Journal of Structural Engineering Seismic Simulation of Integrated Ceiling-Partition Wall-Piping System at E-Defense, Part 2: Evaluation of Nonstructural Damage and Fragilities --Manuscript Draft-Manuscript Number: STENG-3260R1 Full Title: Seismic Simulation of Integrated Ceiling-Partition Wall-Piping System at E-Defense, Part 2: Evaluation of Nonstructural Damage and Fragilities Manuscript Region of Origin: UNITED STATES Article Type: Technical Paper Section/Category: Seismic Effects Abstract: A full-scale, five-story steel moment frame building in base-isolated and fixed-base configurations was subjected to a number of ground motions using the E-Defense shake table. In these experiments, more than 84 m2 (900 sf) of suspended ceiling with lay-in tiles, 90 m (300 linear ft) of partition walls with individual lengths varying from 1.5 to 9.8 m (5 to 32 ft), and 3 sprinkler branch lines were installed below the 5th and 6th (roof) floors of the building. Since the horizontal floor accelerations were generally constrained to relatively low values by the base isolation system, several damage states related to vertical floor system acceleration were observed. One key observation is that use of lateral bracing with compression posts did not improve the seismic response of suspended ceilings when subjected to strong vertical excitation. Acceleration amplification factors of the ceiling-partition-partition components relative to structural floor accelerations were computed. The code prescribed amplification factors for the design of nonstructural components was consistent with the observed amplification in the horizontal direction, but unconservative in the vertical direction because the code neglects the additional amplification produced by slab vibration. Corresponding Author: Keri L Ryan, PhD University of Nevada, Reno Reno, NV UNITED STATES Corresponding Author E-Mail: [email protected] Order of Authors: Siavash Soroushian Emmanuel Manos Maragakis Keri L Ryan, PhD Eiji Sato Tomohiro Sasaki Taichiro Okazaki Gilberto Mosqueda Suggested Reviewers: Tara Hutchinson University of California, San Diego [email protected] Hutchinson led another recent test program of a full-scale building outfitted with nonstructural components and has desirable expertise for all aspects of the companion papers. However, Hutchinson is also a Co-PI on the Grand Challenge project, and so it might be considered a conflict of interest, although she was not involved in this aspect of the project. Amir Gilani Miyamoto International [email protected] Gilani has published experimental research on seismic response of ceiling systems and has expertise to evaluate structural and nonstructural responses observed in this research. Claudia Marin Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation Howard University [email protected] Marin has expertise in seismic isolation and was involved in the test program on fullscale building outfitted with nonstructural components that was led by Hutchinson. Matthew Hoehler Hilti Corporation [email protected] Hoehler has been involved in experimental research on seismic response of ceiling systems and has expertise to evaluate structural and nonstructural responses observed in this research. Opposed Reviewers: Additional Information: Question Response Is the article being considered for more No than one journal? The Journal of Structural Engineering does not review manuscripts that are being submitted simultaneously to another organization or ASCE journal for publication. Is this article already published? No. Manuscript contains material previously presented in a conference publication but Material that has been previously about 50% is new material. published cannot be considered for publication by ASCE. A manuscript that has been published in a conference proceedings may be reviewed for publication only if it has been significantly revised. If you answer YES, please explain. Have all the authors contributed to the Yes study and approved the final version? All authors must have contributed to the study, seen the final draft of the manuscript, and accept responsibility for its contents. It is unethical to list someone as a coauthor who does not want to be associated with the study and who has never seen the manuscript. Was an earlier version of the paper previously considered and declined by ASCE? Declined manuscripts are sent through the review process again. If your manuscript has been submitted to us before under a different title, please provide that title in the space provided below. It is our policy to inform an editor that a manuscript has been previously reviewed, even when it has been reviewed by a different Division, Institute, or Council within ASCE. No Do your table titles/figure captions cite other sources? If you used a figure/table from another source, written permission for print and No Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation online use must be attached in PDF format. Permission letters must state that permission is granted in both forms of media. If you used data from another source to create your own figure/table, the data is adapted and therefore obtaining permission is not required. Does your paper exceed 10,000 words? If Yes YES, please provide justification in your cover letter. If you need help estimating word length, see our sizing worksheet at this link: Sizing Worksheet Estimates for color figures in the printed No journal begin at $924. Cost increases depend on the number and size of figures. Do you intend for any figure to be printed in color? If YES, how many and which ones? Please provide a total count and also list them by figure number. Is this manuscript a companion to one Yes, Part 2 of 2 already submitted/or being submitted? If yes, please note whether this is part I, II, or III. Please make sure all related papers are uploaded on the same day and provide the date of submission, title, and authors of each. Is this manuscript part of a Special Issue? No If yes, please provide the Special Issue title and name of the guest editor. To read ASCE's Data Sharing Policy, please click on the "Instructions" link associated with this question. According to this policy, you are required to report on any materials sharing restrictions in your cover letter. Are you restricted from sharing your data & materials? If yes, did you report on these in your cover letter? No Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation Cover Letter Click here to download Cover Letter: Cover Letter_Revision 1.pdf College of Engineering University of Nevada Reno Oct. 26, 2014 Sherif El-Tawil, Ph.D., P.E., F.ASCE, Dept. of Civil and Env. Engineering University of Michigan Ann Arbor, MI 48109-2125 Ph (734) 764-5617 Fax (734) 764-4292 [email protected] Managing Editor of ASCE Journal of Structural Engineering Dear Professor El-Tawil: On behalf of the NEES/E-Defense collaborative research project on base-isolation and nonstructural components, I hereby submit revised versions of STENG-3259 and STENG-3260 for further consideration as Technical Papers in Journal of Structural Engineering. The titles of the manuscripts have been revised to “Seismic Simulation of Integrated Ceiling-Partition Wall-Piping System at E-Defense, Part 1: Influence of 3D Structural Response and Base Isolation”, and “Seismic Simulation of Integrated Ceiling-Partition Wall-Piping System at E-Defense, Part 2: Evaluation of Nonstructural Damage and Fragilities”. We wish the revised manuscripts to be reviewed as companion papers. The reviewer requested additional information in Part 1 so that the study could be better understood without referring to other documents. Therefore, the length of the submitted manuscript for Part 1 has increased from 9 to 10 pages according to the sizing worksheet estimate. We believe the additions have led to a more readable paper, and the additional length is justified. The length of Part 2 remains at an estimated 10 pages. To reiterate from the original submission, the manuscripts contain color figures, but all figures can be understood in black and white. Thus, our intention is for figures to be published in color electronically and in black and white for the printed journal. We look forward to learning the outcome of the manuscript peer review process. If there are any problems with the submission, please let me know. Sincerely, Keri L. Ryan, Ph.D. Associate Professor Department of Civil and Environmental Engineering University of Nevada, Reno/258 Reno, Nevada 89557-0152 (775) 784-6937 office (775) 784-1390 fax Manuscript Click here to download Manuscript: E-Defense companion paper 2 - Revision 1 - No Figures.docx 1 Seismic Simulation of an Integrated Ceiling-Partition Wall-Piping 2 System at E-Defense, Part 2: Evaluation of Nonstructural 3 Damage and Fragilities 4 Siavash Soroushian,a) E. "Manos" Maragakis,b) Keri L. Ryan,c) Eiji Sato,d) 5 Tomohiro Sasaki,e) Taichiro Okazaki,f) Gilberto Mosquedag) 6 Abstract 7 A full-scale, five-story steel moment frame building in base-isolated and fixed-base 8 configurations was subjected to a number of ground motions using the E-Defense shake 9 table. In these experiments, more than 84 m2 (900 sf) of suspended ceiling with lay-in tiles, 10 90 m (300 linear ft) of partition walls with individual lengths varying from 1.5 to 9.8 m (5 to 11 32 ft), and 3 sprinkler branch lines were installed below the 5th and 6th (roof) floors of the 12 building. Since the horizontal floor accelerations were generally constrained to relatively low 13 values by the base isolation system, several damage states related to vertical floor system 14 acceleration were observed. One key observation is that use of lateral bracing with 15 compression posts did not improve the seismic response of suspended ceilings when 16 subjected to strong vertical excitation. Acceleration amplification factors of the ceiling- 17 partition-partition components relative to structural floor accelerations were computed. The a) Post-doctoral Scholar, Dept. of Civil and Environmental Engineering, University of Nevada, Reno, MS 0258, Reno, NV 89557-0258 b) Dean of Engr., Dept. of Civil and Environmental Engineering, University of Nevada, Reno, MS 0256, Reno, NV 89557-0256 c) Assoc. Prof., Dept. of Civil and Environmental Engineering, University of Nevada, Reno, MS 0256, Reno, NV 89557-0258 d) Dr. Engr., National Research Institute for Earth Science and Disaster Prevention, 1501-21 Nishikameya, Mitsuta, Shijimi-cho Miki, Hyogo, Japan 673-0515 e) Dr. Engr., National Research Institute for Earth Science and Disaster Prevention, 1501-21 Nishikameya, Mitsuta, Shijimi-cho Miki, Hyogo, Japan 673-0515 f ) Assoc. Prof., Graduate School of Engineering, Hokkaido University, Kita 13, Nishi 8, Kita-ku, Sapporo, Hokkaido, Japan, 060-8628 g) Assoc. Prof., Dept. of Structural Engineering, University of California, San Diego, 9500 Gilman Dr. MC0085, La Jolla, CA 92093-0085 18 code prescribed amplification factors for the design of nonstructural components was 19 consistent with the observed amplification in the horizontal direction, but unconservative in 20 the vertical direction because the code neglects the additional amplification produced by slab 21 vibration. 22 Keywords: nonstructural components, vertical ground motion, shake table testing, suspended 23 ceiling, sprinkler-piping, fragility functions 24 Introduction 25 The performance of critical facilities such as hospitals and fire stations during an 26 earthquake depends not only on structural systems, but also on the functionality of 27 nonstructural components (FEMA E-74, 2011). The shaking intensities that can cause 28 damage to the nonstructural components are typically lower than those that induce structural 29 damage (Miranda, 2003). Also, nonstructural systems almost always represent the major 30 portion of the total investment in buildings (Whittaker and Soong, 2003). Therefore, 31 economic losses in buildings due to nonstructural damage or malfunction can be much larger 32 than those directly related to structural performance (Taghavi and Miranda, 2003). 33 In the 2010 Chile earthquake, few buildings suffered structural damage, but economic 34 loss and damage of nonstructural systems was extensive in commercial, residential, office 35 and industrial buildings (Miranda et al., 2012). Damage to the acoustical suspended ceiling 36 systems including fallen ceiling panels was widespread; in some cases, 100% of the ceiling 37 panels fell. Several types of damage such as leakage at pipe joints, failure of braces, and 38 breakage of sprinklers were observed on fire sprinkler piping systems. Damage to partition 39 walls lacking adequate connections to the structure and/or adequate gaps was observed 40 during the earthquake (Miranda et al., 2012). Following the 2011 off the Pacific coast of 41 Tohoku Earthquake, 714 structures were inspected. Major portions of ceiling systems had 2 42 collapsed, and widespread water leakage was observed in inspected buildings due to 43 extensive damage to fire protection systems (Mizutani et al., 2012). These observations from 44 real earthquakes are insightful, but often lack measured responses (both structural and 45 nonstructural) that can correlate the complex system behavior and damage states to demand 46 parameters. 47 Several component and subsystem level nonstructural experiments have been conducted 48 in recent years. Examples of previously tested components include ceiling subsystems with 49 different sizes and aspect ratios (Badillo-Almarez et al., 2007; Gilani et al., 2010), piping 50 assemblies and subsystems (Zaghi et al., 2012; Tian et al., 2012), and individual partition 51 walls in different configurations and boundary conditions (Retamales et al., 2012). 52 However, these component and subsystem level experiments may not accurately reflect 53 the following influences of real buildings on nonstructural response: realistic input excitation, 54 interaction between different types of nonstructural components, realistic boundary 55 conditions, and floor system vibration. Furthermore, with base isolation, which is often 56 chosen to preserve functionality of critical buildings – the nonstructural components are 57 subjected to a different proportion of horizontal to vertical accelerations than prescribed by 58 seismic qualification tests, and so the observations from component tests may not strictly 59 apply. In summary, the limited quantitative data collected from past earthquakes and the 60 limitations of subsystem level experimental studies to reproduce structural and nonstructural 61 system interaction effects necessitates system level experiments and analytical tools to 62 facilitate a better understanding of the seismic response of nonstructural systems. The 63 response of nonstructural components observed in realistic building testbeds for various 64 structural systems that can replicate both horizontal modal effects and floor system vibration 65 is of particular interest. 3 66 As part of a collaborative research project between Network for Earthquake Engineering 67 Simulation (NEES) and the National Research Institute for Earth Science and Disaster 68 Prevention (NIED) of Japan, system-level full-scale shaking experiments of a 5-story 69 building were conducted at E-Defense. The investigation of the ceiling-partition-piping 70 (CPP) component performance in this building was led by the NEES Grand Challenge 71 project “Simulation of the Seismic Performance of Nonstructural Systems”. This is the 72 second of two related papers. In the first paper (Ryan et al., 2013a), the 3-dimensional (3D) 73 structural response in both base-isolated and fixed-base configurations was evaluated, and the 74 relation between severity of CPP damage and horizontal floor and vertical slab accelerations 75 was identified. In this paper, we evaluate the response of the integrated ceiling, partition, and 76 fire sprinkler piping systems, which were installed in the building throughout the test series. 77 Specifically, various observed CPP damage states – many of which were induced directly by 78 strong floor slab acceleration – are identified, and fragility data for these damage states are 79 developed where applicable. 80 Experimental Setup 81 The testbed structure (two base-isolated configurations and one fixed-base configuration) 82 that housed CPP components was described in Ryan et al. (2013a). A partition-ceiling- 83 sprinkler piping subassembly was designed and installed in nearly identical configuration 84 below the 5th and 6th floors of the testbed. Further details about these components are 85 described here. 86 Suspended Ceilings 87 An approximately 84 m2 (900 sf) lay-in-tile suspended ceiling system was designed for 88 each floor that worked around existing boundaries (Fig. 1). However, the ceiling area was 89 interrupted at two locations (total area of 3 m2 or 34 sf) that were impeded by vertical trusses 4 90 used to measure story drifts. The ceilings were installed in the test frame per ASTM 91 E580/E580M-11ae1 standards (ASTM, 2011). The grid was constructed using the heavy-duty 92 USG DONN 24 mm (15/16 in) exposed tee system. Main runners and cross tees were aligned 93 as shown in Fig. 1. The main runners were supported by 12-gauge Hilti X-CW suspension 94 wires spaced 1.2 m (4 ft) apart; additional wires supporting all perimeter grid pieces were 95 placed within 8 in of the partition wall faces. The plenum height (the distance between the 96 bottom of the structural slab and the ceiling system) was 0.9 m (3 ft). 97 A 22 mm (7/8 in) wall molding was attached to the perimeter partition walls. On the 98 North and East sides, the main runners and cross tees were attached tight to the wall molding 99 using USG/ACM7 seismic clips with one partition attached screw and one top hole screw to 100 prevent movement of the ceiling grids (Fig. 2(a)). On the South and West sides, a 19 mm (3/4 101 in) clearance was provided between the main runners/cross tees and the wall molding. This 102 connection used the same seismic clip, but with the second screw attached at the middle of 103 the clip slot to allow the grid members to float freely (Fig. 2(b)). At the hatched areas in Fig. 104 1(a), heavier gypsum board panels were used to simulate the weight of light fixtures. 105 ASTM E580/E580M-11ae1 (ASTM, 2011) requires seismic bracing to be included for 106 ceiling areas larger than 93 m2 (1,000 sf). To compare the behavior of braced and unbraced 107 ceiling systems, seismic braces were installed only on the 6th floor ceiling, while all other 108 ceiling details were identical on both floors. From here forward, the 5th and 6th floor ceilings 109 are understood to be the components suspended from the 5th and roof slabs, respectively. 110 Each seismic brace consisted of: 1) a system of splay wires and 2) a USG/VSA30/40 111 compression post. The seismic braces were placed at 3.7 m (12 ft) on center, in each 112 direction, with the first set occurring within 1.8 m (6 ft) of the wall face. Four wires splayed 113 at 90° from each other were attached to the main runner within 51 mm (2 in) of the 5 114 compression post (Fig. 3). Due to the connection constraints, steel stud compression posts 115 were used instead of VSA30/40 compression posts when the posts were attached to structural 116 girders. In one location on each floor, a 2-way steel stud rigid brace was used in place of two 117 of the splay wires due to space constraints. 118 Fire Sprinkler Piping 119 A standard Schedule 40 piping system was attached to the testbed building per NFPA 13 120 (NFPA, 2011). The 5th and 6th floor piping systems (suspended from the 5th and roof slabs) 121 included one 64 mm (2.5 in) diameter main run and three (North-South) 32 mm (1.25 in) and 122 25 mm (1 in) diameter branch lines per floor connected through a 76 mm (3 in) diameter riser 123 pipe (Fig. 4). All connections on the riser and the main run were groove fit, while the rest of 124 the connections were threaded. Branch Lines 1 and 2, each with three 305 mm (12 in) drops, 125 incorporated armover and straight drops, respectively. For Branch Line 3, Drop 1 was a 305 126 mm (12 in) straight drop, while a Victaulic Aquaflex flexible hose was used at Drop 2 (Fig. 127 5(a)). At Drop 1 of each branch line, a 51 mm (2 in) oversized ring was used to separate the 128 ceiling panel from the sprinkler heads (oversized gap configuration, Fig. 5(b)), while a 129 minimal gap between the ceiling panel and sprinkler head was provided at the rest of the drop 130 locations (no gap configuration, Fig. 5(c)). 131 Lateral resistance was provided by: inclined 25 mm (1 in) diameter longitudinal and 132 lateral pipe sway braces on the main run near the riser pipe (Fig. 6(a)), a lateral pipe sway 133 brace at the end of the main run, and two longitudinal braces at the end of the riser pipe 134 below the 5th floor slab. The ends of the branch lines were restrained with two diagonal splay 135 wires to limit the lateral movement (Fig. 6(b)). 136 Partition Walls 6 137 Approximately 91 m (300 ft) of typical light gauge steel studded gypsum partition walls 138 with individual lengths varying from 1.5 to 9.8 m (5 to 32 ft) were installed in the testbed. 139 The full height partitions were approximately 2.7 m (9 ft) tall. The partition wall details were 140 selected based on the most commonly used commercial and institutional partition walls. 141 Figure 7 presents the plan view of 4th and 5th floor partition walls (understood to be partitions 142 between 4th/5th slabs, and 5th/roof slabs, respectively) of the testbed building. The black and 143 gray labels refer to 4th and 5th floor partition walls, respectively. 144 All partitions were constructed using either 18 mm (0.7in) (350S125-18 and 350T125-18) 145 or 30 mm (1.2 in) (350S125-30 and 350T125-30) studs and tracks with a single ply 13 mm 146 (1/2 in) thick gypsum board on each side of the wall. The 30 mm (1.2 in) thick studs and 147 tracks corresponded to institutional detailing (Retamales et al., 2012), while the 18 mm (0.7 148 in) thick studs and tracks corresponded to commercial detailing. The institutional T-wall and 149 corner connection details incorporated additional studs and more frequent placement of 150 screws than the commercial connection details, as shown in Fig. 8. 151 Full connection detailing was provided for 4th floor partitions while slip track connection 152 detailing was provided for 5th floor partitions. In a full connection, the studs and gypsum 153 boards are fully attached to the top and bottom tracks. Due to the deflection of floors (and top 154 tracks) under live loads, the screws attaching the gypsum boards or studs to the tracks can 155 tear out. To prevent this type of damage, slip track connections have been designed to allow 156 vertical deflection of the top track relative to the gypsum boards and studs. Slip track details 157 can also accommodate horizontal drift with minimal damage. The typical top and bottom 158 details of the full and slip track connections that were installed in this experiment are 159 presented in Fig. 9. 7 160 On the North-East side of the building, a room was constructed with self-standing, partial 161 height partitions on the 4th and 5th floors to house hospital and office contents, respectively 162 (Fig. 10(a)). Full connection detailing was used at the bottom of these partitions while the top 163 boundary, positioned below the ceiling, was unrestrained. Also, bulkhead partitions were 164 built around the drift measurement trusses (at two locations on each floor as shown in Fig. 165 1(a)) to provide realistic ceiling boundary conditions. For the bulkhead partitions, only the 166 corner studs were extended down to the floor to provide access to the trusses (Fig. 10(b)). 167 Instrumentation 168 The table accelerations and the responses of structural and CPP components were 169 monitored by nearly 400 sensor channels (not including the isolation system response, when 170 applicable) recorded at a sampling frequency of 1000 Hz. A 4-pole low-pass Butterworth 171 filter with a cutoff frequency of 25 Hz was applied to all recorded responses. 172 Details of the instrumentation used to measure the structural response (horizontal floor 173 and vertical slab accelerations) are provided in Ryan et al. (2013a). Accelerometers spaced 174 regularly on the ceiling grid members (Fig. 11(a)-(b)) measured grid acceleration. Additional 175 sensors measured accelerations on select ceiling panels and grid points that were seismically 176 braced (Fig. 11(c)-(d)). Displacement transducers on the floating side of the ceiling perimeter 177 measured the movement of the ceiling system relative to the partition walls (Fig. 11(e)). 178 Accelerometers were also placed directly on the sprinkler pipes to measure their 179 longitudinal and transverse movement (Fig. 12(a)-(c)). Displacement transducers located on 180 the main run pipe (near the second branch line) measured the movement of the pipes relative 181 to the structure (Fig. 12(d)). 182 Several uniaxial accelerometers were installed on the full height partitions at or near the 183 ceiling elevation. These accelerometers measured the out-of-plane partition accelerations, 8 184 which corresponded to the acceleration imposed to the ceiling system near the boundaries 185 (Fig. 13(a)). Accelerometers in each horizontal direction were installed on top of the partial 186 height self-standing partitions (Fig. 13(b)). All data discussed in this paper is archived and 187 publicly accessible through the NEES Project Warehouse (Ryan et al. 2013b,c,d). 188 Seismic Response of Suspended Ceiling System 189 The building was subjected to various ground motions over six total days of testing for 190 two base-isolated and one fixed-base building configuration. Out of 41 total excitations, 23 191 were 3D including a vertical component. The achieved or realized excitations encompassed a 192 wide range of shaking intensities and frequencies, which allowed the CPP vulnerability to be 193 critically addressed. The damage mechanisms and the extent of damage were very similar for 194 the three system configurations because, as discussed in Ryan et al. (2013), similar peak 195 demands were observed in the three system configurations and vertical slab vibration was 196 insensitive to the presence of an isolation system. Thus, the damage observations are 197 presented and discussed without further mention of the system configuration during which 198 they were observed, but rather interpreted relative to the horizontal floor acceleration and 199 vertical slab acceleration to which they are closely correlated. The ceiling-piping-partition 200 system was repaired after each test day, but it was never restored to its original configuration. 201 Thus, unless otherwise noted, the results pertaining to repaired CPP components are not 202 reported in this study. 203 A key aspect of the ceiling response is the acceleration of the ceiling components (grid, 204 compression post, and panel) relative to the structural systems to which they are attached 205 (column or slab). Based on the recorded sensor data described previously, Table 1 reports 206 acceleration amplification factors (peak ceiling member acceleration normalized by peak 207 column acceleration). Table 1 also reports peak slab acceleration normalized by peak column 9 208 acceleration to highlight the effect of slab flexibility. The statistics (max, min and median) 209 are based only on valid simulations, which include all 41 simulations for the 5th floor ceiling 210 and only the first 4 simulations for the 6th floor ceiling, because appreciable damage was 211 observed in the 5th simulation. In Table 1, “column” refers to three column sensors while 212 “slab” refers to slab sensors on the given floor, valid only for vertical acceleration. 213 The component amplification factor ap in Eq. 13.3-1 of ASCE 7-10 (2010) accounts for 214 possible amplification of component horizontal response relative to the attached structure due 215 to the inherent component flexibility. The maximum recommended amplification is ap = 2.5 216 for components that are considered flexible; ap can be interpreted as spectral amplification of 217 2.5 relative to the peak column acceleration. Table 1 indicates that the average horizontal 218 (XY) amplification observed during the experiment was 2.6 and 2.8 for the unbraced and 219 braced ceiling, respectively. The observed acceleration correlates well with ap considering 220 that horizontal pounding of the ceiling panels against the grid members amplified the 221 accelerations in some motions. 222 In the vertical direction, the maximum component amplification can be interpreted as: 1) 223 the same value as for the horizontal direction (ap = 2.5) per ICC-AC156 (ICC, 2010) or 2) ap 224 = 2.67, which is the ratio of the constant to short period spectral acceleration (0.8 CV SDS/0.3 225 CV SDS) (ASCE, 2010). As discussed in Ryan et al. (2013a), this component amplification 226 assumes that the vertical ground acceleration is transmitted directly to the nonstructural 227 components, and does not account for amplification of the vertical acceleration as it travels 228 from the ground through the structure to the attachment point of the nonstructural component. 229 In this study, we evaluate component amplification by comparing the peak vertical 230 acceleration of the CPP component to the peak acceleration recorded in the nearest column. 231 Even this definition of component amplification is unconservative, since the vertical 10 232 acceleration was somewhat amplified as it traveled from the shake table to the columns 233 (Ryan et al., 2013a). 234 The median amplification of ceiling vertical acceleration (sensors mounted on the ceiling 235 grid members compared to sensors mounted on the columns) observed in the experiment was 236 4.2 and 5 for the unbraced and braced ceiling, respectively, which is higher than ap by 237 approximately a factor of 2. However, the median vertical amplification of the slab relative to 238 the column was 3 and 5.7 for the 5th and 6th floor, respectively. This suggests that the 239 majority of the vertical acceleration amplification was due to vibration of the floor slab 240 relative to the columns and not increased acceleration in the ceiling relative to the slab. An 241 example of such a trend during 3D-Superstition Hills-Westmorland (TP configuration) was 242 presented in Fig. 14. Some variation in slab acceleration amplification is expected as a 243 function of the slab vibration frequency, which has been shown to significantly vary among 244 realistic floor systems, both in general and in this experiment (Ryan et al., 2013a). Therefore, 245 a constant amplification factor may be inadequate to account for vertical acceleration 246 amplification caused by slab flexibility. The median vertical amplification of the ceiling 247 panel relative to the column sensors was 3.6 and 12 for the unbraced and braced ceiling, 248 respectively. The higher ceiling panel amplification factors in the braced ceiling will be 249 shown to result from pounding of the panels on the ceiling grid, and suggests that anchorage 250 design forces of hanger wires may need to be increased to account for this pounding. 251 In any individual simulation, a maximum of three panels (1%) from the unbraced (4th 252 floor) ceiling were displaced or fell to the floor while up to 40% of the panels in the braced 253 (6th floor) ceiling were displaced and/or fell. Most of the damage was located below the 254 middle of the North-East and South-East slabs, which supported the supplementary roof 255 weight. The condition of the braced and unbraced ceiling after the first 3D-Northridge- 11 256 Rinaldi simulation is compared in Fig. 15. In this simulation, which produced the first 257 significant damage to the ceiling system, the vertical peak table acceleration was 1.2g, which 258 led to peak acceleration of 6.8g on the 5th slab and 6.4g on the roof slab. While the chance of 259 a life threatening injury due to the fallen ceiling panels was low, serious disruption was 260 observed in the office room. Most of the fallen panels were damaged to the extent that 261 replacement would be desirable. In addition, fallen ceiling panels weakened the grid system 262 and increased the chance that more extensive grid repair was necessary. However, over the 263 course of the test program, some of the cross tee sections failed but the main runners always 264 remained intact. 265 In general, the accelerations in all three directions at the slab level, which represent the 266 input excitation to the ceiling system, were slightly higher at the roof slab than the 5th floor 267 slab. However, the input acceleration alone does not explain the difference in damage; 268 observed accelerations in each ceiling suggest that in this experiment, the compression posts 269 used in the lateral bracing increased the damage to the ceiling system. 270 Figure 16 shows the vertical acceleration of a ceiling panel (C9) and ceiling grid (C4) 271 (see Fig. 11(a)) measured for a moderate excitation (3D-Superstition Hills Westmorland/TP 272 configuration) executed prior to the occurrence of ceiling damage. In the unbraced ceiling, 273 little amplification of the panel acceleration relative to the grid was observed (Fig. 16(a)), 274 which suggests that these two components moved together. However, in the braced ceiling, 275 the acceleration of the ceiling panel was significantly amplified relative to the compression 276 post attachment location (Fig. 16(b)), which suggests that the panel pounded on the grid 277 system. The median acceleration amplification values presented in Table 1 imply the same 278 trend. 12 279 The acceleration trends observed in Fig. 16 are explained as follows: consider the 280 diagram of the ceiling system in Figs. 16 and 17, where the vertical acceleration of the slab, 281 grid, and panel are labeled Aslab, Agrid, and Apanel, respectively. Figure 17 depicts the unbraced 282 ceiling, which was supported only by hanger wires. When the hanger wires were in tension 283 (case 1), the accelerations of the slab, grid and panel were the same. However, when the 284 hanger wires were loose (case 2), which could have been initiated by downward slab 285 acceleration of more than 1g while the panels and grid system were limited to a maximum of 286 1g downward acceleration, the slab acceleration differed from that of the panels and grid 287 system. Since the panel and grid system had almost the same acceleration, the panels 288 remained in place between the grid members and the probability of dislodging panels was 289 low. 290 Figure 18 depicts the braced ceiling with compression posts at regular intervals. Due to 291 the constraint imposed by the compression posts, the entire system (slab, grid, and panels) 292 generally moved together with equal accelerations, as depicted in case 1. However, during 293 downward slab acceleration of more than 1g, the grid system moved with the slab (assuming 294 the compression posts are rigid) at the compression post locations while the panels were 295 limited to 1g downward acceleration. As a result, the slab and grid accelerations differed 296 from the panel accelerations, causing a gap to form between the ceiling grid and panels. 297 Once the gap formed, the ceiling panels were no longer constrained by the horizontal forces 298 of the grid system, and hence the panels tended to “pop out” of the grid. Furthermore, the 299 ceiling panels impacted the grid system when they fell, weakening the grid members. Gilani 300 et al. (2010) observed this type of damage in ceiling component tests, but concluded that the 301 damage was atypical of that observed in the field due to large vertical accelerations 302 experienced at the roof. This study confirms that the vertical acceleration and associated 13 303 damage pattern can be observed in a realistic building system subjected to significant vertical 304 input shaking. 305 Ceiling Perimeter Attachment Damage 306 Figure 19 shows minor damage observed at the unattached joints between grid members 307 and wall molding in the first 3D-Northridge-Rinaldi simulation. The mechanism is 308 interpreted as follows: when the grid member moved away from the wall, the grid member 309 lost contact with the wall molding (Fig. 20(a)). Since the middle slot was large relative to the 310 screw dimensions, the grid member settled slightly due to a combination of vertical 311 movement of the grid member, rotation of the seismic clip over its attachment point, and 312 popping out of the middle screw (Fig. 20(b)). As the settled grid member moved back toward 313 the wall, it hit the wall molding to cause the observed damage (Fig. 20(c)). This damage 314 could perhaps be avoided by increasing the width, and hence seat length, of the wall molding. 315 Note that ASTM E580/E580M-11ae1 (ASTM, 2011) permits the use of either 22-mm (7/8- 316 in) or 51-mm (2-in) wall molding to support seismic clips, and therefore, the tested design 317 met code requirements. 318 Fragility Methodology 319 Due to the large numbers of applied motions, a fragility curve methodology was used to 320 interpret and extend the test results to assess the seismic vulnerability of the CPP systems. 321 The experimental results were used to estimate the seismic demands, or engineering demand 322 parameter (EDPs), on CPP systems. Seismic fragility curves are conditional probability 323 statements about the vulnerability of a system under the seismic loading. Vulnerabilities are 324 generally expressed in terms of damage states that are physically meaningful in terms of 325 repair (cost and/or time) and the system functionality, and the fragility statement shows the 326 probability that the seismic demand exceeds a threshold capacity associated with the damage 14 327 state. The conditioning parameter of these probabilistic statements is often a single seismic 328 intensity measure (IM) (e.g. horizontal peak floor acceleration or PFA, as used in this study). 329 Seismic fragility curves can be represented by a lognormal cumulative distribution function 330 (Nielson and DesRoches, 2007) as: ln( Sd / Sc ) P EDP DS | IM 2 2 C d IM 331 332 (1) 333 where Sd is the median seismic demand estimate as a function of IM, Sc is the median 334 estimate of the damage state capacity, βd|IM is the logarithmic standard deviation of the 335 demand estimate, βc is the dispersion of the damage state capacity, and Φ[·] is the standard 336 normal cumulative distribution function. 337 The fragility study reported here highlights the relation between seismic performance, 338 wall molding width, and clearance (seismic gap) of the grid members from the wall molding. 339 To do so, the relative ceiling-partition displacements demands were conditioned on the 340 experimentally observed PFA of the associated floor level. A regression analysis of this data 341 was used to estimate the parameters Sd and βd|IM of the probabilistic seismic demand models 342 according to (Cornell et. al., 2002): 343 (2) 344 (3) 345 where di is the peak demand at the ith floor. In developing the fragility curves, all test data 346 obtained from the unbraced ceiling was considered, while only the first five applied motions 347 for the braced ceiling were included. 15 348 The maximum observed forward or reversing relative displacement between the ceiling 349 and partition was evaluated with respect to PFA for each horizontal direction. The 350 displacement versus PFA trends are shown in Fig. 21 for both unbraced and braced ceilings 351 on a log-log scale along with regression lines. Figure 21(a) compares all observed data and 352 fitted curves for both ceilings, while Fig. 21(b) compares All Data and Limited Data 353 (undamaged only) and associated curves for the braced ceiling. Table 2 presents regression 354 parameters a and b along with the dispersion βd|IM . 355 As mentioned previously, the grid members were installed with a clear space of 19 mm 356 (3/4 in.) from the partition walls on the floating side, which left a travel distance of only 3 357 mm (1/8 in) before the grid members unseat from the wall angle. To identify the optimum 358 clearance of grid members from the partition walls, three different limit states were defined. 359 These limit states were classified according to their clearance [19 mm (3/4 in), 16 mm (5/8 360 in), and 13 mm (1/2 in)] between the grid members and partition walls, where 19 mm (3/4 in) 361 represents the tested condition. Each limit state was defined by the median displacement 362 for pounding (equal to the wall clearance) or grid unseating (computed as 22 mm or 363 7/8 in minus the clearance). A constant value of 0.4 was assigned to βC (Table 3), as 364 frequently used for nonstructural components (FEMA, 2012). 365 The fragility curves for the limit states defined above were obtained from Equation (1) 366 using the data of Fig. 21 as demands and the limit state displacements as capacities, and are 367 plotted in Fig. 22 for both the unbraced and braced ceiling configurations. Figure 22(a) 368 implies that the unbraced ceiling is slightly more vulnerable than the braced ceiling. 369 However, Fig. 22(b) shows that the braced ceiling based on Limited Data was more fragile 370 than the braced ceiling based on All Data. Therefore, the differences in vulnerability of the 16 371 braced ceiling may have been influenced by the effect of ceiling repair in the All Data case or 372 insufficient data points in the Limited Data case. 373 The median and logarithmic standard deviation (dispersion) of each fragility curve from 374 Fig. 22 is tabulated in Table 4. The results obtained for the tested condition – a 19 mm (3/4 375 in) clearance between the grid ends and the partitions and a 3 mm (1/8 in) seat length – were 376 considered as the reference point. The fragility data suggests that for a constant length wall 377 molding, reducing the clearance and simultaneously increasing the seat length reduces the 378 vulnerability of the ceiling. Since unseating has a greater probability of occurrence than 379 pounding, increasing the seat length from 3 mm (1/8 in) to 6 mm (1/4 in) and 10 mm (3/8 in) 380 increases the median PFA of the unseating damage state by 77% and 149%, respectively, in 381 the unbraced ceiling and by 93% and 185%, respectively, in the braced ceiling (All Data). 382 Increasing the seat length at the expense of clearance increases the probability of pounding. 383 However, increasing the overall width of the wall molding reduces the probability of 384 unseating without any increase in the probability of pounding. As an example, use of 32 mm 385 (1.25 in) wide wall molding with 19 mm (3/4 in) clearance and 13 mm (1/2 in) seat length 386 achieves the lowest probability of damage for both pounding and unseating (Fig. 22). 387 Seismic Response of Sprinkler Piping 388 Using a similar approach to that reported for the ceiling system, the accelerations 389 recorded on the piping systems and the amplification factors of the piping acceleration with 390 respect to column accelerations were evaluated and compared to the code component 391 amplification factors. Table 5 tabulates acceleration amplification factors for sensors 392 mounted on the main runs, branch lines, and sprinkler heads relative to sensors mounted on 393 the columns, with slab to column amplification included for reference. The median horizontal 394 amplification observed during the experiment was 2.6 and 2.28 for the 5th and 6th floor main 17 395 runs relative to the columns, respectively, which is comparable to the code component 396 amplification factor (ap = 2.5). However, these amplification factors were increased to 5.7 397 and 4.9 on the branch lines and 6.4 and 5.5 on sprinkler heads of the 5th and 6th floor piping 398 system, respectively. The responses of the branch lines, armovers, and drops were further 399 amplified since the acceleration of the main runs (already amplified relative to the structural 400 acceleration by a factor of 2.5) served as input excitation to the branch lines. In general, the 401 largest horizontal acceleration was observed in the main run perpendicular direction at the 402 middle compared to other main run locations. The transition from loose to tension wire 403 restrainers induced large pulse accelerations in the direction of restrainers. 404 The median vertical acceleration amplification relative to the column sensors was greater 405 than 4 for the main runs on both floors, 6.3 for the 5th floor branch line, and 4.9 for the 6th 406 floor branch line. Thus, the observed vertical amplification factors were more than twice ap. 407 On the other hand, the median vertical amplification of the slab relative to the column sensors 408 was 3 and 3.9 for the 5th and roof slabs, respectively, which implies that much of the piping 409 acceleration amplification relative to the columns can be attributed to structural slab 410 flexibility. Thus, design of the piping supporting elements and anchorages based on a vertical 411 component amplification factor ap = 2.5 may be inadequate. 412 Damage Near Sprinkler Heads 413 Wherever rigid drop pipes were used, the ceiling panels sustained damage from pounding 414 of the sprinkler heads regardless of whether the oversized gap configuration, which 415 conformed to code requirements (ASTM, 2011), or the no gap configuration was used. 416 During the XY-Tohoku-Iwanuma excitation/fixed-base configuration with PFA = 1.12g, up 417 to 203 mm (8 in) of material was knocked out of the ceiling panel (Fig. 23(a)-(b)), which is 418 much larger than the 51 mm (2 in) gap required by code. Tearing of ceiling panels due to the 18 419 piping interaction is a function of maximum piping movement as ceiling panels are 420 composed from very weak fibers. Therefore, this type of damage might not be sensitive to the 421 duration of a motion. However, 203 mm (8 in) tearing of ceiling panels was observed during 422 XY-Tohoku-Iwanuma excitation, which is a strong motion with a very long duration. On the 423 other hand, no damage was observed around the flexible hose fittings that were mounted at 424 the end of Branch Line 3 (Fig. 23(c)). 425 Similar to the ceiling-partition relative displacement, a fragility procedure was used here 426 to study the probability of ceiling-sprinkler head interaction. The peak ceiling-piping relative 427 displacement was determined as the peak relative displacement between the piping and 428 ceiling systems, both of which were recorded by displacement transducers. A regression 429 analysis of the relative displacement with respect to PFA in the direction perpendicular to the 430 main run was performed using Equation (1). Figure 24 plots the ceiling-piping (sprinkler 431 head) displacement demand on a log-log scale along with the regression curve. As shown in 432 this figure, higher heteroscedasticity was observed between the demands and IMs at PFAs 433 smaller than 0.1g. While linear regression was applied to the corresponding data to estimate 434 the median response, alternative approaches such as weighted linear least squares regression 435 can be used to accommodate such a non-constant variability (heteroscedasticity) (Ang and 436 Tang, 1975). 437 Several limit states have been defined herein to interpret the seismic response of ceiling- 438 piping interaction during the experiment (Table 6). These limit states were defined by 439 (median clearance of ceiling panel and sprinkler head) and βC (logarithmic standard deviation 440 of this clearance). Three values of 3 mm (0.1 in), 13 mm (0.5 in), and 25 mm (1 in) were 441 considered as the clearance limit states ( 442 "no gap"), 25 mm (1 in), and 51 mm (2 in) diameter of ceiling panel holes around the ), which correspond to the 6 mm (0.2 in) (named 19 443 sprinkler heads. A value 0.4 was assigned to βC as before. The fragility curves for ceiling- 444 piping interaction computed by Equation (1) are shown in Fig. 25. Table 7 lists the median 445 and dispersion of the fragility curve for each limit state. 446 The fragility curves show that during this experiment, even a 25 mm (2 in) oversized ring 447 around the sprinkler heads was not sufficient to prevent ceiling-piping interaction for floor 448 accelerations exceeding 0.65g. The experiment showed that use of a flexible hose drop was 449 effective in eliminating this type of damage. Decreasing the spacing of lateral sway braces, 450 currently spaced no more than 12 m (40 ft) (NFPA, 2011), might reduce the ceiling-piping 451 interaction. Decreasing the spacing of braces along the main runs can also limit the piping 452 deflection due to bending and subsequently reduce the ceiling-sprinkler head interaction. 453 Permanent Rotation of Armover Drops 454 A vulnerability of armover drops compared to straight drops was observed in these 455 experiments. During the first 3D-Northridge-Rinaldi excitation (input vertical peak table 456 acceleration = 1.2g and resultant peak slab vibration = 6.4g), the entire 6th floor Branch Line 457 1 with three armover drop pipes twisted around its connection point to the main run (Fig. 458 26(a)). During vertical acceleration, a vertical inertia force was generated proportional to the 459 mass of the armover drop. The twisting moment around the branch line was the summation of 460 the torque generated at each drop (Fig. 26(b)). The current code (NFPA, 2011) permits the 461 connections along this branch line to be designed without torsional resistance when the 462 cumulative horizontal length of the unsupported armover < 610 mm (24 in), which was true 463 for the experimental test setup. However, the torsional resistance of the threaded joints was 464 not sufficient to resist the cumulative torsional demand generated in the large vertical 465 excitation, and permanent twisting of the branch line was observed. 20 466 Next, the twisting moment of the threaded connection of the armover branch to the main 467 run in the 6th floor piping system was estimated by a simple calculation. Based on videos 468 recorded during the experiment, the branch line was observed to twist at ~ 9.5-10.5 sec into 469 the excitation. The rotational acceleration of each armover drop was estimated from the 470 relative acceleration between the corresponding sprinkler head and the closest point on the 471 branch line, which happened to be the main run (Fig. 12(a)). The relative acceleration history 472 of each armover is plotted in Fig. 27. The instantaneous peak of the relative acceleration 473 summed over the three armovers was 6.51g at 10.37 sec. The corresponding rotational 474 acceleration = 210 rad/s2 was used to compute the twisting moment (T) that triggered the 475 rotation of branch line around its connection to the main run: T I 476 (4) 477 where I is the effective rotational inertia of armover, drop pipes, and sprinkler head masses 478 about the branch line, assuming a lumped mass approximation. The average twisting moment 479 resistance per unit (circumferential) length is defined as follows: sT 480 T D0 (5) 481 where D0 is the outside diameter of branch line pipe. The moment was estimated as 0.05 kN- 482 m (0.4 kip-in), leading to sT = 4 kN-m/m (0.08 kip-in/in). If sT is assumed to be independent 483 of the threaded pipe diameter, Equation (5) approximates the maximum torsional resistance 484 of the piping. 485 Seismic Response of Partitions 486 Past component-level testing suggests that drift-related damage to partition walls occurs 487 at median story drifts of 0.3% (Retamales et al., 2012). Maximum drifts observed during the 488 experiment were 0.78% on the 4th story and 0.62% on 5th story, which occurred during the 21 489 3D-Northridge-Rinaldi excitation on the fixed-base configuration. Drift-related damage to the 490 partition walls was not noticeable. 491 On the other hand, atypical damage states were observed that were attributed to the 492 relative vertical acceleration between the floors. After the first 3D-Northridge-Rinaldi 493 excitation, large diagonal and vertical cracks appeared on the gypsum wallboards of full 494 height partitions (Fig. 28(a)). Also, at a few locations on the 5th story partitions, which 495 utilized slip track connections, the top of the studs were observed to move laterally or “pop 496 out” permanently from their constrained position within the top tracks (Fig. 28(b)). This 497 damage state, which – to our knowledge – has not been observed prior to this experiment, 498 was due to the large differential movement between the top track (attached to the roof slab) 499 and the studs (mounted on the 5th floor slab). 500 Damage to bulkhead and partial height self-standing partitions exceeded that of the full 501 height partitions. During the first 3D-Northridge-Rinaldi excitation, the studs of bulkhead 502 partitions buckled, again due to the differential acceleration between the 5th floor and roof 503 slabs (Fig. 29(a)). Bulkhead partitions are not common in practice, but were designed to meet 504 test constraints. Also, during the final motion of the test program, the top portion of the door 505 opening separated from the adjoining wall, and the bottom connection of the track to the slab 506 failed (Fig. 29(b)). These failures were likely due to the cumulative pounding against the 507 walls of loose objects from inside the rooms. Table 8 lists the out-of-plane horizontal 508 acceleration amplification of the full height partition walls compared to the slabs that they 509 were mounted on or hung from, and amplification of partial height self-standing partitions 510 relative to the slabs below. The data for full height partitions was categorized into full 511 connections (Full) and slip track connections (Slip) with respect to slab below (Below) or 512 slab above (Above). Partial height partitions were categorized based on their institutional and 22 513 commercial details. Amplification factors were computed for each relevant ground excitation 514 along with statistics (maximum, minimum and median). 515 The median amplification of the partition out-of-plane acceleration observed during the 516 experiment was 2.7 and 2.8 for partitions with full connections, and 3.8 and 3 for partitions 517 with slip track connections relative to the below and above slab, respectively. These values 518 are slightly higher than the code component amplification factor (ap = 2.5), especially for slip 519 track partitions. As mentioned previously, the gypsum board and steel studs of slip track 520 partitions were not attached to the top tracks, which caused their out-of-plane acceleration to 521 be higher than those with full connection details. These results also suggest that ceiling tests 522 need to incorporate the flexible partition wall boundaries to represent realistic ceiling 523 behavior. Experimental ceiling specimens that are attached to rigid boundaries cannot 524 account for the influence of flexible partition walls on overall ceiling system performance. 525 The median acceleration amplification factor of the partial height partitions was > 5, 526 which exceeded that of full height partitions. This higher amplification could have been 527 caused by pounding of the unrestrained objects against the walls. The observed amplification 528 factors suggest that current code factors for attachment and anchorage design forces for 529 partial height partitions should be reconsidered. 530 Summary and Conclusions 531 532 The major findings of this experiment are summarized below: In the horizontal direction, the median CPP component acceleration amplification 533 factors (peak component acceleration relative to corresponding floor acceleration) 534 observed during the experiments were closely correlated to the recommended code 535 factor ap. 536 The component amplification factors in the vertical direction observed during the 23 537 experiment were much greater than the recommended code factor, which can 538 primarily be attributed to flexibility of the floor slabs. 539 Use of lateral bracing with compression posts may not improve the seismic response 540 of the ceiling when subjected to strong vertical excitation. Further research is needed 541 to confirm the generality of this observation. 542 Seismic clips were used around the ceiling perimeter in conjunction with reduced 543 width 22 mm (7/8 in) wall molding and a 19 mm (3/4 in) grid-partition clearance. 544 This detail was found to be vulnerable to unseating followed by pounding due to load 545 reversal. Although the seismic clip detail is expected to be an improvement over 546 standard detailing, increasing the width of wall molding and/or or reducing the grid- 547 partition clearance may alleviate the behavior observed in the experiment. 548 Due to the twisting moment generated from the armover drops, branch line pipes with 549 several unsupported armover drops can twist around the branch line threaded 550 connection point. A simple approach was proposed to estimate the maximum 551 torsional resistance as a function of threaded pipe diameter. 552 The oversized gap configuration with 51 mm (2 in) ring was not effective to prevent 553 damage to ceiling panels resulting from sprinkler head pounding; however, the use of 554 flexible hose drops substantially reduced the piping-ceiling interaction. 555 These observed response mechanisms may be sensitive to specific circumstances of 556 the experiments, such as building configuration (e.g. slab vibration characteristics) 557 and acceleration demands (e.g. large vertical acceleration relative to horizontal 558 acceleration). 559 Acknowledgements 24 560 This material is based upon work supported by the National Science Foundation under 561 Grant No. CMMI-0721399. This GC project to study the seismic response of nonstructural 562 systems is under the direction of M. Maragakis from the University of Nevada, Reno and Co- 563 PIs: T. Hutchinson (UCSD), A. Filiatrault (UB), S. French (G. Tech), and B. Reitherman 564 (CUREE). Any opinions, findings, conclusions or recommendations expressed in this 565 document are those of the authors and do not necessarily reflect the views of the sponsors. 566 The input provided by the NEES Nonstructural Project Practice Committee, composed of W. 567 Holmes (Chair), D. Allen, D. Alvarez, and R. Fleming; by the Advisory Board, composed of 568 R. Bachman (Chair), S. Eder, R. Kirchner, E. Miranda, W. Petak, S. Rose and C. Tokas, has 569 been critical for the completion of this research. The authors recognize and thank the 570 following companies for providing product donations and technical support: USG Building 571 systems, Victaulic, Tolco, Hilti, Allan Automatic Sprinkler and CEMCO steel. 572 References 573 American Society of Civil Engineers (ASCE), (2010). “Minimum Design Loads for Buildings and 574 Other Structures”, ASCE Standard ASCE/SEI 7-10, Reston, VA, USA. 575 American Society for Testing and Materials (ASTM), (2011). E580/E580M-11ae1: Standard Practice 576 for Installation of Ceiling Suspension Systems for Acoustical Tile and Lay-in Panels in Areas 577 Subject to Earthquake Ground Motions. ASTM International, Volume 04.06. 578 579 580 581 Ang AHS., Tang WH. (1975). “Probability Concepts in Engineering Planning and Design. Volume IBasic Principles”. New York: John Wiley. Badillo-Almaraz, H., Whittaker, A., Reinhorn, A. (2007). “Seismic Fragility of Suspended Ceiling Systems”, Earthquake Spectra, Earthquake Engineering Research Institute, 23(1):23-40. 582 Cornell, A. C., Jalayer, F., Hamburger, R. O., and Foutch, D. A. (2002). “Probabilistic basis for 2000 583 SAC Federal Emergency Management Agency steel moment frame guidelines”, J. Struct. Eng. 584 128, 526–532. 25 585 586 587 588 FEMA (2011). “Reducing the Risks of Nonstructural Earthquake Damage: A Practical Guide”, FEMA E-74, Federal Emergency Management Agency, Washington, D.C. FEMA, (2012). “Seismic Performance Assessment of Buildings, Volume 2 – Implementation Guide”, FEMA P-58-2, Federal Emergency Management Agency, Washington, D.C. 589 Gilani, A. S. J., Reinhorn, A. M., Glasgow, B., Lavan, O., Miyamoto, H. K. (2010). “Earthquake 590 Simulator Testing and Seismic Evaluation of Suspended Ceilings”, Journal of Architectural 591 Engineering, ASCE, 16(2):63-73. 592 593 594 595 596 597 ICC Evaluation Service (2010). AC 156 Acceptance Criteria for Seismic Certification by Shake Table Testing of Nonstrucutral Components, ICC Evaluation Service. Miranda, E., (2003). “Building Specific Loss Estimation for Performance Based Design”, 2003 Pacific Conference on Earthquake Engineering, Christchurch, New Zealand. Miranda, E., Mosqueda. G., Retamales, R., and Pekcan, G. (2012). “Performance of Nonstructural Components during the February 27, 2010 Chile Earthquake”, Earthquake Spectra, 28, 453-471. 598 Mizutani, K., Kim, H., Kikuchihara, M., Nakai, T., Nishino, M., Sunouchi, S. (2012). “The damage of 599 the building equipment under the 2011 Tohoku pacific earthquake”, 9th International Conference 600 on Urban Earthquake Engineering & 4th Asia Conference on Earthquake Engineering, Tokyo 601 Institute of Technology, Tokyo, Japan. 602 603 604 605 National Fire Protection Association (NFPA), (2011). NFPA 13: Standard for the Installation of Sprinkler Systems." National Fire Protection Association, 2010 Edition, Quincy, MA. Nielson, G. B. and DesRoches, R. (2007). “Analytical Seismic Fragility Curves for Typical Bridges in the Central and Southeastern United States”, Earthquake Spectra, 23(3), 615-633. 606 Retamales, R., Davies, R., Mosqueda, G., Filiatrault, A. (2012). “Experimental Seismic Fragility of 607 Cold-Formed Steel Framed Gypsum Partition Walls”, Journal of Structural Engineering, ASCE, 608 139, Special Issue: NEES 2: Advances in Earthquake Engineering, 1285-1293. 26 609 Ryan, K. L., Soroushian S., Maragakis, E. M., Sato, E., Sasaki, T., Okazaki, T. (2013a). “Seismic 610 simulation of integrated nonstructural systems at E-Defense, Part 1: Influence of 3D structural 611 response and base isolation”, Under Review in J. Struct. Eng. 612 Ryan K, Sato E, Sasaki T, Okazaki T, Guzman J, Dao N, Soroushian S, Coria C (2013b). "Full Scale 613 5-story Building with Triple Pendulum Bearings at E-Defense", Network for Earthquake 614 Engineering Simulation (database), Dataset, DOI:10.4231/D3X34MR7R. 615 Ryan K, Sato E, Sasaki T, Okazaki T, Guzman J, Dao N, Soroushian S, Coria C (2013c). "Full Scale 616 5-story Building with LRB/CLB Isolation System at E-Defense", Network for Earthquake 617 Engineering Simulation (database), Dataset, DOI:10.4231/D3SB3WZ43. 618 Ryan K, Sato E, Sasaki T, Okazaki T, Guzman J, Dao N, Soroushian S, Coria C (2013d). "Full Scale 619 5-story Building in Fixed-Base Condition at E-Defense", Network for Earthquake Engineering 620 Simulation (database), Dataset, DOI:10.4231/D3NP1WJ3P. 621 Taghavi, S., and Miranda, E. (2003). “Response Assessment of Nonstructural Building Elements”, 622 PEER Rep. 2003/05, Pacific Earthquake Eangineering Research Center (PEER), Univ. of 623 California, Berkeley, CA. 624 Tian, Y., Filiatrault, A., Mosqueda, G. (2013). “Experimental Seismic Study of Pressurized Fire 625 Sprinkler Piping Subsystems”, Technical Report MCEER-13-0001, Multidisciplinary Center for 626 Earthquake Engineering Research, State University of New York at Buffalo, Buffalo, NY, USA, 627 2013. 628 Whittaker, A.S. and Soong, T.T., (2003). “An Overview of Nonstructural Component Research at 629 Three U.S. Earthquake Engineering Research Centers”, in Proceedings of Seminar on Seismic 630 Design, Performance, and Retrofit of Nonstructural Components in Critical Facilities, Applied 631 Technology Council, ATC-29-2, pp. 271-280, Redwood City, California. 632 Zaghi, A. E., Maragakis, E. M., Itani, A., and Goodwin, E. (2012). “Experimental and Analytical 633 Studies of Hospital Piping Subassemblies Subjected to Seismic Loading”, Earthquake Spectra, 634 Earthquake Engineering Research Institute. 28(1):367-384. 635 27 636 637 Tables Table 1. Maximum and Minimum Ceiling to Column or Slab to Column Acceleration Amplification 5th Floor Motion Number 1 2 3 4 5 ... 41 Max=* Min=* Median* Grid Column Slab Column 6th Floor Panel Column Grid Column Slab Column Panel Column XY Z Z XY Z XY Z Z XY Z 1.02 1.25 2.55 3.27 3.00 ... 1.37 7.19 1.00 2.57 0.94 1.71 5.14 4.82 2.23 ... 4.41 6.25 2.09 4.24 1.00 1.13 4.51 4.26 2.45 ... 1.25 4.95 1.50 2.97 1.00 1.16 3.25 3.23 3.72 ... NA 6.57 0.97 2.49 0.93 1.01 5.61 3.72 2.26 ... NA 5.61 1.94 3.58 1.03 1.15 5.75 4.52 4.27 ... 2.29 5.75 1.03 2.83 1.31 1.43 5.27 4.67 2.03 ... 6.44 5.27 4.67 4.97 1.08 1.69 6.46 4.88 2.14 ... 1.35 6.46 4.88 5.67 1.01 1.08 7.70 5.88 16.78 ... NA 7.70 1.01 3.48 1.83 2.89 12.8 10.9 2.54 ... NA 12.8 10.9 11.9 638 639 640 641 Table 2. Demand Parameter Estimations for Ceiling Displacements System 5th Floor Ceiling a 4.83 b 1.20 βd|PFA 0.53 6th Floor Ceiling-All 6th Floor Ceiling-Limited 4.32 7.37 1.05 1.40 0.72 0.78 Table 3. Limit States of Grid Perimeter Pounding or Unseating (mm) Ceiling System (5th and 6th Floor) 19 mm (3/4 in) Clearance 16 mm (5/8 in) Clearance 13 mm (1/2in) Clearance Pounding Unseating θdisp βc θdisp βc 19 16 13 0.4 0.4 0.4 3 6 10 0.4 0.4 0.4 642 643 644 645 Table 4. Medians and Dispersions for 3 Different Ceiling Perimeter on the Floating Side System 3/4" Clearance Median PFA(g) Dispersion 5th Floor Ceiling-Pounding 5th Floor Ceiling-Unseating 6th Floor Ceiling- Pounding 6th Floor Ceiling-Unseating 3.16 0.71 4.00 0.74 0.66 0.66 0.82 0.82 6th Floor Ceiling- Pounding 6th Floor Ceiling-Unseating 1.95 0.54 0.88 0.88 5/8" Clearance Median Difference PFA(g) Dispersion (%) All Data Points 2.71 0.66 -14 1.26 0.66 77 3.43 0.82 -14 1.43 0.82 93 Limited Data Points 1.71 0.88 -12 0.89 0.88 65 646 28 Median PFA(g) 1/2" Clearance Difference Dispersion (%) 2.25 1.77 2.78 2.11 0.66 0.66 0.82 0.82 -29 149 -31 185 1.46 1.19 0.88 0.88 -25 120 Table 5. Maximum Piping/Floor Acceleration Amplification 5th Floor Piping Motion Number 1 2 3 4 5 ... 41 Max=* Min=* Median=* 647 648 Main Column Slab Column 6th Floor Piping Branch Column Head Column Main Column Slab Column Branch Column Head Column XY Z Z XY Z XY XY Z Z XY Z XY 1.11 1.78 4.37 4.14 3.21 ... 1.98 7.74 0.97 2.57 1.66 2.17 5.78 6.12 2.45 ... 15.9 6.91 2.45 4.44 1.00 1.13 4.51 4.26 2.45 ... 1.25 4.95 1.50 2.97 1.37 2.21 14.8 12.1 7.87 ... 3.43 14.9 1.37 5.74 1.19 1.49 11.48 10.20 2.26 ... 38.12 11.85 2.26 6.31 1.38 3.25 15.10 16.17 5.24 ... 2.76 16.59 1.38 6.38 1.08 1.72 3.05 3.80 3.40 ... 2.28 4.71 1.08 2.28 2.11 2.36 8.02 11.9 3.79 ... 23.5 12.0 2.50 4.16 1.08 1.69 6.46 4.88 2.14 ... 1.35 6.46 1.55 3.88 1.19 1.98 11.17 10.50 8.30 ... 3.85 14.35 1.19 4.91 1.39 2.00 11.9 9.10 2.22 ... 54.7 11.9 2.22 4.85 1.43 2.46 19.48 12.22 7.37 ... 4.58 19.48 1.44 5.45 * Damaged Pipes (Highlighted numbers, first branch line data) were not included. * Only 3D motions were considered for z direction calculation Table 6. Limit States of Ceiling-Piping Interaction Clearance Ceiling - Sprinkler Head 3 mm (1/10 in) Clearance 13 mm (1/2 in) Clearance 25 mm (1 in) Clearance θint βc 3 13 25 0.4 0.4 0.4 649 650 651 Table 7. Medians and Dispersions of Ceiling-Piping Interaction 3 mm (1/10 in) Clearance Median PFA(g) Dispersion 0.07 0.77 1.10 1.34 13 mm (1/2 in) Clearance Median PFA(g) Dispersion 0.33 0.77 3.55 1.34 25 mm (1 in) Clearance Median PFA(g) Dispersion 0.64 0.77 N/A* 1.34 * Estimated median values are much larger than can be appropriately extrapolated from regression analyses. 652 653 654 Table 8. Maximum Partition/Floor Acceleration Amplification Partial Height Self Standing Partitions Full Height Partitions Motion Number 1 ... 34 35 ... 41 Max=* Min=* Median=* Full Below Full Above Slip Below Slip Above Commercial Institutional 1.51 ... 5.08 3.14 ... 1.40 6.35 1.16 2.62 1.15 ... 4.91 2.29 ... 1.16 7.43 1.00 2.79 1.87 ... 8.40 2.24 ... 1.24 8.40 1.17 3.77 1.31 ... 7.63 1.93 ... 1.11 8.09 0.92 2.95 3.34 ... 7.69 4.75 ... 3.58 13.69 1.40 5.12 3.66 ... 7.00 3.41 ... 3.35 15.54 2.75 5.75 * Damaged partitions (highlighted numbers) were not considered. 655 29 656 657 List of Figures 658 Figure 1. Overall view of ceiling system layout 659 Figure 2. Connection of runners/cross tees and wall molding: (a) attached detail, and (b) 660 unattached 661 Figure 3. Connection of seismic bracing to the ceiling grid 662 Figure 4. Overall plan view of piping system 663 Figure 5. Sprinkler heads and drops: (a) flexible drop, (b) 50 mm (2 in) oversized gap 664 configuration, (c) no gap configuration 665 Figure 6. Bracing for piping system: (a) lateral and longitudinal brace near riser, and (b) 666 diagonal splay wires and pipe hanger at the end of each branch line 667 Figure 7. Overall partition plan view; black labels: 4th story, gray labels: 5th story 668 Figure 8. Typical institutional (a) T and (b) corner; and commercial (c) T and (d) corner 669 connections 670 Figure 9. Typical institutional connection details for full connection: (a) top, (b) bottom; and 671 slip track connection: (c) top, (d) bottom 672 Figure 10. (a) Partial height self-standing partitions, (b) bulkhead partitions 673 Figure 11. (a) Summary of accelerometers on ceiling system (plan view); accelerometer on: 674 (b) ceiling grid, (c) ceiling panel, (d) compression post; (e) displacement transducer at ceiling 675 perimeter 676 Figure 12. Instrumentation view: (a) plan view of accelerometers placed on piping system; 677 accelerometer on: (b) sprinkler head, (c) branch line; (d) displacement transducer on main run 678 Figure 13. Accelerometers installed on (a) full height partitions, (b) partial height partitions 679 Figure 14. Vertical acceleration histories recorded at slab, ceiling, and column locations in 680 roof SE slab during 3D-Superstition Hills-Westmorland (TP configuration) 30 681 Figure 15. Condition of (a) 6th floor braced ceiling and (b) 5th floor unbraced ceiling after 682 3D-Northridge-Rinaldi (TP configuration) 683 Figure 16. Vertical acceleration in panel (C9) versus grid (C4) in (a) 5th floor unbraced and 684 (b) 6th floor braced ceiling due to 3D-Superstition Hills-Westmorland (TP configuration) 685 Figure 17. Vertical dynamics of unbraced ceiling 686 Figure 18. Vertical dynamics of braced ceiling 687 Figure 19. Ceiling perimeter attachment failure after 3D-Northridge-Rinaldi (TP 688 configuration) 689 Figure 20. Grid-wall molding interaction mechanism 690 Figure 21. Ceiling-partition relative displacement seismic demand: (a) All Data for braced 691 and unbraced (b) All Data and Limited Data for braced 692 Figure 22. Ceiling perimeter fragility curves on the floating side: (a) unbraced and braced - 693 All Data, (b) braced - All and Limited Data 694 Figure 23. Comparison of pounding damage on 6th floor after XY-Tohoku-Iwanuma 695 excitation: (a) conventional sprinkler head in no gap configuration and (b) oversized gap 696 configuration; (c) flexible hose sprinkler head 697 Figure 24. Ceiling-piping relative displacement perpendicular to the main run 698 Figure 25. Ceiling-piping interaction fragility curves in direction perpendicular to main runs 699 Figure 26. (a) Armover permanent rotation following 3D-Northridge-Rinaldi (TP 700 configuration), and (b) torsional demand on armover branch line 701 Figure 27. Sprinkler head-branch line relative acceleration at armover branch line in 6th floor 702 during 3D-Northridge-Rinaldi excitation (TP configuration) 703 Figure 28. (a) Typical cracks on full height partitions, and (b) lateral movement of studs 704 from the top tracks on slip track partitions 31 705 Figure 29. (a) Stud buckling of bulkhead partitions, (b) damage to partial height self- 706 standing partitions 707 32 Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Figure Click here to download high resolution image Copyright Agreement Click here to download Copyright Agreement: Copyright Transfer - Paper 2.pdf Sizing worksheet (.xls) Click here to download Sizing worksheet (.xls): Sizing Worksheet_Paper 2.xls Journals Sizing Worksheet ***Please complete this form for all new manuscripts*** October 26, 2014 This worksheet will automatically calculate the total number of printed pages your article will occupy in the journal. Please fill in all fields in green below. If you do not know your Manuscript Number, you may leave that field blank. Technical Paper/Case Study = 8 pgs. Technical Note = 3 pgs. Manuscript number: Journal name: Corresponding author name: Email address: Length Limits: Forum = 4 pgs. Discussion/Closure = 2 pgs. STENG-3260 Journal of Structural Engineering (ASCE) Keri Ryan [email protected] Information on the maximum allowed length for each article type can be found online at: http://www.asce.org/Content.aspx?id=29559 Number of pages in your manuscript: 27 Number of figure pages: 10 - Please include figure captions when indicating the size of your manuscript. Number of table pages: 2 - Manuscripts should use 12 pt. font, double spaced, with 1 inch margins. Estimated article pages: 10 Note: The total displayed above is only an estimate. Final page count will depend on a number of factors, including the size of your figures and tables, and the number of display equations in your manuscript. Additional author resources can be found online using the ASCE Author Guide located at: http://www.asce.org/Content.aspx?id=18107 = total ms pages = # of figs = # of tables 10 10 10 0 10 7 3.5 0 11 6 2.5 0.7 9 6 3.5 0.7 10 0 10 0 0 0 0 0 1 0 0 0 1 0 0 Response to Reviewers Comments Click here to download Response to Reviewers Comments: Response-to-reviewer-comments-Paper 2.pdf Seismic Simulation of Integrated Nonstructural Systems at E-Defense, Part 2: Evaluation of Nonstructural Damage and Fragilities Reviewer #2: 1) One of the unique features of this study is that it is a system-level investigation. However, in the presentation and analysis of the nonstructural data, the results of the three tested configurations (fixed base, isolated with LRB+cross linear bearings, and isolated with TFP bearings) are all thrown into one bin. Doesn't the type of isolation system, or lack thereof, have an effect? This needs to be addressed. One purpose of Part 1 of the companion papers was to describe the structural response observed in the different configurations and its relation to the nonstructural response. Certainly, horizontal accelerations are generally different in isolated buildings compared to fixed-base buildings. However, the variation of horizontal floor accelerations observed in the study was minimized due to the need to apply different scale factors to the motions for the isolation and fixed-base configurations. More importantly, Part 1 demonstrated that the response of the CPP system (at the amplitudes observed in the experiment) were more closely tied to the vertical slab acceleration than the horizontal floor acceleration, and that slab vibration was insensitive to the presence of an isolation system. A new section titled “Influence of Isolation System on Vertical Amplification Factors” with an expanded discussion of the propagation of vertical motion through the structure has been added to strengthen this argument. As a result, the presence of the isolation system did not have a noteworthy effect on the CPP system response in this experiment. We have correlated the CPP damage to the observed horizontal and vertical accelerations recorded at the floor level, which is the best indicator of damage to those components. To clarify for the reader, we have expanded the discussion related to this issue: “The damage mechanisms and the extent of damage were very similar for the three system configurations because, as discussed in Ryan et al. (2013), similar peak demands were observed in the three system configurations and vertical slab vibration was insensitive to the presence of an isolation system. Thus, the damage observations are presented and discussed without further mention of the system configuration during which they were observed, but rather interpreted relative to the horizontal floor acceleration and vertical slab acceleration to which they are closely correlated.” (lines 193 to 199). 2) As for part 1, the title of paper needs to be more specific. What types of nonstructural systems are investigated? A correction has been made. The title of paper was changed to: “Seismic Simulation of an Integrated Ceiling-Partition Wall-Piping System at E-Defense, Part 2: Evaluation of Nonstructural Damage and Fragilities” 3) The first paragraph of the Introduction needs references. The following references have been added to the text: FEMA E-74, (2011). “Reducing the Risks of Nonstructural Earthquake Damage: A Practical Guide”, Federal Emergency Management Agency, Washington, D.C. Miranda, E., (2003). “Building Specific Loss Estimation for Performance Based Design”, 2003 Pacific Conference on Earthquake Engineering, Christchurch, New Zealand. Taghavi, S., and Miranda, E. (2003). “Response Assessment of Nonstructural Building Elements”, PEER Rep. 2003/05, Pacific Earthquake Eangineering Research Center (PEER), Univ. of California, Berkeley, CA. Whittaker, A.S. and Soong, T.T., (2003). “An Overview of Nonstructural Component Research at Three U.S. Earthquake Engineering Research Centers”, in Proceedings of Seminar on Seismic Design, Performance, and Retrofit of Nonstructural Components in Critical Facilities, Applied Technology Council, ATC-29-2, pp. 271-280, Redwood City, California. 4) There is a logical gap between the last two sentences of the first paragraph in the Introduction. Something needs to be said about the value of structural vs nonstructural components. A correction has been made. The following sentence has been added: “Also, nonstructural systems almost always represent the major portion of the total investment in buildings (Whittaker and Soong, 2003).” 5) Line 165. Many types of nonstructural components have high natural frequencies. Is a 25 Hz cutoff frequency appropriate for the Butterworth filter? The reviewer raises a valid question. To clarify, the cutoff frequency of 25 Hz was used in the previous publications of this experiment by the authors, and therefore, the same cutoff frequency was chosen in this study to provide consistency of presented results. While 25 Hz may be on the low side for nonstructural components, it is larger than the cutoff frequency for rigid components (16.7 Hz, according to ICC-AC156), and thus encompasses the frequency range of the interest. Knowing that the use of larger cutoff frequencies may result in signals with high frequency noise, the authors decided to stay with cutoff frequency of 25 Hz to process the data. ICC Evaluation Service (2010). AC 156 Acceptance Criteria for Seismic Certification by Shake Table Testing of Nonstrucutral Components, ICC Evaluation Service. 6) Paragraph starting on line 223. A time history window of the involved responses, for a typical case, might aid the discussion. As suggested by the reviewer, a time history window of SE slab, ceiling, and columns was added to show the trend discussed in the mentioned paragraph. 7) What are the consequences of falling ceiling panels. Do they pose a credible threat to the safety of building occupants? Do they break easily? If not, can they be easily popped back into place, or is repair associated with this type of damage costly? Etc. Offering some perspective may be useful. The following statement was added to the text in order to address the reviewer’s concern: “While the chance of a life threatening injury due to the fallen ceiling panels was low, serious disruption was observed in the office room. Most of the fallen panels were damaged to the extent that replacement would be desirable. In addition, fallen ceiling panels weakened the grid system and increased the likelihood that more extensive grid repair was necessary. However, over the course of the test program, some of the cross tee sections failed but the main runners always remained intact.” 8) Line 233. For what type of building and site location are the median vertical amplification factors from this study suitable for design or evaluation? This goes back to the choice of ground motions used in this study, being criticized as uncharacteristic in the vertical direction (see review of STENG-3259). As shown in Figure 5 of Part 1 (STENG-3259), a wide range of vertical shake intensities and frequencies were included in the test program. In addition, the variation of amplification factor was found to be independent of the shaking intensity. Therefore, the median values presented in this study may not correspond to specific types of buildings or site locations. This is also consistent with the usual interpretation of ap, which is a function of component type (or their flexibility) rather than site location or building types. 9) Line 308. This sentences does not quite make sense. The relationship between the demand and the IM is expressed through a functional dependence similar to Equation 2 but with randomness attached to it. The correction has been made. Below is the modified sentence: “Seismic fragility curves can be represented by a lognormal cumulative distribution function (Nielson and DesRoches, 2007) as:” 10) The numerator in Equation 3 does not look correct. Hopefully this is just a typo. The correction has been made. It was a typo mistake. 11) Line 394 and Figure 22. The damage described and shown in this case corresponds to the Iwanuma motion, a strong motion with uncharacteristically very long duration. The motion resulted in the highest PFA and EDP. Isn't it fair to say that damage like this is atypical? Duration of motion may not be an important factor on causing damage to ceiling panels. Experimental studies by Soroushian et al. (2014) and Tian et al. (2013) showed similar damage in motions with 10 sec. and 48 sec. in durations. As shown by Soroushian et al. (2014), 8 in. tearing of ceiling panels happened during a 10 sec motion with 1.06g PFA in a full scale two-story experiment, which is very comparable to the observation of this study. Note that ceiling panels are composed from very weak fibers and this type of damage is more function of piping movement. The effect of motion amplitude on the extent of panel tearing is presented in Figure 23. The following statement is added to the paper to address the reviewer’s comment: “Tearing of ceiling panels due to the piping interaction is a function of maximum piping movement as ceiling panels are composed from very weak fibers. Therefore, this type of damage might not be sensitive to the duration of a motion. However, 203 mm (8 in) tearing of ceiling panels was observed during XY-Tohoku-Iwanuma excitation, which is a strong motion with a very long duration.” Soroushian, S., Rahmanishamsi, E., Ryu, K. P., Maragakis, E. M., Reinhorn, A.M., “A Comparative Study of Sub-System and System Level Experiments of Suspension Ceiling Systems”, Tenth U.S. National Conference on Earthquake Engineering, July 2014, Anchorage, AK. Tian, Y., Filiatrault, A., Mosqueda, G. (2013) “Experimental Seismic Study of Pressurized Fire Sprinkler Piping Subsystems,” Technical Report MCEER, State University of New York at Buffalo, NY, MCEER 13-0001. 12) Figure 23 shows that the dispersion of the data in log-log space is nonuniform. The dispersion becomes huge in the lower range of the independent variable. Ordinary least squares, which is used in this study to estimate the parameters of the regression model, is problematic in this case. The reviewer suggests that the authors look up "heteroscedasticity" in statistics literature for a description of the phenomenon and for possible ways to deal with it. The following statement has been added to address the reviewer’s comment: “ As shown in this figure, higher heteroscedasticity was observed between the demands and IMs at PFAs smaller than 0.1g. While linear regression was applied to the corresponding data to estimate the median response, alternative approaches such as weighted linear least squares regression can be used to accommodate such a non-constant variability (heteroscedasticity) (Ang and Tang, 1975).” 13) Figure 8 in Part 1 (STENG-3259) shows PFAs that exceed 2g for the fixed-base case. However, in Part 2 the highest PFA appears to be 1.1 g, which is high for an isolated building, but rather low for a fixed base building. Is the data for higher PFA omitted because it corresponds to repaired ceiling systems (line 190)? If this is the case, then the upper end of the IM used in the regression analysis is low and extrapolating to more realistic values of PFA for fixed base buildings may or may not be applicable. This need to be noted. Figure 8 in Part 1 (STENG-3259) shows the ratios of PFA/PGA that exceed 2 for the fixedbase case. This ratio is unitless and the (g) provided in the label of the x axis was included by mistake. Please note that this figure was replaced (now Figure 9) in the revised version of part 1, and now shows both absolute acceleration and acceleration ratio. 14) Lines 409-411. Unclear. The sentence referred to was rephrased as follows: “Three values of 3 mm (0.1 in), 13 mm (0.5 in), and 25mm (1 in) were considered as the clearance limit states (θint), which correspond to the 6mm (0.2in) (named "no gap"), 25 mm (1 in), and 51 mm (2 in) diameter of ceiling panel holes around the sprinkler heads.” Reviewer #3: 1) Was there any correlation between the amplification of vertical acceleration and the number of stories (i.e., how high up in the structure the accelerations are measured)? We believe the reviewer is referring to amplification of CPP component acceleration relative to floor slabs (or other appropriate measure) and not the amplification of column and slab vibration relative to the ground. The latter was discussed in Part 1 (STENG-3259) (now Figures 10 and 11). However, the amplification of CPP component response with respect to height cannot be well addressed since the nonstructural components were only installed on the top two floors. Also, as shown in new Figure 14 of this paper, vertical acceleration amplification in the nonstructural components are related to the slab vibration, which may vary more as a function of slab vibration frequency than building height. 2) The data has been processed as "All data" and "Limited data" based on the damage state. No specific conclusions were drawn based on the existence of the isolation. Does that mean that the base condition of the structure does not have any effect on the nonstructural damage? Can this be due to the order of the 6 day testing without full repair of the damages? One purpose of Part 1 of the companion papers was to describe the structural response observed in the different configurations and its relation to the nonstructural response. Certainly, horizontal accelerations are generally different in isolated buildings compared to fixed-base buildings. However, the variation of horizontal floor accelerations observed in the study was minimized due to the need to apply different scale factors to the motions for the isolation and fixed-base configurations. More importantly, Part 1 demonstrated that the response of the CPP system (at the amplitudes observed in the experiment) were more closely tied to the vertical slab acceleration than the horizontal floor acceleration, and that slab vibration was insensitive to the presence of an isolation system. A new section titled “Influence of Isolation System on Vertical Amplification Factors” with an expanded discussion of the propagation of vertical motion through the structure has been added to strengthen this argument. As a result, we believe the presence of the isolation system did not have a noteworthy effect on the CPP system response in this experiment. We have correlated the CPP damage to the observed horizontal and vertical accelerations recorded at the floor level, which is the best indicator of damage to those components. To clarify for the reader, we have expanded the discussion related to this issue: “The damage mechanisms and the extent of damage were very similar for the three system configurations because, as discussed in Ryan et al. (2013), similar peak demands were observed in the three system configurations and vertical slab vibration was insensitive to the presence of an isolation system. Thus, the damage observations are presented and discussed without further mention of the system configuration during which they were observed, but rather interpreted relative to the horizontal floor acceleration and vertical slab acceleration to which they are closely correlated.” (lines 193 to 199). The lack of full repair certainly has some effect on the vulnerability of the system but we believe it does not change the observations in a major way. Since we cannot quantify the influence of the lack of full repair, it is stated in the paper as a limitation, and care is taken not to draw conclusions based on specific components that were not repaired.
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project