalyptus timber. Constr Build Mater 2018;161:9–15.

525 https://doi.org/10.1016/j.conbuildmat.2017.11.027.

526 [5] Thomas RE, Buehlmann U. Using Low-Grade Hardwoods for CLT Production : A Yield

527 Analysis. 6th Int Sci Conf Hardwood Process 2017:199–206.

528 [6] Fredriksson M, Bomark P, Broman O, Grönlund A. Using small diameter logs for cross-
529 laminated timber production. BioResources 2015;10:1477–86.

530 https://doi.org/10.15376/biores.10.1.1477-1486.

531 [7] Jahedi S. Defining Project‐Specific Custom CLT Grade Utilizing Low‐Value Ponderosa

532 Pine Lumber from Logs Harvested in SW Oregon and Northern California Forest

533 Restoration Programs. Oregon State University, 2022.

534 [8] ANSI/APA. ANSI/APA PRG 320:2019 Standard for Performance-Rated Cross-Laminated

535 Timber. 2020.

536 [9] Bhandari S. Modular Cross Laminated Timber Structures Using Underutilized Ponderosa

537 Pine. Oregon State University, 2022.

538 [10] Rothoblaas. X-RAD Catalogue. Rothoblaas; 2019.

539 [11] Mohammad M, Douglas B, Rammer D, Pryor SE. Connections in cross-laminated timber

540 buildings. In: Karacabeyli E, Douglas B, editors. CLT Handb. cross-laminated timber,

541 Leesburg, VA: FPInnovations; 2013.

542 [12] Hossain A, Danzig I, Tannert T. Cross-Laminated Timber Shear Connections with Double-

543 Angled Self-Tapping Screw Assemblies. J Struct Eng (United States) 2016;142:1–9.

544 https://doi.org/10.1061/(ASCE)ST.1943-541X.0001572.

545 [13] Polastri A, Giongo I, Piazza M. An Innovative Connection System for Cross-Laminated

546 Timber Structures. Struct Eng Int 2017;27:502–11.

547 https://doi.org/10.2749/222137917X14881937844649.

548 [14] Angeli A, Polastri A, Callegari E, Chiodega M. Mechanical characterization of an

549 innovative connection system for CLT structures. WCTE 2016 - World Conf Timber Eng

550 2016.

551 [15] Popovski M, Karacabeyli E. Seismic behaviour of cross-laminated timber structures. World
552 Conf. Timber Eng. 2012, WCTE 2012, vol. 2, Auckland: 2012, p. 335–44.

553 [16] Nolet V, Casagrande D, Doudak G. Multipanel CLT shearwalls: an analytical methodology

554 to predict the elastic-plastic behaviour. Eng Struct 2019;179:640–54.

555 https://doi.org/10.1016/j.engstruct.2018.11.017.

556 [17] Casagrande D, Doudak G, Mauro L, Polastri A. Analytical Approach to Establishing the

557 Elastic Behavior of Multipanel CLT Shear Walls Subjected to Lateral Loads. J Struct Eng

558 (United States) 2017;144:1–13. https://doi.org/10.1061/(ASCE)ST.1943-541X.0001948.

559 [18] Lukacs I, Björnfot A, Tomasi R. Strength and stiffness of cross-laminated timber (CLT)

560 shear walls: State-of-the-art of analytical approaches. Eng Struct 2019.

561 https://doi.org/10.1016/j.engstruct.2018.05.126.

562 [19] Gavric I, Fragiacomo M, Ceccotti A. Cyclic behavior of CLT wall systems: Experimental

563 tests and analytical prediction models. J Struct Eng (United States) 2015;141:1–14.

564 https://doi.org/10.1061/(ASCE)ST.1943-541X.0001246.

565 [20] Shahnewaz M, Alam S, Tannert T. In-plane strength and stiffness of cross-laminated timber

566 shear walls. Buildings 2018;8:1–13. https://doi.org/10.3390/buildings8080100.

567 [21] Izzi M, Polastri A, Fragiacomo M. Investigating the Hysteretic Behavior of Cross-

568 Laminated Timber Wall Systems due to Connections. J Struct Eng (United States)

569 2018;144:1–10. https://doi.org/10.1061/(ASCE)ST.1943-541X.0002022.

570 [22] Polastri A, Giongo I, Pacchioli S, Piazza M. Structural analysis of CLT multi-storey

571 buildings assembled with the innovative X-RAD connection system: Case-study of a tall-

572 building. WCTE 2016 - World Conf Timber Eng 2016.

573 [23] Polastri A, Giongo I, Angeli A, Brandner R. Mechanical characterization of a pre-fabricated

574 connection system for cross laminated timber structures in seismic regions. Eng Struct
575 2018;167:705–15. https://doi.org/10.1016/j.engstruct.2017.12.022.

576 [24] OpenSeesWiki. SAWS Material 2010.

577 https://opensees.berkeley.edu/wiki/index.php/SAWS_Material (accessed October 19,

578 2021).

579 [25] Mitra N. Pinching4 Material. OpenSees Wiki 2012.

580 https://opensees.berkeley.edu/wiki/index.php/Pinching4_Material (accessed March 6,

581 2022).

582 [26] McKenna F, Scott MH, Fenves GL. Nonlinear Finite-Element Analysis Software

583 Architecture Using Object Composition. J Comput Civ Eng 2010;24:95–107.

584 https://doi.org/10.1061/(asce)cp.1943-5487.0000002.

585 [27] Benedetti F, Rosales V, Opazo-Vega A, Norambuena-Contreras J, Jara-Cisterna A.

586 Experimental and numerical evaluation of hold-down connections on radiata pine Cross-

587 Laminated-Timber shear walls: a case study in Chile. Eur J Wood Wood Prod 2019;77:79–

588 92. https://doi.org/10.1007/s00107-018-1365-1.

589 [28] Shen YL, Schneider J, Tesfamariam S, Stiemer SF, Mu ZG. Hysteresis behavior of bracket

590 connection in cross-laminated-timber shear walls. Constr Build Mater 2013;48:980–91.

591 https://doi.org/10.1016/j.conbuildmat.2013.07.050.

592 [29] Gagnon S, Bilek ET, Podesto L, Crespell P. CLT Handbook. FPInnovations; 2013.

593 [30] Kirkpatrick S, Gelatt CD, Vecchi MP. Optimization by simulated annealing. Science (80- )

594 1983;220:671–80. https://doi.org/10.1126/science.220.4598.671.

595 [31] Mahdavifar V, Barbosa AR, Sinha A, Muszynski L, Gupta R, Pryor SE. Hysteretic

596 Response of Metal Connections on Hybrid Cross-Laminated Timber Panels. J Struct Eng

597 (United States) 2019;145:1–14. https://doi.org/10.1061/(ASCE)ST.1943-541X.0002222.

598 [32] Morrell I, Soti R, Miyamoto B, Sinha A. Experimental Investigation of Base Conditions

599 Affecting Seismic Performance of Mass Plywood Panel Shear Walls. J Struct Eng

600 2020;146:04020149. https://doi.org/10.1061/(asce)st.1943-541x.0002674.

601 [33] MathWorks. MATLAB 2022.

602 [34] Zhu M, McKenna F, Scott MH. OpenSeesPy: Python library for the OpenSees finite

603 element framework. SoftwareX 2018;7:6–11. https://doi.org/10.1016/j.softx.2017.10.009.

604 [35] Jahedi S, Muszyński L, Riggio M, Blengino BB, Bhandari S. MoE distribution in visually

605 graded Ponderosa Pine lumber harvested from restoration programs in Southern Oregon and

606 Northern California. Wood Fiber Sci 2022;[In Press].

607 [36] Aranha C, Branco J, Lourenço P, Flatscher G, Schickhofer G. Finite element modelling of

608 the cyclic behaviour of CLT connectors and walls. WCTE 2016 - World Conf. Timber Eng.,

609 Vienna, Austria: 2016, p. 3501–8.

610