Fire Science and Technology 2015 The Proceedings of 10th Asia-Oceania Symposium on Fire Science and Technology 9789811003769, 9789811003752, 9811003750, 9811003769

Citation preview

Kazunori Harada · Ken Matsuyama Keisuke Himoto · Yuji Nakamura Kaoru Wakatsuki Editors

Fire Science and Technology 2015 The Proceedings of 10th Asia-Oceania Symposium on Fire Science and Technology

Fire Science and Technology 2015

ThiS is a FM Blank Page

Kazunori Harada • Ken Matsuyama • Keisuke Himoto • Yuji Nakamura • Kaoru Wakatsuki Editors

Fire Science and Technology 2015 The Proceedings of 10th Asia-Oceania Symposium on Fire Science and Technology

Editors Kazunori Harada Kyoto University Kyoto, Japan Keisuke Himoto National Institute for Land and Infrastructure Management Tsukuba, Japan

Ken Matsuyama Tokyo University of Science Chiba, Japan Yuji Nakamura Toyohashi University of Technology Toyohashi, Japan

Kaoru Wakatsuki Shinshu University Ueda, Japan (formerly, National Research Institute of Fire and Disaster, Tokyo, Japan)

ISBN 978-981-10-0375-2 ISBN 978-981-10-0376-9 DOI 10.1007/978-981-10-0376-9

(eBook)

Library of Congress Control Number: 2016944922 # Springer Science+Business Media Singapore 2017 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. Printed on acid-free paper This Springer imprint is published by Springer Nature The registered company is Springer Science+Business Media Singapore Pte Ltd.

Foreword

The 10th Asia-Oceania Symposium on Fire Science and Technology (10th AOSFST) was held in Tsukuba, Japan, 5–7 October 2015. The symposium was jointly hosted by the Building Research Institute (BRI), the Japan Association for Fire Science and Engineering (JAFSE), and Tokyo University of Science (TUS), and was supported by the National Research Institute of Fire and Disaster (NRIFD), and the National Institute for Land and Infrastructure Management (NILIM). About 250 registrants attended three parallel sessions in which 94 peerreviewed papers and 6 invited lectures were presented. Also, one poster session was held at the symposium in which 61 posters were displayed. Delegates from 15 countries and regions joined: Australia, Canada, China, France, Hong Kong, Indonesia, New Zealand, Japan, Russia, Singapore, Sweden, South Korea, Taiwan, the United Kingdom, and the United States of America. Papers and poster abstracts were accepted on the basis of their quality and originality in fire science and technology. The opening ceremony was conducted by representatives from the host organizations and AOAFST/IAFSS: Professor Wan-Ki Chow, president of AOAFST, presented the greetings, Professor Patrick van Hees, president of IAFSS, made remarks, and local organizers gave welcome speeches. Following the opening ceremony, Professor Andrew Buchanan from the University of Canterbury delivered a plenary lecture entitled “Fire Resistance of Multi-story Timber Buildings”. Five other invited lectures were also presented during the symposium by Professors Hiroyuki Suzuki, Jinhua Sun, Yulianto Sulistyo Nugroho, Wan-Ki Chow, and Brian Meacham. Prior to the banquet, the Hirano Memorial Session was held by Professors Ritsu Dobashi, Kazunori Kuwana, and Lijing Gao. Professors Wan-Ki Chow, Bogdan Dlugogorski, and Jinhua Sun conveyed their sympathies for the great loss of one of important founders of AOAFST. At the awards reception during the banquet, Professor Naian Liu, Chair of the Symposium Awards Committee, presented the Lifetime Contribution Award to three dedicated leaders in fire science and technology in Asia-Oceania district: Professors Wan-Ki Chow, Takeyoshi Tanaka, and Yuji Hasemi. At the closing ceremony, Professor Naian Liu, Chair of the Symposium Awards Committee, presented four excellent paper awards, three excellent student paper awards, and four excellent poster awards. Professor Takeyoshi Tanaka delivered closing remarks as representative of the host country. The next AOSFST (11th AOSFST) was announced and Professor Kuang-Chung Tsai from National Kaohsiung First University of Science and Technology presented the potential date and place with appealing social events. In the final minutes, I, as the Chair of the Local Organizing Committee, conveyed my sincere thanks to all attendees in the symposium. The Local Organizing Committee would like to thank the Building Research Institute (BRI), the Japan Association for Fire Science and Engineering (JAFSE), the Tokyo University of Science (TUS), the National Research Institute of Fire and Disaster (NRIFD), and the National Institute for Land and Infrastructure Management (NILIM) for being the supportive sponsors of the symposium. Also the committee would like to express its gratitude to all the organizations, advisory board and all committee members, and other volunteers who assisted v

vi

Foreword

in making this symposium successful. A special thanks is given to Professor Kazunori Harada, Chair of the Program Committee, for organizing and leading the committee on the selection of papers and posters, and all of the Program Committee Members including track chairs/cochairs who interacted with authors and reviewers as well as reviewing final manuscripts to ensure all reviewers’ comments were properly addressed. Many thanks should go to Professor Ken Matsuyama, Chair of the Publication Committee for his great effort in checking the format of the papers and posters presented in this volume. Sincere efforts by Professor Naian Liu and the Awards Committee were highly appreciated to successfully select the best candidates for awards. Finally, all dedicated support and help contributed by members of the Logistics Committee, chaired by Dr. Tomohiro Naruse, are gratefully acknowledged.

Chair, Local Organizing Committee, 10th AOSFST Director, Department of Fire Engineering Building Research Institute, Japan

Ichiro Hagiwara, Dr. Eng.

Preface

This proceedings of the 10th Asia-Oceania Symposium on Fire Science and Technology (10th AOSFST) held in Tsukuba, Japan, October 5–7, 2015, collects selected contents of presentations at the symposium. This symposium covered a very wide range topics in the fire research field, and the final edition of the proceedings includes about 100 topics. The editors acknowledge all those involved, including authors, reviewers, and volunteers, to successfully complete the 10th AOSFST and to publish this proceedings. In addition, the editors would like to explain the review and selection process of papers and the selection process of posters in 10th AOSFST. Also, we would like to introduce and congratulate all award recipients.

Paper Submission and Selection Process The papers were solicited and selected by a peer-review process. The first call for papers was issued in June 2014, followed by second call in September 2014. Papers were submitted during October and December 2014 in the regular submission period. During this time, 84 papers were submitted. In the extended period, up to 14 January 2015, 43 more papers were submitted. All the submitted papers were allocated to 1 of 14 tracks, then subjected to review by multiple reviewers. Selection of papers was based on their quality, but priority was given to papers submitted during the regular submission period. In total, 96 papers were selected but two papers were withdrawn during the finalization process. As a result, 94 papers were selected for presentation at the symposium.

Selection of Posters Poster presentations were also accepted at the symposium. The program committee received 64 poster abstracts. As poster presentations aim to exchange ideas on ongoing research, selection was mostly based on the clarity of the contents. Of the submissions, 61 posters were accepted for presentation. The poster abstracts are not included in this book, however.

Excellent Paper Award and Excellent Student Paper Awards The award committee selected excellent papers and excellent student papers by voting. The recipients of excellent paper awards were: • Heisuke Yamashita, Toru Yoshida, and Takeo Hirashima. Influence of Water Content on Total Strain of Super-high-strength Concrete Under Elevated Temperature (Japan Testing Center for Construction Materials) vii

viii

Preface

• Yi-chul Shin, Yoshifumi Ohmiya, Shin-ichi Tsuburaya, Yuki Yoshida, Kazumasa Tashima, and Jun-ichi Suzuki. Study on Fire Plume Behavior in Vertical Shafts of Buildings (Tokyo University of Science) • Tatyana Bolshova, Sergey Yakush, Vladimir Shvartsberg, Andrey Shmakov, Oleg Korobeinichev, and Anatoly Chernov. Development and Validation of Skeletal Mechanism for Flame Inhibition by Trimethylphosphate (Institute of Chemical Kinetics and Combustion, Russia) • Charles Fleischmann. Defining the Heat Release Rate per Unit Area for Use in Fire Safety Engineering Analysis (University of Canterbury, NZ) The recipients of excellent student paper awards were: • Wei Gao, Naian Liu, Xieshang Yuan, Yueling Bai, Linhe Zhang, and Koyu Satoh. Neutral Plane and Length Scale of Spill Fire Plume Considering the Effects of Cross-Ventilation (SKLFS, USTC, China) • Toichiro Okawa, Wataru Ebina, Hiroyoshi Naito, and Akira Yoshida. Inhibition of Propane/Air Premixed Flame by Water Mist, (Tokyo Denki University) • Fietrysia Leonita, Harfan Sakti, and Yulianto Sulistyo Nugroho. Study of the Overall Movement Speed on Medium and High-Rise Buildings in Indonesia (Universitas Indonesia)

Excellent Poster Awards Excellent poster awards were decided in terms of the on-site performance of the poster presentations. Ballots were collected from all participants during the Poster Session. Final decisions were made by the ballots for each poster. The recipients were: • Supan Wang, Xinyan Huang, Haixiang Chen, Naian Liu, and Guillermo Rein. Expandable Polystyrene Foam Spot Fire Ignition by Hot Metal Particle (SKLFS, USTC, China) • Ken Mizutani, Kyosuke Miyamoto, Nozomu Hashimoto, and Osamu Fujita. Comparison of Limiting Oxygen Concentrations of Spreading Flame over Different Electric Wire Insulations in Microgravity (Hokkaido University, Japan) • Cheng-Chun Lin and Liangzhu (Leon) Wang. Real-Time Forecasting of Building Fires Using Data Assimilation (Concordia University, Canada) • Ping Ping, Qingsong Wang and Jinhua Sun. Study of the Fire Behavior of High-Energy Lithium-Ion Batteries with Full-Scale Burning Test (SKLFS, USTC, China)

Lifetime Contribution Awards To recognize their lifetime contribution to fire science and technology, the following three leaders were given awards: • Professor Wan-Ki Chow, Polytechnic University of Hong Kong • Professor Emeritus Takeyoshi Tanaka, Kyoto University • Professor Yuji Hasemi, Waseda University Finally, the editors acknowledge the support of many organizations and committees to accomplish the success of 10th AOSFST.

Preface

ix

Advisory Board Emeritus Prof. Takeyoshi Tanaka, Kyoto University, Japan (Chair) Mr. Greg Baker, SP Technical Research Institute of Sweden, Sweden Prof. Shen-Wen Chien, National Central Police University, Taiwan Prof. Wan-Ki. Chow, Hong Kong Polytechnic University, Hong Kong Dr. Dhionis Dhima, Centre Scientifique et Technique du Baˆtiment, France Prof. Bogdan Dlugogorski, Murdoch University, Australia Prof. Charles Fleischmann, University of Canterbury, New Zealand Prof. Ed Galea, University of Greenwich, UK Prof. George Hadjisophocleous, Carleton University, Canada Dr. Anthony Hamins, National Institute of Standards and Technology, USA Prof. Yuji Hasemi, Waseda University, Japan Prof. Yaping He, University of Western Sydney, Australia Prof. Patrick Van Hees, Lund University, Sweden Prof. Juncheng Jiang, Nanjing University of Technology, China Prof. Naian Liu, University of Science and Technology of China, China Prof. Yulianto Sulistyo Nugroho, University of Indonesia, Indonesia Prof. Hideo Ohtani, Yokohama National University, Japan Prof. Ai Sekizawa, Tokyo University of Science, Japan Prof. Arnaud Trouve, University of Maryland, USA Prof. Kuang-Chung Tsai, National Kaohsiung First University of Science and Technology, Taiwan Prof. Zhisheng Xu, Central South University, China Awards Committee Prof. Naian Liu, University of Science and Technology of China, China (Chair) Prof. Wan-Ki. Chow, Hong Kong Polytechnic University, Hong Kong Prof. Bogdan Dlugogorski, Murdoch University, Australia Prof. Charles Fleischmann, University of Canterbury, New Zealand Prof. Ed Galea, University of Greenwich, UK Prof. George Hadjisophocleous, Carleton University, Canada Prof. Yuji Hasemi, Waseda University, Japan Prof. Yulianto Sulistyo Nugroho, University of Indonesia, Indonesia Prof. Arnaud Trouve, University of Maryland, USA Prof. Kuang-Chung Tsai, National Kaohsiung First University of Science and Technology, Taiwan Local Organizing Committee Dr. Ichiro Hagiwara, Building Research Institute, Japan (Chair) Prof. Ritsu Dobashi, University of Tokyo, Japan Prof. Kazunori Harada, Kyoto University, Japan Mr. Kouichi Hasegawa, Nohmi Bosai Ltd., Japan Prof. Yuji Hasemi, Waseda University, Japan Prof. Takeo Hirashima, Chiba University, Japan Prof. Ken Matsuyama, Tokyo University of Science, Japan Dr. Tensei Mizukami, National Institute for Land and Infrastructure Management, Japan Dr. Tomohiro Naruse, Building Research Institute, Japan Mr. Masahisa Nashimoto, Japan Association for Fire Science and Engineering, Japan Prof. Yoshifumi Ohmiya, Tokyo University of Science, Japan Prof. Hideo Ohtani, Yokohama National University, Japan Dr. Masahiko Shinohara, National Research Institute of Fire and Disaster, Japan Emeritus Prof. Takeyoshi Tanaka, Kyoto University, Japan

x

Prof. Kaoru Wakatsuki, Shinshu University, Japan Dr. Tokiyoshi Yamada, National Research Institute of Fire and Disaster, Japan Logistics Committee Dr. Tomohiro Naruse, Building Research Institute, Japan (Chair) Dr. Yoshihiko Hayashi, National Institute for Land and Infrastructure Management, Japan Dr. Keisuke Himoto, National Institute for Land and Infrastructure Management, Tsukuba, Ibaraki, Japan Mr. Tatsuya Iwami, National Institute for Land and Infrastructure Management, Japan Dr. Koji Kagiya, Building Research Institute, Japan Dr. Tensei Mizukami, National Institute for Land and Infrastructure Management, Japan Dr. Daisaku Nii, Building Research Institute, Japan Dr. Tomoaki Nishino, Building Research Institute, Japan Dr. Junichi Suzuki, National Institute for Land and Infrastructure Management, Japan Dr. Hideki Yoshioka, National Institute for Land and Infrastructure Management, Japan Also, during the review and selection process, immeasurable time was spent by track chairs and reviewers. The editors would like to thank all the chairs and reviewers for their time and expert service. English mentors were allocated to some of the papers to help non-native English authors. The Chairman of the International Association for Fire Safety Science, IAFSS, supported the process of English mentoring by nominating volunteers. Special thanks go to the following: Program Committee Local Program Committee Members Prof. Kazunori Harada, Kyoto University, Japan (Chair) Dr. Keisuke Himoto, National Institute for Land and Infrastructure Management, Tsukuba, Ibaraki, Japan Prof. Ken Matsuyama, Tokyo University of Science, Japan Prof. Yuji Nakamura, Toyohashi University of Technology, Japan Prof. Kaoru Wakatsuki, Shinshu University, Japan

Track Chairs Track 1 – Fire/Explosion Physics and Chemistry Prof. Bogdan Dlugogorski, Murdoch University, Australia Prof. Masataro Suzuki, Nagaoka Institute of Technology, Japan Track 2 – Fire and Smoke Modeling and Experiments Prof. Yasushi Oka, Yokohama National University, Japan Track 3 – Human Behavior in Fire Prof. Tomonori Sano, Waseda University, Japan] Track 4 – Fire Statistics and Risk Assessment Prof. Richard Yuen, City University of Hong Kong, Hong Kong Prof. Akihiko Hokugo, Kobe University, Japan Track 5 – Fire Safety Design and Codes Prof. Charles Fleischmann, University of Canterbury, New Zealand Dr. Daisaku Nii, Building Research Institute, Japan Track 6 – Structural Behavior in Fire Prof. Linhai Han, Tsinghua University, China Prof. Takeo Hirashima, Chiba University, Japan Track 7 – Fire Properties and Testing Methods of Materials Prof. Kuang-Chung Tsai, National Kaohsiung First University of Science and Technology, Taiwan

Preface

Preface

xi

Dr. Hideki Yoshioka, National Institute for Land and Infrastructure Management, Japan Track 8 – Industrial Fires Prof. Sung-Chan Kim, Kyungil University, South Korea Dr. Masakatsu Honma, National Research Institute of Police Science, Japan Track 9 – Urban, Wildland/Urban Interface and Forest Fires Prof. Naian Liu, University of Science and Technology of China, China Dr. Masahiko Shinohara, National Research Institute of Fire and Disaster, Japan Track 10 – Fire Detection and Suppression Prof. Cheol-Hong Hwang, Daejeon University, South Korea Prof. Hiroyuki Torikai, Hirosaki University, Japan Track 11 – Fire Investigation/Fire Services Prof. Fa-lin Chen, National Taiwan University, Taiwan Dr. Yoshio Ogawa, National Research Institute of Fire and Disaster, Japan Track 12 – Fire Protection of Cultural Heritages Prof. Yuji Hasemi, Waseda University, Japan Track 13 – Fire Protection of High-Rise Buildings Prof. Jose Torero, University of Queensland, Australia Dr. Keich Suzuki, Shimizu Corporation, Japan Track 14 – Fires of Recyclable Fuels and Renewable Resources Dr. Qingsong Wang, University of Science and Technology of China, China Prof. Takashi Tsuruda, Akita Prefecture University, Japan Publication Committee Prof. Ken Matsuyama, Tokyo University of Science, Japan (Chair) Dr. Yusaku Iwata, National Research Institute of Fire and Disaster, Japan Prof. Kazunori Kuwana, Yamagata University, Japan Dr. Tomoaki Nishino, Building Research Institute, Japan

Reviewers Nobuyuki Abe Mohammednoor Altarawneh Shigeaki Baba Abu Bakar Truchot Benjamin Hubert Biteau Serge Bourbigot Ricky Carvel Haixiang Chen Shen-Wen Chien Francesco Colella Siaka Dembele Michelle Donnelly Mario Fontana Michael Gollner Ichiro Hagiwara Yuji Hasemi Masayuki Hirota Xinyan Huang Yuka Ikehata Jie Ji Bjorn Karlsson

Anthony Abu Petra Andersson Vyto Babrauskas Greg Baker Craig Beyler Per Blomqvist Karen Boyce Franc¸ois Chatelon Xianfeng Chen Wan-Ki Chow Peter Collier Bogdan Dlugogorski Vince Dowling Kevin Frank Eric Guillaume Tuula Hakkarainen Yaping He San-Ping Ho Yu-Hsiang Huang Tomohiko Imamura Nils Johansson Sung-Chan Kim

Yuki Akizuki Vivek Apte Ariza Sharikin Dave Barber Subrata Bhattacharjee Lars Bostro¨m Andy Buchanan Prateep Chatterjee Sherman Cheung Ed Claridge Michael A. Delichatsios Ritsu Dobashi Rita Fahy Xiangpeng Gao Steven Gwynne Kazunori Harada Patrick van Hees Longhua Hu Zhaohui Huang Ken Isman Koji Kagiya Michael Klippel

xii

Yuji Kudo Mariano Lazaro Jiao Lei Ying Zhen Li Weiming Liu Jian Ma Ken Matsuyama Tensei Mizukami Khalid Moinuddin Jun Nakamura Vedha Nayagam Daniel Nilsson Hideo Ohtani Sandra Olson Ezgi Oztekin Hugues Pretrel Rene Rossi Jaime Santos-Reyes Yusuke Shintani Michael Spearpoint Junichi Suzuki Fumiaki Takahashi Futoshi Tanaka Ian Thomas Arnaud Trouve Kaoru Wakatsuki Kelvin Wong Sergey Yakush Setsuko Yoshino

Preface

Kazunori Kuwana Eric Lee Kaiyuan Li Tianshui Liang Xuan Liu Jianqiang Mai Bart Merci Masayuki Mizuno Yoshinori Murakami Yuji Nakamura Xiaomin Ni Xuemin Niu Yasushi Oka Rosaria Ono Beom Jin Park Enrico Ronchi Kozo Saito Ai Sekizawa Albert Simeoni Osami Sugawa Keichi Suzuki Ken-Ichi Takanashi Zhong Tao Hiroyuki Torikai Takahiro Tsukame Yi Wang Huahua Xiao Yukio Yamauchi Hideki Yoshioka

Angus Law Yee-Ping Lee Wei Li Naian Liu Siu Ming Lo Mariusz Maslak Yoshikazu Minegishi Huai-Chu Mo Shinji Nakahama Tomohiro Naruse Daisaku Nii Hiroaki Notake Stephen Olenick Fuminobu Ozaki Seul-Hyun Park Jean-Louis Rossi Iain Sanderson Masahiko Shinohara Tian-Yi Song Patrick Summers Masataro Suzuki Franco Tamanini Geoff Thomas Svetlana Tretsiakova-Mcnally Takashi Tsuruda Yong Wang Yibing Xin Guan Heng Yeoh Byeonghun Yu

English Mentors Francine Amon Ricky Carvel Xinyan Huang Jose Torero Cullen

Edmund Ang Rita Fahy Jianqiang Mai

Vivek Apte Virginia Alonso Gutierrez Samuel L. Manzello

Sincerely yours, Editors Fire Science and Technology 2015 - The Proceedings of 10th Asia-Oceania Symposium on Fire Science and Technology

Contents

Part I 1

In Memorial

Memories of Professor Toshisuke Hirano . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ritsu Dobashi, Takeshi Suzuki, Masataro Suzuki, Lijing Gao, Kazunori Kuwana, Wan-Ki Chow, Bogdan Dlugogorski, Patrick van Hees, and Jinhua Sun

Part II

3

Plenary and Invited

2

Fire Resistance of Multistorey Timber Buildings . . . . . . . . . . . . . . . . . . . . . . Andrew H. Buchanan

3

Ultimate Strength and Its Application to Post-Earthquake Fire Resistance of Steel Frames in Fire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hiroyuki Suzuki

17

Integrating Wildland and Urban Fire Risks in Local Development Strategies in Indonesia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yulianto Sulistyo Nugroho

31

Thermal Analysis and Flame Spread Behavior of Building-Used Thermal Insulation Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jinhua Sun and Lin Jiang

45

A Discussion on Tall Building Fire Safety in the Asia-Oceania Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wan-Ki Chow

61

Toward a Risk-Informed Performance-Based Approach for Post-Earthquake Fire Protection Design of Buildings . . . . . . . . . . . . . . . . Brian J. Meacham

73

4

5

6

7

Part III 8

9

10

9

Compartment Fire

Interaction of a Pool Fire in a Compartment with Negative Pressure Generated by Mechanical Ventilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yasuo Hattori, Ken Matsuyama, Hitoshi Suto, Seiji Okinaga, and Eiji Onuma

89

Pool Fire Behavior in a Small and Mechanically Ventilated Compartment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alexis Coppalle, Alvin Loo, and Philippe Aine´

97

Discussion on Heat Lost Through Solid Boundaries in Modelling Atrium Fires Under Mechanical Exhaust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 Liang Yi, Danyang Sun, Yuanzhou Li, Ran Huo, Wan-Ki Chow, and Nai-Kong Fong

xiii

xiv

11

Contents

Experimental Study on Fire Behavior in a Compartment Under Mechanical Ventilated Conditions: The Effects of Air Inlet Position . . . . . . . . . . . . . . . . . 111 Ken Matsuyama, Seiji Okinaga, Yasuo Hattori, and Hitoshi Suto

Part IV

Egress Safety

12

Study of the Occupant Characteristics During Evacuation in Mediumand High-Rise Buildings in Indonesia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 Fietrysia Leonita, Harfan Sakti, and Yulianto Sulistyo Nugroho

13

Effect of Weibull Distributed Pre-movement Time on Evacuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Yiping Zeng, Weiguo Song, Feizhou Huo, and Xiaoge Wei

14

Travel Speed and Visibility in a Pedestrian Space with Inhomogeneous Distribution of Floor Illuminance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 Yuki Akizuki, Norikazu Yamaguchi, Ai Konno, and Shino Okuda

15

Analysis of Commuter Line’s Passenger Movement in the Station During Peak Hours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 Harfan Sakti, Fietrysia Leonita, Muhammad Agung Santoso, and Yulianto Sulistyo Nugroho

Part V

Electrical and Industrial Fires

16

The Effect of Multicomponent Electrolyte Additive on LiFePO4-Based Lithium Ion Batteries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 Lihua Feng, Qingsong Wang, Chengying Ai, and Jinhua Sun

17

PVC Cable Fire Toxicity Using the Cone Calorimeter . . . . . . . . . . . . . . . . . . 175 Wadie A. Al-Sayegh, Omar Aljumaiah, Gordon Edward Andrews, and Herodotos N. Phylaktou

18

Lessons Learned from the 0801 Petrochemical Pipeline Explosions in Kaohsiung City . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 Huei-Ru Hsieh, Chung-Sheng Lee, Min-Cheng Teng, Wen-Ray Su, and Wei-Sen Li

Part VI

Fac¸ade Fires

19

Experimental Study of the Fire Behaviour of Wooden Facades . . . . . . . . . . . 193 Dhionis Dhima and Jean-Marie Gaillard

20

Verification Methodology of Vertical Fire Spread to the Upstairs Room via Openings and Facade Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Hideki Yoshioka, Ko Muraoka, Masatoshi Nakamura, Yoshikazu Deguchi, Takeshi Morita, Kouta Nishimura, Masaki Noaki, Yoshifumi Ohmiya, and Tomohiro Naruse

21

Experimental Study on Confluence of the Emerged Flames Ejected from Two Adjacent Openings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 Yoshikazu Deguchi, Hideki Yoshioka, and Yoshifumi Ohmiya

22

Neutral Plane and Length Scale of Spill Fire Plume Considering the Effects of Cross-Ventilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 Wei Gao, Naian Liu, Xieshang Yuan, Yueling Bai, Linhe Zhang, and Koyu Satoh

Contents

xv

Part VII

Fire Protection of High-Rise Buildings

23

Performance-Based Fire Safety Design for a Skyscraper: A Case in Japan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 Naohiro Takeichi and Yoshikazu Minegishi

24

Study on Fire Plume Behavior in Vertical Shafts of Buildings . . . . . . . . . . . . 245 Yi-chul Shin, Yoshifumi Ohmiya, Shin-ichi Tsuburaya, Yuki Yoshida, Kazumasa Tashima, and Jun-ichi Suzuki

25

Experimental Investigation on Glass Cracking for Wind Load Combined with Radiant Heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255 Han Zhao, Qingsong Wang, Yanfei Su, Yu Wang, Guangzheng Shao, Haodong Chen, and Jinhua Sun

Part VIII

Fire Resistance

26

Comparison of Existing Time-Equivalence Methods and the Minimum Load Capacity Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263 Philip Xie, Anthony Abu, and Michael Spearpoint

27

Components and Consequences of Cross-Laminated Timber Delamination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273 Richard Emberley, Arne Inghelbrecht, Nicholas Doyle, and Jose´ L. Torero

28

Fire Behavior of Cross-Laminated Timber (CLT) Slabs: Two-Way Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281 Nicholas Doyle, Richard Emberley, and Jose´ L. Torero

29

Influence of Water Content on Total Strain of Super High-Strength Concrete Under Elevated Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 Heisuke Yamashita, Toru Yoshida, and Takeo Hirashima

30

A Study on Structural Behavior of Reinforced Concrete Walls Exposed to Hydrocarbon Fire Under Vertical Load . . . . . . . . . . . . . . . . . . . . . . . . . . . 299 Takeshi Morita, Heisuke Yamashita, Masuhiro Beppu, and Makoto Suzuki

31

Response of Steel Members Subject to Temperature Gradient in Localized Fires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309 Chao Zhang and Guo-Qiang Li

32

Tensile Resistance of Splice Connections with High-Strength Bolts in Fire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319 Shuhei Ando, Robert Dwiputra, and Takeo Hirashima

Part IX

Fire Safety Design

33

Statistical Analysis on the Reliability of Sprinkler Systems: Study on a Risk-Based Evacuation Safety Design Method . . . . . . . . . . . . . . . . . . . . 331 Yuka Ikehata, Jun-ichi Yamaguchi, Yoshikazu Deguchi, and Takeyoshi Tanaka

34

Determination of Design Fire Load for Structural Fire Safety in the Compartment Subdivided by Non-Fire-Rated Partitions . . . . . . . . . . . 341 Tensei Mizukami and Takeyoshi Tanaka

35

Linking Safety Factor and Probability of Failure Based on Monte Carlo Simulation in Fire Safety Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 Depeng Kong, Shouxiang Lu, and Ping Ping

xvi

Contents

36

Domestic Sprinkler: It Is Time to Consider Mandatory Requirement in Hong Kong . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361 Yuen Kwong Woo, Tai Keung Tam, Wan-Ki Chow, and Nai-Kong Fong

37

Discuss Performance-Based Fire Safety Design in Taiwan: Case Study of “Wei-Wu-Ying Center for the Arts” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367 Hung-Chieh Chung, Shen-Wen Chien, and Tzu-Sheng Shen

38

To Improve the Care Environments by Using Fire Safety Engineering for Existing Small-Scale Hospitals in Taiwan . . . . . . . . . . . . . . . . . . . . . . . . . 373 Wei-Wen Tseng, Tzu-Sheng Shen, Shu-Feng Liao, and Chih-Chi Tseng

39

Risk Communication Applied to Community-Based Fire Mitigation and Management for Historic Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383 Pei-Chun Shao

Part X

Fire Source Model

40

Experimental Study on the Effect of Frame Height of Bed Mattress upon Fire Behavior in Compartment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397 Kye-Won Park, Jong-Jin Jeong, Masayuki Mizuno, Yoshifumi Ohmiya, Kenichi Ikeda, and Yoshihiko Hayashi

41

Analysis of Combustion Expansion and Heat Release Rate During Combustion of Mattress Installed at Different Heights . . . . . . . . . . . . . . . . . . 409 Jong-Jin Jeong, Kye-Won Park, Masayuki Mizuno, Yoshifumi Ohmiya, and Kenichi Ikeda

42

Defining the Heat Release Rate per Unit Area for Use in Fire Safety Engineering Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419 Charles Fleischmann

43

Algebraic Equations for Calculating Surface Flame Spread and Burning of a Cubical-Shaped Polyurethane Foam Block . . . . . . . . . . . . . . . . . . . . . . . 427 Kazuhiko Ido, Kazunori Harada, Yoshifumi Ohmiya, Ken Matsuyama, Masaki Noaki, and Junghoon Ji

44

An Experimental Study on the Mass Flow Rate from a Line Fire Source Along a Vertical Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437 Yusuke Shintani, Tsutomu Nagaoka, Yoshikazu Deguchi, and Kazunori Harada

45

The Simulation of Release Consequence with a Modified Gaussian Plume Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445 Hongya Zhu, Xuanya Liu, Qingsong Wang, and Jinhua Sun

46

A Simple Hand Calculation Method to Estimate the Pyrolysis Kinetics of Plastic and Wood Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455 Xiaoyun Wang, Charles Fleischmann, Michael Spearpoint, and Kaiyuan Li

Part XI

Firefighting

47

Research on Firefighting Simulation Training System of Large Oil Tank Zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465 Zhi-hui Li, Li-biao Pan, and Qiang Li

48

Research on Indoor Firefighter Positioning Based on Inertial Navigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473 Di-Ping Yuan, Zhi-hui Li, and Mu Li

Contents

xvii

49

Comparison of Bench-Scale and Manikin Tests of Protective Clothing Systems During Low-Level Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485 Ming Fu, Wenguo Weng, Jie Yang, and Qian Zhang

50

Use of Water Curtains to Prevent Ignition by Firebrands . . . . . . . . . . . . . . . 491 Vasily Novozhilov

Part XII

Fires in Transportation Systems

51

Full-Scale Experiments on Ship Accommodation Cabin Fire . . . . . . . . . . . . . 499 Bosi Zhang, Jiaqing Zhang, Xiaomin Wang, Shouxiang Lu, and Changhai Li

52

Underground Station Fire Simulation with Multilayer Zone Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 511 Keichi Suzuki and Takeyoshi Tanaka

53

Performance-Based Approach for Fire Safety in Aboveground Mass Rapid Transit Stations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521 S. Leong Poon, Adrian Cheong Wah Onn, and Boon Tong Tan

Part XIII

Fires of Recyclable Fuels and Renewable Resources

54

Toxic Gas Emissions from a Timber-Pallet-Stack Fire in a Full-Scale Compartment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533 Abdulaziz Alarifi, Herodotos N. Phylaktou, Gordon Edward Andrews, Jim Dave, and Omar Aljumaiah

55

An Experimental Study on the Spread of Burning Gasoline Spilling on Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543 Yuntao Li, Hong Huang, Zheng Wang, Jianzhong Zhang, and Chunming Jiang

56

An Experimental Study of Complex Fuel Burning Behavior Using Characteristic Fuel Unit Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 549 Yibing Xin, Yi Wang, Marcos Chaos, and Sergey Dorofeev

Part XIV

Flame Characteristics

57

An Analytical Method for the Estimation of Radiation Heat Flux from Open Pool Fires and Pool Fires Impinging on Ceilings . . . . . . . . . . . . . 559 Alexandros Vouros, Michael A. Delichatsios, and Jianping Zhang

58

PIV Measurements of Velocity Fields of Three Hot Air Jet Plumes Impinging on a Horizontal Ceiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 569 Xiangyang Zhou

59

Evaluation of Temperature Rise Under an Eave Due to Flame Impingement: Toward the Mitigations of Fire Spread Risk in Japanese Historic Urban Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577 Keisuke Himoto and Yoshikazu Deguchi

60

A Procedure for Measuring Flame Spread Properties of Materials by Cone Calorimeter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587 Tatsuya Tarumoto, Junghoon Ji, Tsuneto Tsuchihashi, Kazunori Harada, Woon-Hyung Kim, Kye-Won Park, and Jong-Hoon Kim

Part XV 61

Flame Retardant

Correlations Between Measurements of Flame-Retarded High-Density Polyethylene Composites Subjected to Three Conventional Fire Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 599 Zhi-Sheng Xu, Long Yan, Ding-Li Liu, Tian-Xiao Ni, Jin-Zhi Peng, and Ye Xu

xviii

Contents

62

Fire Behavior of Intumescent Polyurethane: Synergy, Morphology, and Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 609 Serge Bourbigot, Maryska Muller, and Sophie Duquesne

63

Development and Validation of Skeletal Mechanism for Flame Inhibition by Trimethylphosphate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 619 Tatyana Bolshova, Vladimir Shvartsberg, Andrey Shmakov, Oleg Korobeinichev, Sergey Yakush, and Anatoly Chernov

Part XVI

Flashover

64

A Validation Study of Existing Formulas for Determining the Critical Heat Release Rate for Flashover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631 SungChan Lee and Kazunori Harada

65

Experimental Study of Time to Onset of Flashover in Classroom Size Compartment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639 Tomohiro Naruse, Koji Kagiya, Jun-ichi Suzuki, Noboru Yasui, and Yuji Hasemi

66

Experimental Study and Model Analysis of Flashover in Confined Compartments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649 Ruowen Zong, Ruxue Kang, Weifeng Zhao, and Changfa Tao

Part XVII

Material Flammability

67

Experimental Study on Radiation Blockage of Small-Scale Vertical PMMA Fires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 661 Zhen Li, Naian Liu, Shaojie Zhang, Xiaodong Xie, and Wei Gao

68

Thermal Performance of Magnesium Oxide Wall Board Using Numerical Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 667 Mohamed Rusthi, Poologanathan Keerthan, Mahen Mahendran, and Anthony Deloge Ariyanayagam

69

Burning Behaviors of Thermoplastic Pellets in Pans Under Different Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 677 Qiyuan Xie, Tangqing Wu, Nan Wang, Ran Tu, and Xi Jiang

70

Investigation of the Contribution to Fire of Electrical Cable by a Revisited Mass Loss Cone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 687 Gae¨lle Fontaine, Franck-Estime Ngohang, Laurent Gay, and Serge Bourbigot

71

Experimental Study on Thermal Insulation Properties of Charred Woods Under Radiative Heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695 Masaki Noaki, Jun-ichi Suzuki, Yoshifumi Ohmiya, and Michael A. Delichatsios

72

Thermal Boundaries of Intumescent-Type Insulations in Cone Calorimeter Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 705 Sungwook Kang, Sengkwan Choi, and Joungyoon Choi

73

The Effect of Specimen Thickness on Critical Heat Flux and Effective Thermal Inertia Calculations Using Cone Calorimeter and Ignitability Test Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715 Tsuneto Tsuchihashi and Kazunori Harada

Part XVIII 74

Measurement Methods

Low-Temperature CO Catalytic Oxidation over KOH-Hopcalite Mixtures and In Situ CO2 Capture from Fire Smoke . . . . . . . . . . . . . . . . . . . . . . . . . . 725 Yafei Guo, Chuanwen Zhao, Changhai Li, and Shouxiang Lu

Contents

xix

75

Comparison and Assessment of Particle Mass Concentration Measurements in Fire Smokes with a Microbalance, Opacimeter, and PPS Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735 Axel Bellivier, Alexis Coppalle, Alvin Loo, Je´roˆme Yon, Louis Decoster, Sylvie Dupont, and Herve´ Bazin

76

Improvements to the Hartmann Dust Explosion Equipment for MEC Measurements that Are Compatible with Gas Lean Limit Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743 Muhammad Saeed, Gordon Edward Andrews, Herodotos N. Phylaktou, Hamed Sattar, Clara Huescar-Medina, David Slatter, Pradeep Herath, and Bernard Gibbs

77

Numerical Simulations of Gas-Phase Interactions of Phosphorus-Containing Compounds with Cup-Burner Flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 751 Fumiaki Takahashi, Viswanath Katta, Gregory Linteris, and Valeri Babushok

Part XIX

Outdoor Fires

78

CFD Study of Large City Fires in Windy Conditions Relevant to Aerial Firefighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 761 Koyu Satoh, Naian Liu, Xiaodong Xie, and Wei Gao

79

The Velocity and Structure of the Flame Front at Spread of Fire Across the Pine Needle Bed Depending on the Wind Velocity . . . . . . . . . . . . . . . . . . 771 Oleg Korobeinichev, Alexander Tereshchenko, Alexander Paletsky, Andrey Shmakov, Munko Gonchikzhapov, Anatoly Chernov, Lilia Kataeva, Dmitriy Maslennikov, and Naian Liu

Part XX

Smoke Control and Ventilation

80

Smoke Filling in a Confined Compartment with Single Ceiling Vent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 783 Qiang Li, Jinmei Li, Shijing Ren, and Jiaqing Zhang

81

Full-Scale Tests and CFD Modeling to Investigate the Effect of Opening Arrangement on Smoke Layer Height in Atrium Fires . . . . . . . . . . . . . . . . . 793 Amir Rafinazari and George Hadjisophocleous

82

Experimental Study on Influence of Air Supply System Difference on Smoke Shielding Performance in Air Pressure Smoke Control . . . . . . . . . 801 Masashi Kishiue, Jun-ichi Yamaguchi, Seiji Okinaga, Ken Matsuyama, and Takayuki Matsushita

83

An Experimental Investigation of the Fire-Induced Oscillatory Vent Flow Through a Horizontal Opening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 811 Xiao Chen, Shouxiang Lu, Xiaomin Wang, Kim Moew Liew, Shengshi Huang, and Changhai Li

Part XXI

Solid Combustion Fundamentals

84

Experimental Study of Sidewall and Pressure Effect on Vertical Downward Flame Spread Over Insulation Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 823 Weigang Yan, Yang Shen, Lin Jiang, Weiguang An, Yang Zhou, Zhen Li, and Jinhua Sun

85

Heat Loss and Kinetic Effects on Extinction and Critical Self-Sustained Propagation of Forced Forward Smoldering . . . . . . . . . . . . . . . . . . . . . . . . . 831 Jiuling Yang, Haixiang Chen, and Naian Liu

xx

86

Contents

Numerical Study of the Radiative and Turbulent Heat Flux Behavior of Upward Flame Spread Over PMMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 841 Alexander Karpov, Artem Shaklein, Mikhail Korepanov, and Artem Galat

Part XXII

Suppression

87

Validating the Function of Absorber Plates for Auto-sprinkler System Activation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 851 Kuang-Chung Tsai, Yukio Yamauchi, and Ken Matsuyama

88

An Experimental Study on the Smoke-Logging Phenomenon Using Sprinkler for Performance-Based Evacuation Safety Design . . . . . . . . . . . . . 859 Dong-Goo Seo, Ung-Gi Yoon, In-Hyuk Koo, Bong-Chan Kim, Dong-Eun Kim, Ken Matsuyama, and Young-Jin Kwon

89

Effect of Fire Detection Function on Fire Suppression in Home Stay Facilities in Taiwan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 869 Chung-Hwei Su, Kuang-Chung Tsai, Ming-Hui Dai, and Chun-Chou Lin

90

Full-Scale Experimental Study on Fire Suppression Performance of a Water Mist System for Large Shipboard Machinery Spaces . . . . . . . . . . 877 Xiaowei Wu and Shouxiang Lu

91

Examination of Extinguishment Method with Extinguishing Powder Packed in a Spherical Ice Capsule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 887 Miho Ishidoya, Hiroyuki Torikai, Akihiko Ito, and Yuji Shiibashi

92

Experimental Study on Transformer Oil Pool Fire Suppression by Water Mist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 895 Pei Zhu, Xishi Wang, Zhigang Wang, Haiyong Cong, and Xiaomin Ni

93

Inhibition of Propane/Air Premixed Flame by Water Mist . . . . . . . . . . . . . . . 903 Toichiro Okawa, Wataru Ebina, Hiroyoshi Naito, and Akira Yoshida

Part XXIII

Tunnel Fires

94

The Influence of Train on the Smoke Movement in Subway Tunnel with Longitudinal Ventilations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 915 Shaogang Zhang, Xudong Cheng, Ruifang Zhang, Kaiyuan Li, Song Lu, Hui Yang, and Heping Zhang

95

The Effect of Aspect Ratios on Critical Velocity in Tunnel Fires . . . . . . . . . . 925 Changcheng Liu, Song Lu, Ruifang Zhang, Hui Yang, Xudong Cheng, and Heping Zhang

96

Influence of Different Longitudinal Wind on Natural Ventilation Efficiency with Vertical Shaft Under Different Fires in Tunnel . . . . . . . . . . . . . . . . . . . 933 Haiyong Cong, Xishi Wang, Pei Zhu, Zhigang Wang, Tonghui Jiang, and Qiong Tan

Part I In Memorial

1

Memories of Professor Toshisuke Hirano Ritsu Dobashi, Takeshi Suzuki, Masataro Suzuki, Lijing Gao, Kazunori Kuwana, Wan-Ki Chow, Bogdan Dlugogorski, Patrick van Hees, and Jinhua Sun

Abstract

It is with great sadness that we have learned of the death on February 13, 2014 of Professor Toshisuke Hirano, who was a “father of AOAFST”. At the occasion of the 10th AsiaOceania Symposium on Fire Science and Technology (10th AOSFST), we share the memories of Prof. Hirano and renew his will on the progress of fire science. In this article, his achievements are summarized, such as the activities in academic societies, achievements on scientific researches, major awards received, and so on. The eulogies given by Prof. Wan-Ki Chow (Hong Kong), Prof. Bogdan Dlugogorski (Australia) and Prof. Patrick van Hees (Sweden), and Prof. Jinhua Sun (China) are included in this article. Keywords

Professor Toshisuke Hirano  Father of AOAFST  Achievements

R. Dobashi (*) The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan e-mail: [email protected] T. Suzuki National Research Institute of Fire and Disaster, 4-35-3 Jindaijihigashi-machi, Chofu, Tokyo 182-8508, Japan M. Suzuki Nagaoka University of Technology, 1603-1 Kamitomioka, Nagaoka, Niigata 940-2188, Japan L. Gao Chiba Institute of Science, 3 Shiomi-cho, Choshi, Chiba 288-0025, Japan K. Kuwana Yamagata University, 4-3-16 Jonan, Yonezawa, Yamagata 992-8510, Japan W.-K. Chow The Hong Kong Polytechnic University, Kowloon, Hong Kong B. Dlugogorski Murdoch University, Perth, Australia P. van Hees Lund University, Lund, Sweden J. Sun University of Science and Technology of China, Hefei, China

Professor Toshisuke Hirano (1939–2014)

It is with great sadness that we have learned of the death on February 13, 2014 of Professor Toshisuke Hirano, who was a “father of AOAFST”. Prof. Hirano, aged 74 and Emeritus Professor at the University of Tokyo, passed away after 2 months medical treatment at Tokyo Metropolitan Neurological Hospital in Tokyo, Japan. At the occasion of the 10th Asia-Oceania Symposium on Fire Science and Technology (10th AOSFST) held in Japan, we would like to

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_1

3

4

share the memories of Prof. Hirano and renew his will on the progress of fire science. Prof. Hirano made enormous contributions to the globalization of fire science. He had rendered great services to the International Association of Fire Safety Science (IAFSS) continuously from the time of its founding. He served as a secretary (1985–1991), vice chairman (1991–1997), and chairman (1997–2002) of IAFSS. He had dedicated his efforts especially to developing the Association into a more globalized community for the fire science. He worked as a bridge across areas of Asia, Oceania, and Russia, and broadened international activities of IAFSS in these areas. He established the Asia-Oceania Association for Fire Science and Technology (AOAFST) as a branch of IAFSS in 1992 to promote international collaboration of fire safety science within, and beyond, the Asia-Oceania section. AOAFST organizes Asia-Oceania Symposiums for Fire Science and Technology (AOSFSTs). The first AOSFST was held in 1992 in Hefei, China. The symposium was followed by the 2nd in 1995 (Russia), the 3rd in 1998 (Singapore), the 4th in 2000 (Japan), the 5th in 2001 (Australia), the 6th in 2004 (Korea), the 7th in 2007 (Hong Kong), the 8th in 2010 (Australia), the 9th in 2012 (China), and the 10th in 2015 (Japan). At the time when AOAFST was established, the acceptance rate of papers submitted from Asian countries was lower than those from Europe or the United States. Prof. Hirano supposed two reasons: one was the difficulty in writing in English; the other was research topics closely related to Asian-specific problems, with which the European and U.S. people were not familiar. With this in view, he established AOAFST to improve mutual understanding and to enlarge opportunities of exchanging ideas in English. At the present day, papers from Asian countries have a considerable presence among whole papers from the world. AOSFST also becomes a very important regular symposium, which many researchers attend from not only Asia-Oceanian countries but also countries outside that region. Prof. Hirano performed pioneering researches in fire safety science and engineering. For example, he studied flame spread phenomena above combustible solid and liquid. He carefully examined the airflow near the flame and succeeded in describing the exact flame structure as a diffusion flame in a boundary layer. He studied also flame propagation phenomena in combustible particle clouds and flammable gas mixtures (dust explosion and gas explosion, respectively). Propagation mechanisms were analyzed on the basis of aero-thermo dynamics. These fundamental studies revealed how these phenomena of flame spread and

R. Dobashi et al.

propagation are governed by processes including mass and heat transfers, combustion reaction, and flow behavior. His results have contributed immensely to scientific understandings of flame spread and flame propagation that occur in fire and/or explosion incidents. Those results were presented in more than 170 original papers, 65 review papers, and 27 books. Toward these achievements, he was granted the Award for the Outstanding Paper (Japan Society for Safety Engineering, 1981), the Komo Prize (Tanikawa Fund Promotion of Thermal Technology, 1987), and other noted awards. His achievements in research are of great importance for the promotion of fire safety science and engineering, and also have had great impact on other academic areas such as combustion science, transport phenomena, and energy engineering.

Prof. Hirano was an active person and his leadership was exceptional. He served not only for AOAFST and IAFSS, but also for the Combustion Institute (Executive Committee member, Secretary for Foreign Affairs, Board of Directors), and Japan Association for Fire Science and Engineering (President). He has also served as a member of editorial board for the Journal of Loss Prevention in Process Industries, the Journal of Japan Society for Safety Engineering, and Combustion and Flame. Toward these great contributions, he was awarded Bernard Lewis Gold Medal (Combustion Institute, 2004), Award for Prominent Contribution (Japan Association for Fire Science and Engineering, 1985), Dionizy Smolenski Medal (Polish Academy of Sciences, 1997), Award for Prominent Contribution in Science (Japan Institute of Energy, 1999), International Science and Technological Cooperation Award of the People’s Republic of China (2003), AOSFST Award for Founding and Active Promotion (2007), and AOSFST Lifetime Contribution Award (2012). Prof. Hirano made considerable efforts to build young generations. He was known as a strict teacher; however, he

1

Memories of Professor Toshisuke Hirano

trained them with warm heart. Many of his pupils have played important roles in the research field of fire safety science. He was an attractive person who loved sports (association football) and flowers (especially roses). We believe his legacy will endure in many ways. Following this writing, eulogies are given by Prof. Wan-Ki Chow (Hong Kong), Prof. Bogdan Dlugogorski (Australia) and Prof. Patrick van Hees (Sweden), and Prof. Jinhua Sun (China).

1.1

Memorial Statement for Professor Toshisuke Hirano (1939–2014)

W. K. Chow President of AOAFST, The Hong Kong Polytechnic University, Kowloon, Hong Kong Professor Toshisuke Hirano passed away on 13 February 2014 at the age of 74. In addition to being a pioneer researcher and a well-known expert in the fields of combustion and fire research, he had made significant contributions to develop fire science and technology in the Asia-Oceania areas. His passing away is a great loss to the science community. Professor Hirano was one of the founders of the AsiaOceania Association for Fire Science and Technology (AOAFST) and being the President of the AOAFST from 1992 to 1995. In recognition of his outstanding contributions to the development of Fire Science and Technology, he was award the Lifetime Contribution Award of AOAFST in 2012. He had served on many international and local professional bodies such as the International Association for Fire Safety Science (IAFSS), the Combustion Institute, the Combustion Society of Japan and the Japan Association for Fire Science and Engineering; and many more government

5

committees overseeing fire safety matters. He chaired the IAFSS from 1997 to 2002. In the Asia-Oceania areas, he has pushed hard in promoting fire research. More importantly, the problems facing in the Asia-Oceania areas are very different from those in the developed areas. There are very tall buildings, deep subway stations in urban areas, long tunnels for vehicles, trains and subways, large halls and many others. The cities used to be densely populated, with those spaces crowded with people, with high fire load, and different safety culture from the western world. Therefore, the fire hazards are very different, and a unified fire profession in this part of the world is necessary. In the AOSFST held in Hong Kong in 2007, the fire hazards of karaoke music boxes were clearly pointed out by him. He had worked at the University of Tokyo for over 20 years since 1976, promoted to Professor in 1985, and retired in 1999. He then served as President of National Research Institute of Fire Disaster from 2001 to 2004, and also President of Chiba Institute of Science from 2004 to 2010. We are greatly saddened by the passing away of Professor Hirano. He will always remain in our memories.

1.2

Tribute to Professor Toshisuke Hirano at the 10th AOSFST in Tsukuba

Bogdan Dlugogorski Immediate Past Chairman of IAFSS, Murdoch University, Perth, Australia Patrick van Hees Chairman of IAFSS, Lund University, Lund, Sweden We were surrounded by sorrow and sadness by the news that Professor Hirano had passed away during the week of the 11th International Symposium of the International

6

R. Dobashi et al.

Association of Fire Safety Science in Christchurch. At that time, Professor Patrick van Hees, the new Chair of the IAFSS shared the sad news with the membership of our Association. It is difficult to think about the world around us and about our fire safety community without Professor Hirano being its member. Professor Hirano was a gentleman, a scholar, a colleague, a mentor and a pre-eminent leader in fire safety research. He was one of the founding fathers of the International Association for Fire Safety Science, and for 18 years served as the Association Secretary, Vice-Chairman and the Chairman. He was the local organiser of the 2nd Symposium and acted as the Program Chair for the 5th and 6th Symposia of the IAFSS. Through his work, he has defined who we are as the fire safety community. He mentored one of us, Professor Bogdan Dlugogorski, the past Chair of IAFSS, to take first steps as a fire safety researcher in Australia. His passing away is a great loss to everyone. As a scientist, he has made important contributions to the archival literature on fires and explosions, including seminal papers on flame spread. But his contributions are too many to mention. He has authored papers in almost every area of fire research. In the Asia-Oceania region, he was the one who brought together fire researchers from Australia, China, Japan and Russia to establish the Asia-Oceania Association for Fire Science and Technology. We all own him a great deal of gratitude for everything he has done for improving fire safety in our region. Professor Hirano will remain in our memories and in our thoughts.

1.3

In Memory of My Mentor: Professor Toshisuke Hirano

Jinhua Sun University of Science and Technology of China, Hefei, China Today, we are getting together in Tsukuba, attending the 10th Asia-Oceania Symposium on Fire Science and Technology, to exchange views on issues of fire safety and discuss the mysteries of fire science. However, it is such a pity that we will never see the founding father of the AsiaOceania Association for Fire Science and Technology, the past chair of the International Association for Fire Safety Science, my PhD supervisor and true mentor, Professor Toshisuke Hirano. As we all know, Professor Hirano was a great scientist in combustion and fire science and a pre-eminent leader in fire

safety research. In his whole life, he authored more than 20 books and more than 200 original research papers, which were published in the best journals on combustion and fire science, such as Combustion and Flame, Proceedings of the Combustion Institute, Fire Safety Journal. He founded and established the Asia-Oceania Association for Fire Science and Technology in early 1990s and served as the first Chair. He also served as a chair of the International Association for Fire Safety Science, an executive committee member and secretary of the Combustion Institute. Professor Hirano was also very active in international academic cooperation, education and students training. In particular, he made a great contribution to the development of fire science in China. Professor Toshisuke Hirano started the cooperation with University of Science and Technology of China in 1986 and participated the cooperative program between University of Science and Technology of China and the University of Tokyo sponsored by Chinese and Japanese governments, and served as the committee chair of the cooperative program between University of Science and Technology of China and the University of Tokyo from 1992 to 1994. Since then he systematically introduced the theory system, academic thoughts and research achievements on fire science to China. He also promoted the cooperation between China and many developed countries, such as Japan, UK, US, Russia and Australia. In addition, he personally conducted substantive cooperation with Chinese researchers, and helped them to train young researchers and build research institutes. With his great help, the fire science in China has been developing rapidly. He greatly assisted the founding and establishment of “State Key Laboratory of Fire Science”, which is the unique national research institution in the field of fire science in China, and nowadays becomes one of the well-known fire research institutes in the world with his help and support. Professor Hirano visited China for cooperation and education over 40 times and trained tens of talents on fire science and technology for China. He made an outstanding contribution to the development of fire safety science in China. Thanks for his great work, he was awarded “Friendship Award” by the Chinese government in 1992, and the China International Science and Technology Cooperation Award in 2003. His passing away is a huge loss to the combustion and fire communities all over the world. His passing away is particularly a great grief to the researchers and colleagues in the field of fire safety. Although Professor Hirano has passed away 1.5 years ago, his spirit of seeking truth, his meticulousness and noble morality will remain in our memories and in our thoughts. Professor Hirano will live forever in my heart.

Part II Plenary and Invited

2

Fire Resistance of Multistorey Timber Buildings Andrew H. Buchanan

Abstract

Tall timber buildings are becoming popular around the world. This paper discusses options for setting the level of fire resistance in multistorey timber buildings. Fire resistance of timber structures involves a paradox, because it is well known that heavy timber construction has excellent fire resistance in severe fires, but it is also well known that burning of wood has contributed to severe damage and loss of life in some fires. This paradox often leads to contradictory statements on fire safety in timber buildings. This paper summarises recent international reports and proposes a consistent method of deciding which parts of multistorey timber buildings need additional fire protection. Modern fire engineering designs of steel and concrete buildings rely on full “burnout” of any fire compartment, with no fire spread and no collapse, through the full period of fire development and decay, after which the building can be repaired. For timber buildings, the achievement of burnout is less certain, because of the residual fuel which is always present in the large timber structural elements. Design for burnout may require full or partial encapsulation of the timber, which may not be acceptable to the building’s owners or architects if they have selected wood for aesthetic reasons. Keywords

Buildings
Tall
Timber
Wood
Fire
Burnout

2.1

Recent Developments in Tall Timber Buildings

2.1.1

International Popularity

Tall timber buildings are becoming popular around the world. Recent examples include the Forte 9-storey apartment building in Melbourne, Australia (Fig. 2.1), the UNBC 8-storey educational building in Prince George,

Canada (Fig. 2.2), and the Treet 14-storey building under construction in Bergen, Norway (Fig. 2.3). In addition, there are many proposals for multistorey timber buildings, including a 15-storey office building in Ottawa (Fig. 2.4), a 24-storey building in Vienna, a 30-storey apartment building in Canada (Fig. 2.5) and a 42-storey mixed-use building in Chicago, USA (Fig. 2.6).

2.1.2 A.H. Buchanan (*) Department of Civil and Natural Resources Engineering, University of Canterbury, Christchurch, New Zealand e-mail: [email protected]

Fire Concerns in Tall Buildings

Fire safety is a major concern in all tall buildings. The main additional issues for tall timber buildings include the following:

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_2

9

10

A.H. Buchanan

Fig. 2.3 Treet building, Bergen, Norway (www.woodskyscrapers. com) Fig. 2.1 Forte building, Melbourne (www.woodskyscrapers.com)

Fig. 2.2 University of Northern British Columbia, Prince George, Canada (www.unbc.ca)

Fig. 2.4 Proposed 15-storey building, Ottawa, Canada (Douglas Consultants, Quebec City)

2

Fire Resistance of Multistorey Timber Buildings

11

– Burnout. What happens after the fire burns itself out? – Fire resistance, for fire containment and structural performance. This is the subject of the paper. Different countries have very different requirements for fire safety in timber buildings. At the open end of the spectrum, countries like New Zealand have no area or height limits for structural timber, simply requiring that the performancebased requirements are met. On the other hand, countries like Australia require that only “non-combustible” construction is used for buildings taller than two storeys, without special dispensation and special study. Some European countries require a non-combustible stairwell and lift shaft for tall timber buildings. This paper attempts to provide a rational and consistent framework for specification of fire resistance in tall and very tall timber buildings.

2.1.3

Fig. 2.5 Possible 30-storey building, Canada [4]

Recent Reports on Fire Safety in Timber Buildings

There are a number of recent major international reports on fire safety in timber buildings: • • • • • • •

Technical Guideline for Europe [1] The Case for Tall Wood Buildings [2] Fire Safety Challenges of Tall Wood Buildings [3] Tall Wood Buildings in Canada [4] The Timber Tower Research Project [5] Use of Timber in Multi-Storey Buildings [6] Fire Resistance of Timber Structures NIST White Paper [7]

All of these reports confirm that well-designed timber buildings can have very good fire safety. Careful design is needed to ensure that safety is ensured in every phase of construction and use, throughout all possible fire scenarios. Automatic fire sprinkler systems are strongly recommended for all tall buildings, regardless of materials. Most of the reports do not consider the potential problem of burnout in the decay phase of an uncontrolled fire if the sprinklers do not work for any reason. The reports all recommend protection of wood by partial or full encapsulation which will avoid rapid fire spread and which will increase the fire resistance of protected structural elements. Fig. 2.6 Possible 42-storey building, Chicago [5]

– Early fire hazard. Flammability of large areas of exposed internal wood surfaces. – Exterior fire spread. Vertical fire spread via combustible facades or cavities. – Fires during construction. This is a particular problem for light timber buildings.

2.1.4

The Dilemma of Encapsulation

Encapsulation refers to covering wood surfaces with non-combustible materials. Encapsulation can improve fire safety, but it does not immediately solve all fire problems in timber buildings. Full encapsulation will require that the

12

A.H. Buchanan

wood be encapsulated with enough layers of protective material to prevent any ignition or charring of the wood in a complete burnout of the fire compartment, giving the same fire resistance as any non-combustible materials. Partial encapsulation will prevent any rapid fire spread on wood surfaces, but the smaller number of layers may fall off before burnout, exposing structural timber to the later stages of a severe fire. In both cases, there is an aesthetic dilemma because building owners and users and their architects want to see the wood, whereas the fire engineers may need to cover it all from view.

2.1.5

and ceiling linings are the worst hazard. Performance improves in rooms with protected ceilings and only the walls exposed. A more detailed assessment of pre-flashover fire safety is beyond the scope of this paper.

2.2.2

Post-flashover Fire Safety

For performance-based fire design, it is essential that the performance requirements be clearly established. The objectives can be summarised in the bullets below:

Fire Resistance of Heavy Timber What Are We Trying to Achieve?

Heavy timber has excellent fire resistance, which is well documented in the literature [8]. This excellent behaviour is a result of the slow and predictable rate of surface charring in severe fires, leading to simple calculation of fire resistance by subtracting the charred area and a thin layer of heataffected wood from the original cross-section. As a result of this charring behaviour, unprotected heavy timber structural elements have excellent fire resistance, much better than unprotected structural steel, for example. The big question for designers is this: does good fire resistance make a safe building? The answer is “not always” depending on the height and use of the building and what kinds of fire might occur, as will be discussed below.

2.2

Performance-Based Design

Design for fire safety is rapidly moving from prescriptive codes into performance-based design, where designers provide the minimum level of performance required by performance-based (or objective-based) national fire codes. The fire safety design must consider all phases of fire development from the pre-flashover fire through flashover to the fully developed fire, and eventual decay.

2.2.1

Pre-flashover Fire Safety

The pre-flashover fire is of great concern to life safety, but of very little concern for structural fire performance. In the pre-flashover phase, the concern is about the “early fire hazard”. For timber buildings, the main early fire hazard concern is about the flammability and combustibility of the internal lining materials in the room of fire origin, or adjacent spaces. It has been shown that large areas of combustible surfaces lead to rapid fire growth and faster time to flashover. Rooms with combustible materials for all wall

• • • •

The same level of fire safety as in other buildings. Design to meet code-specified performance levels. Occupant safety in all fires. Tall buildings must remain standing after the fire, even if no firefighting services are available.

To meet these objectives, the worst case scenario must be considered. That is the case where a disaster such as an earthquake or terrorist attack disables the automatic fire sprinkler system, leaving the building to resist the fire only through the passive fire protection built into the fabric of the building. If the worst happens, the compartment of fire origin must have sufficient fire resistance to contain the fire and prevent structural collapse throughout the full process of fire growth, development and decay. Designers of steel and concrete buildings work on the basis that if the compartmentation works and the structure is designed for adequate fire resistance, the fire will eventually go out after the fuel is all consumed, and the structure will cool to ambient temperatures after any heat remaining in the building has dissipated. This scenario is less certain for timber buildings, because there will always be some fuel present in the wood components of the timber structure, leading to the possibility of slow charring which might continue long after the main fire has gone out. In this case, it is only possible to design a tall timber building with the same level of fire safety as a steel or reinforced concrete building if design for burnout can be accomplished.

2.2.3

Can We Design Timber for Burnout?

The only certain way to design for burnout in a timber building is to apply full encapsulation, so that none of the

2

Fire Resistance of Multistorey Timber Buildings

13

structural timber ever begins to char, throughout the full process of fire growth, development and decay. The required encapsulation will depend on several factors: • The fire severity and duration of the burning period • The rate of temperature drop due to ventilation in the decay phase of the fire • The effectiveness of encapsulation materials • Intervention after the fire is out

2.3

Active and Passive Fire Protection

2.3.1

Active Fire Protection

The main active fire protection measure is an automatic sprinkler system. The biggest question for designers and code writers is the effectiveness of the sprinkler system in tall buildings. If sprinklers can be guaranteed to work, the additional requirements for fire safety and fire resistance are minimal, but if the sprinklers do not work for any reason, timber buildings are in a special category, needing more careful attention. The possible options are: 1. The sprinklers work as intended: what happens? • No flashover, no charring, no problems. 2. Sprinklers don’t work: what is the fire service response? • Rapid response – Only with good access routes and the fire on the lower floors • Slow response – Due to poor access or fire on upper floors of the building • No response – Due to a major earthquake or a terrorist attack

2.3.2

Passive Fire Protection

Design of tall timber buildings will require all aspects of passive fire protection to be considered, including egress, fire resistance, concealed cavities, etc. This paper concentrates on passive fire protection by protecting the wood surfaces, by paint, other applied finishes or encapsulation. Several options are: 1. All timber walls and ceilings exposed to view 2. All walls exposed to view, ceilings protected 3. All beams and columns exposed to view, walls and ceilings protected

Fig. 2.7 Wood walls and ceiling exposed

Fig. 2.8 Wood ceiling and furniture exposed

4. All wood protected with one layer of gypsum plaster board (partial encapsulation) 5. Full encapsulation of all wood with several layers of plaster board Figures 2.7, 2.8, 2.9, 2.10, and 2.11 show a range of interior views with different amounts of structural timber exposed to view (and exposed to fire). Figure 2.12 shows a multistorey timber building in Zurich during the process of protecting all the wood ceilings, beams and columns with fire-resistant board encapsulation. Table 2.1 shows all the combinations of active and passive fire protection in a matrix format. It can be seen that with a good sprinkler system in operation there is no problem because there is no flashover and no charring of structural timber (column 2). However the potential problems get

14

A.H. Buchanan

Fig. 2.12 Wood ceiling, beams and columns during process of encapsulation with fire-resistant board Fig. 2.9 Wood columns, beams and part ceiling exposed

worse and worse with if the sprinklers do not work and the response of the fire service is slow (columns 3–5). Things get more complicated when we combine the options for passive fire protection. With full encapsulation, a timber building is just as safe as a steel or concrete building, regardless of fire service intervention (the bottom row in Table 2.1.) However with partial encapsulation or no encapsulation, the potential problems get worse because the likely extent of charring can increase dramatically. The thick solid line separates the areas of no damage from those areas of moderate or serious damage.

2.3.3 Fig. 2.10 Wood columns and beams exposed

Fig. 2.11 Wood columns exposed

Effect of Building Height

The matrix in Table 2.1 is dependent on the height of the building, because minor problems for low-rise buildings can become huge problems for tall buildings. Table 2.2 defines different building heights from low-rise to high-rise, with suggestions for passive fire protection depending on sprinkler protection and the height of the building. Table 2.2 shows suggestions for passive fire protection depending on sprinkler protection and height of the building. In Table 2.2, the rows for “No sprinklers” and “Normal sprinklers” have their usual meanings. The additional row for “Special sprinklers” has been added to allow designers to take special precautions, such as a dedicated water tank in the building to ensure that the sprinklers have water, even if the street mains are destroyed by an earthquake or landslide, in which case a greater area of visible wood could be provided in tall buildings. The diagonal shading across Table 2.2 helps to divide up the options into sensible groupings.

2

Fire Resistance of Multistorey Timber Buildings

15

Table 2.1 Matrix of fire damage scenarios for combinations of active and passive fire protection

Active fire intervention:  Fire size:



Passive fire protection: Walls and ceilings exposed

Sprinklers work No flashover

No charring

Sprinklers don’t work Rapid response Flashover, one room. Extinguished by fire service Minor charring. Stopped by fire service.

Slow response

No response

Flashover and decay, one floor. Cooled by fire service

Full burnout, one floor.

Moderate charring. Stopped by fire service.

Very Severe charring. (large areas)

Severe charring continues.

Walls exposed

No charring

Minor charring. Stopped by fire service.

Moderate charring. Stopped by fire service.

Severe charring (smaller areas)

Beams and columns exposed

No charring

Minor charring. Stopped by fire service.

Limited charring. Stopped by fire service.

Severe charring (limited areas)

Wood protected with one layer of gypsum Full encapsulation.

No charring

No charring

Some gypsum falls off. Minor charring.

Gypsum falls off Severe charring later

No charring

No charring

No charring

No charring

Table 2.2 Suggestions for passive fire protection depending on sprinkler protection and height of the building

Height

Low-rise

Mid-rise

Tall

Very tall

Hi-rise

Storeys

1-2

3-5

6-8

9-15

>15

Likely escape

Quick escape

Slow escape

Assisted escape

Assisted escape

Difficult escape

No sprinklers

Local areas exposed

No exposed wood

Not allowed

Not allowed

Not allowed

Normal sprinklers

Large areas exposed

Local areas exposed

No exposed wood

Full encapsulation

Full encapsulation

Special sprinklers

Large areas exposed

Large areas exposed

Local areas exposed

No exposed wood

Full encapsulation

16

2.3.4

A.H. Buchanan

Will This Work? Who Decides?

The options in Table 2.2 are simply suggestions to show the range of possibilities. As tall wood buildings become more popular, it will be necessary for code writers in different countries to adopt requirements which reflect these ideas in a rational way. More research including quantitative risk assessment may help to further define the options.

2.4

Research Needs

This paper has raised many questions, many of which do not have clear answers without more research. Research in the following areas will help design engineers and code writers to make good decisions about the issues raised in this paper. 1. Severity of design fires (needed for all structural materials) 2. Contribution of timber to the fire load 3. Charring rate of wood as a function of fire exposure 4. Self-extinguishment of charred wood 5. Fire performance of encapsulated timber 6. Dangers of combustible wooden fac¸ade claddings 7. Effect of different combinations of passive and active fire protection 8. Quantitative risk assessment, to include all these items

2.5

Conclusions

This paper has suggested a range of solutions to current concerns about design for fire resistance in tall timber buildings. Different levels of protection or encapsulation of the wood structure should be provided, depending on the height of the building and the reliability of the water supply for the mandatory sprinkler system. The biggest question for designers and code writers is the effectiveness of automatic sprinkler systems in tall

buildings. If sprinklers could be guaranteed to work, the additional requirements for fire safety are minimal, but if they may not work for any reason, timber buildings are in a special category, needing more careful attention than buildings from non-combustible materials. The paper concludes that tall timber buildings can have equivalent fire safety to buildings of traditional materials, provided that attention is given to the overall fire design strategy, combined with careful design and detailing, which will require some or all of the wood structural elements to be hidden from view by full or partial encapsulation. Acknowledgements The author acknowledges the assistance of Prof ¨ stman at SP Sweden and Andrea Frangi at ETH Zurich, Birgit O colleagues Charley Fleischmann, Michael Spearpoint and Anthony Abu at the University of Canterbury, as well as numerous fire engineering designers.

References ¨ stman B et al (2010) Fire safety in timber buildings – technical 1. O guideline for Europe. SP Report 2010:19. SP Technical Research Institute of Sweden, Stockholm 2. Green M (2012) The case for tall wood buildings – how mass timber offers a safe, economical, and environmentally friendly alternative for tall building structures. mgb Architecture þ Design, Vancouver 3. Gerard R, Barber D, Wolski A (2013) Fire safety challenges of tall wood buildings. Arup North America Ltd, San Francisco, and Fire Protection Research Foundation Quincy, MA, U.S.A. 162pp 4. FPInnovations (2013) Chapter 5 – Fire safety and protection. In: Technical guide for the design and construction of tall wood buildings in Canada. FPInnovations, Vancouver, pp 223–282 5. SOM (2013) The timber tower research project. Skidmore, Owings and Merrill LLP, Chicago. 66pp 6. Smith I, Frangi A (2014) Use of timber in tall multi-storey buildings, structural engineering document SED 13. International Association for Bridge and Structural Engineering IABSE 7. Buchanan AH, Ostman B, Frangi A (2014) Fire resistance of timber structures. NIST white paper. National Institute of Standards and Technology, Washington, DC 8. Buchanan AH (2001) Structural design for fire safety. Wiley, West Sussex: 421pp

3

Ultimate Strength and Its Application to PostEarthquake Fire Resistance of Steel Frames in Fire Hiroyuki Suzuki

Abstract

Predicted ultimate temperatures of steel frames in fire are derived based on the plastic nature of frames of stress redistribution and redundancy, which is applicable to practical fire resistance design. The primary ultimate temperatures of portal as well as multi-story frames are determined with simple calculation. Comparing these with refined numerical solutions verifies applicability of the method of prediction to practice. With this method and numerical verification, theoretical buckling temperatures are found to be also useful in practice for heated columns. As an application of this prediction, post-earthquake fire resistance of steel frames is clarified in terms of ultimate temperature of pre-drifted frames, which is as high as the ultimate temperature intact frames show, if fire is limited within not so enlarged space on the floor. Thermal post-earthquake resistance of two-ply gypsum board wall is also studied based on loading and subsequent heating tests. Although the present prescription for dry walls is found insufficient to make pre-damaged walls fire resistant if damage is not slight, an easy and practical improved construction is also shown to keep damaged walls from premature failure in fire. Keywords

Stress redistribution
Redundancy
Primary ultimate temperature
Permanent story drift
Gypsum board

Nomenclature E, Et F l leff MpB, MpC N

Young’s and tangent moduli resp. (N/mm2) Lifting up force of key beams (N) Full span length of a beam (mm) Effective buckling length of a column (mm) Full plastic moments of a beam and column resp. (m) Post-buckling residual strength of a column (N)

P q T

Compressive force a column carries (N) Uniform beam load (N/mm) Member temperature ( C)

Greek Symbols 2, 2y κ λ σ, σy

Strain and yield strain of steel at ambient temp. resp. Ratio of residual high temp. Yield strength of steel to that at ambient temp. Normalized slenderness ratio of a column Stress and yield stress of steel at ambient temp. resp.

H. Suzuki (*) Institute of Engineering Mechanics and Systems, University of Tsukuba, Tennodai 1-1-1, Tsukuba 305-8573, Japan e-mail: [email protected] # Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_3

17

18

3.1

H. Suzuki

Introduction

For fire resistance of frames it is desired to rationalize the design method, since ways dependent on detailed prescription have long controlled design procedures and been accepted. The rationalization must be oriented toward enhancing logical clarity of the design methodrogy, safety of frames against disaster and freedom in design procedures. For steel frames, the study on mechanics of heated frames enables us to overcome the situation involved, since the high ductility of the material performs advantageous roles mechanically when frames undergo fire. For earthquake resistant frames in steel, constituent stockier members and thicker plate elements help frames, even when they are heated, to utilize this ductility of the material more successfully. In view of this, it is the objective of this paper to summarize both the theoretical importance and practical applicability of the plastic nature of heated frames. To consider plastic behavior of a heated frame means not to consider high thermal stresses generated at local parts but to consider overall stability of the frame instead. To consider overall behavior means to consider the process of stress redistribution working on a frame. Stress redistribution is derived from redundancy of a frame. Fire resistive ultimate state of a frame is the state when the redundancy is fully exhausted in the frame in fire. In the following section, we discuss theoretical aspects for the above mentioned problems and solve the ultimate states of example frames in fire including both simple portal and general multi story frames. Two topics relevant to ultimate fire resistance of steel frames are also discussed. One is structural post-earthquake fire resistance of frames. This has not been much discussed previously despite its importance, for it is difficult to find a practically appropriate answer. Although the difficulty remains unchanged, an important view point can be drawn and is given in the subsequent section for this problem. Another is thermal post-earthquake fire resistance of dry partition walls. After a frame suffers large permanent story drift during a strong earthquake and the installed partition walls are inevitably damaged at the same time, does not such damage of walls cause ill effect on resistance against subsequent fire? This is also an important problem on postearthquake fire. In the final section, an experimental study on this problem is introduced. Since gypsum boards used for walls are not ductile, present authorized prescription for dry walls is found to be insufficient for post-earthquake fire if pre-damage of walls is significant. However a very simple and practical way is also found, which not only makes damaged walls restore the performance but also improves the method of construction itself for dry walls.

3.2

Ultimate State of a Frame in Fire

3.2.1

Thermal Stress, Stress Redistribution and Ultimate State

Throughout the period a frame is exposed to fire without collapse, some or large thermal stresses are generated and continue to exist in its constituent members, whether they are heated or not. Thermal stresses are generated when heated and lengthened longitudinal fibers of a member are constrained elastically and redundantly by the member’s neighborhood. This results in the fact that all thermal stresses within a whole frame form a self-balancing force system. When a frame is subjected to fire and the temperatures of some constituent members rise, the resultant stress, which is the sum of mechanical stress balancing with external force plus such thermal stress, becomes sooner or later as large at certain points as the frame yields and goes into elasticplastic behavior. Reduction in yield stress of heated steel may accelerate yielding of the frame somewhat. However, for most cases, this yielding does not indicate direct collapse of the frame in that a limited amount of plasticity cannot make the frame unstable locally nor overall, since elasticplastic deformation of an individual member is not only stable more or less but a building frame itself is also highly redundant usually. We assume herein that major individual members of a frame are proportioned to have some plastic deformation capacity. Frames designed for earthquake resistance can sufficiently satisfy such requirement. Then we can increase member temperatures further to know post yielding behavior of a frame in fire. In this way a frame in fire starts to exhibit rather stable elastic-plastic behavior but with several weakened members in strength due to increasing member temperatures. The post-yielding behavior of an increasingly heated frame is characterized by two consecutive distinct phases, i.e. first and second halves, in view of mechanical effect. The first half is characterized by rather severe increase in resultant stress everywhere in the frame, which is an extension of frame’s pre-yielding behavior and caused mainly by increasing thermal strains. However increase in resultant stress ceases at a certain stage of heating and it turns to decrease subsequently, which indicates that the state of the frame enters into the second half. The mechanics within this latter half throughout is characterized now by fiber to fiber and/or member to member stress redistribution. This stress redistribution is always done from heated, larger strained and plastic parts of the frame to other elastic parts with smaller strains which are not always heated. It is because the shortage to carry external

3

Ultimate Strength and Its Application to Post-Earthquake Fire Resistance of Steel. . .

load of elastic-plastic and highly heated parts with reduced yield strengths must be compensated by stress increase of the elastic parts where there are still rooms to increase burden. Such stress redistribution continues to work as far as the redundant nature of the frame remains effective. However the redundancy deteriorates gradually, since, in accordance with the process of the stress redistribution, existing plastic deformation in progress and newly generated plastic zones come to behave like plastic hinges and they therefore spend one redundancy of the frame after another. It must be noticed that release of thermal stress takes place as exactly synchronous with the above stress redistribution. One reason for this is because the above shortage in strength of highly heated parts becomes a cause of weakening of mutual constraint action between them and their neighborhood. Another is because the stress redistribution spends redundancy of the frame, which weakens also the mutual constraint action. However, as far as stress redistribution works and therefore the frame behaves stably, we always find thermal stresses in key members. It is right to say conversely that existence of any thermal stress in a frame indicates that stress redistribution is still working and therefore the frame does not collapse immediately in fire. We then reach an important remark that the ultimate state of the heated frame against fire is not only the state when stress redistribution is no more possible, but also the state when all thermal stresses which have been generated and existed until this state come to be fully released and vanished in all key members forming mechanism motion. It is not denied however that some thermal stresses remain to exist in other members of the frame. Therefore, at the ultimate state, all stresses occurring in such key members are spent fully and only to carry the external load without being spent for mutual constraint, from the fact of which we can find the equation of equilibrium in forces at the ultimate state of heated frames. Plastic theory of structures says that the plastic limit state of a frame under monotonic load does not depend on initial thermal strains. The reasoning of the theorem has been made originally for elastic-perfectly plastic frames with the small displacement theory. Substantial part of the theory can be applicable to frames made of practical steel materials which, when heated, show significant yield strength reduction. In these our practical cases, the following theoretical remarks can be drawn. As shown in the above, the ultimate state of a heated steel frame in fire is determined simply by obtaining balance of forces between all occurring stresses in the key members and external load. In addition, since generated stresses in the key members at ultimate state are approximately equal to yield strengths in steel because of slight strain hardening at high temperatures. Therefore the ultimate

19

state so found is almost independent of initial thermal strains as the theory says if no premature failure such as member or plate element buckling is predominant.

3.2.2

Portal Frame in Fire

3.2.2.1 Primary Ultimate State of a Portal Frame I [1] The mechanical behavior of the frame shown in Fig. 3.1a is considered up to its ultimate state, which has sound columns but a beam subjected to a uniform beam load q and to uniformly increasing temperature T. As the frame is heated, combined bending moment M and axial forces P are generated in the beam. The latter P is compressive and of thermal stress caused by constraint between the neighborhoods. In this and next subsections, we ignore any geometrical nonlinearities but consider solely the small displacement together with the simple plastic theories of frames. The primary ultimate temperature of the frame will be predicted based on this assumption as the following. At a certain state of heating with a temperature T1, the mechanical system of the frame becomes full plastic. Figure 3.1a shows a corresponding mechanism with three plastic hinges whose full plastic states are identical and represented by a point A(P1, M1) on the interaction curve as shown in Fig. 3.1b. This curve represents combined strength of the plastic hinge of the beam at constant T1. However the frame does not collapse yet then, since this mechanism is not active at T1. In fact, if the mechanism were active, the plastic hinges should not only rotate but shrink longitudinally due to compressive P1. Knowing that M1 of the three plastic hinges remains constant balancing with q, this shrink should release the constraint and lower P1 itself under constant M1. This might make the stress state fall from point A into elastic zone inside the interaction curve. In this way temperature can increase to T2 and a new full plastic state B(P2, M1) is attained as shown in Fig. 3.1b with a

b M

C

B T3

A T2

T = T1 P

Fig. 3.1 A collapsing heated beam in a portal frame (a) formed mechanism (b) changing full plastic state during stress redistribution

20

H. Suzuki

the same mechanism. Point B is on a shrunk interaction curve since this curve accords with the reduced material strength of higher temperature T2(> T1) and has the same M1 as point A because of constant q. We then see in Fig. 3.1b that P2 is smaller than P1, which indicates that the compressive thermal stress resultant is released from P1 to P2 resulting from stress redistribution from state A to B. As before, state B is not ultimate yet. The ultimate state of the frame is at state C(0,M1) on the interaction curve at T3, since, at higher temperatures than T3 and on more shrunk interaction curves, the frame could not resist M1 which is needed to carry q. Therefore the ultimate temperature T3 can be predicted by solving the following simple equilibrium equation.

shown in Fig. 3.2a. However, state E is still stable. In fact, if the system were unstable and the column heads moved outward with rotation of the plastic hinges, thermally generated beam’s compression and column’s M should decrease simultaneously due to decreasing mutual constraint and this might make state E fall into the elastic zone. This means the column heads cannot move outward at state E. Via another stable full-plastic state F(P, M5) at a higher temperature T5 in Fig. 3.2b, the state of stress attains the ultimate G(P, 0) at T6. Keeping P constant, thermal bending moment M is released from M4 to zero until state G is attained and this entire process corresponds to stress redistribution. Since M is vanished at state G, the ultimate temperature T6 becomes the solution of the equation

ql2 =4 ð¼ M1 Þ ¼ 4 MpB ðT 3 Þ;

MpC ðP, T 6 Þ ¼ 0;

ð3:1Þ

where l denotes beam length and MpB(T3) is exactly the beam’s full plastic moment without axial force component at T3. Ignoring strain hardening, (Eq. 3.1) says that all fibers with a yield stress in the three plastic hinges work only to carry external load q at the ultimate state. This means that the beam becomes free from thermal axial force at the ultimate temperature. Therefore the primary ultimate state is independent of thermal stress and of initial thermal strain.

3.2.2.2 Primary Ultimate State of a Portal Frame II [1] A similar observation is drawn for the uniformly heated frame shown in Fig. 3.2a, where both columns are under constant vertical load P respectively and the beam is assumed to be sound in spite of heating. As the frame is heated, both columns are pushed outward by thermal elongation of the beam and therefore the columns are under combined bending and axial compression P. Let M denote bending moment at the column ends which represents thermal stress. At temperature T4, certain full-plastic state E(P, M4) in Fig. 3.2b may be attained, in which M4 is a magnitude of M at this state. Corresponding formation of plastic hinges is a

b M T = T4 E T 5

T6

F G

Fig. 3.2 Collapsing heated columns in a portal frame

P

ð3:2Þ

where MpC(P,T6) is full plastic bending moment of the column cross section under P and T6. Ignoring strain hardening, (Eq. 3.2) indicates that all fibers with a compressive yield stress of the entire column cross section work only to carry external P at state G. Therefore the predicted primary ultimate temperature is free from initial thermal strain similarly as before. The predicted temperature from Eqs. 3.1 and 3.2 is called herein a primary ultimate temperature of a frame in that this is obtained based on the small displacement and simple plastic theories of frames.

3.2.2.3 Numerical Examples for Portal Frames [1, 2] Validity of Eqs. 3.1 and 3.2 is examined in the following by comparing the prediction from these equations with refined numerical fire response analysis which takes realistic nonlinear material behavior at high temperatures as well as geometrical nonlinearities of displaced frames into consideration. Sample frames used are the left half subassemblages of two story steel portal frames shown in Fig. 3.3. It is assumed that a beam and a column of the lower floor are both heated uniformly and increasingly while the upper floor remains at ambient 20  C. The frames are made of JIS mild steel SS400/SN400. AIJ recommends their σ – E curves under various constant temperatures for fire resistance design, which are shown in Fig. 3.5 [2]. These curves are used herein and function σ(E, T ) is used to represent them. Constant thermal expansivity 1.2  105 is assumed for any temperature of steel. It is assumed that, following seismic design, the lower column is made stocky with normalized slenderness ratio of 0.25 and is subjected to axial compression P with 30 % of its ambient yield strength. Assuming further various magnitudes of beam load, a series of numerical analysis

3

Ultimate Strength and Its Application to Post-Earthquake Fire Resistance of Steel. . .

21

500 RT∼ 300

400

q

400

300

stress

h

h

500

200

600

100 0

P

L

700 0

0.02

0.04 0.06 strain

0.08

0.1

Fig. 3.5 Stress vs. strain relations of SS400/SN400 steel under constant temperatures, which are specified by AIJ Recommendations [2]

Fig. 3.3 A structural subassemblage in fire

theoretical buckling temperature

1.5 existing data of heated SS400

1

beam load

°C

state Robertson's reached

numerical ultimate state

1

0.5 0.5

primary ultimate state

0

200

400

κ(T) 600

0

800

ulmate temperature [⬚C]

0

200

400

600

800

1000

temperature in steel [⬚C]

Fig. 3.4 Ultimate temperatures of frames shown in Fig. 3.3 in fire, in which MpBRT is beam’s full plastic moment at ambient temperature

Fig. 3.6 Effective yield stress of heated steel which is compatible with the stress according to strain 1 % on a σ(E,T ) curve

have been conducted. Figure 3.4 summarizes the result of analysis. Ordinate of a hollow circle in Fig. 3.4 represents the temperature beyond which no convergent balance can be obtained between stress and external load in the analysis and it is thought of as the refined ultimate temperature of the frame. A solid circle represents the state when analyzed beam displacement reaches Robertson’s criterion 4L2/(800d) [3] in which d denotes the depth of the beam cross section. In parallel the predicted ultimate temperatures obtained from Eqs. 3.1 and 3.2 are also shown by a solid folded line in Fig. 3.4, in which the stress at 1 % strain of a curve in Fig. 3.5 is assumed to be effective yield strength of the heated steel. These effective strengths are represented in nondimensionalized form as a function

in which F is yield stress 235 N/mm2 of SS400/SN400 steel used as standard for domestic structural design. Shape of function κ(T) is shown in Fig. 3.6 along with several existing test results of heated SS400. This shape of κ(T ) is also recommended by AIJ [2]. In Fig. 3.4 the inclined line segment accords with Eq. 3.1 and vertical one Eq. 3.2, while the horizontal one represents the maximum load the beam can carry. Comparison between numerical and predicted results in Fig. 3.4 tells us 2 points. First is that the refined ultimate temperatures of frames with column failure mode are slightly lower than the simple primary one predicted by Eq. 3.2 due to column buckling. Since the columns are stocky and their post-buckling behaviors are less unstable, errors of the prediction remain small. Second point is that, for beam failure cases, the numerically obtained ultimate temperatures are significantly higher than the primary ones

κðT Þ  σ ð0:01, T Þ =F

ð3:3Þ

22

3.2.2.4 High Temperature Buckling of a Column Installed in a Portal Frame High temperature buckling of columns must be considered under the same mechanical, geometrical and heating conditions as they are installed in a frame. A previously illustrated simple frame shown in Fig. 3.3 is adopted again and buckling of its installed column is discussed under the same loading and heating conditions as before. Several factors complexly influence buckling of heated installed columns. They are enumerated as follows. (i) Such a column has experienced full-plastic states due to combined bending and axial compression like the states E, F, etc. in Fig. 3.2 before buckling occurs. This means the column has behaved as an elastic-plastic beam-column subjected to large antisymmetric thermal bending moment distribution. However this bending moment disappears completely when the column buckles, since this thermal stress owes to stable mutual constraint between the neighborhoods before buckling. (ii) Buckling and post-buckling of the installed column is affected by rotational constraint at its both ends by neighborhoods. (iii) Preceding plastic deformation in the form of a beam-column gives geometrical imperfection to buckling of the column. (vi) Unstable buckling motion accompanies elastic unloading at various fibers of the column cross sections, since the mode of longitudinal strain distribution when the column buckles is quite different from the past distribution in the form of a beam-column. In view of the complexity above, a comparison is made in the following between refined numerical solutions which can take all above factors into account together and simple theoretical solutions. The former is the aforementioned numerical ultimate temperature of an entire frame. The latter is a predicted critical temperature Tcr when a constant central compression P working on a simply supported column attains the tangent modulus load under uniformly increasing member temperatures. In the prediction, the cross section of the lower column is taken and the column length leff is assumed to be 0.7 times the story height. Constant P is

assumed to be the vertical nominal load which the lower column carries initially at ambient temperature. Tcr is then the solution of the following equation.

 π 2 Et T cr , PA I Et T cr , PA 1 P P¼ or p ¼ ¼ ð3:4Þ Py E λ2 l2ef f where the following notation is used. I and A are moment of inertia and area of the column cross section, respectively. E is Young’s modulus and Et tangent modus at σ ¼ P/A on σ  E curve under temperature pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Tcr. λ ¼ lef f Py =π 2 EI is normalized slenderness ratio of the column where Py is its yield compressive force at ambient temperature. Figure 3.7 shows the results of above both analyses together with ALJ recommendation. It is found in the figure that the numerical ultimate temperatures of the frames with column buckling are nearly equal to or a little greater than the theoretical ones. We may therefore say the following on this problem. The ultimate state of the frames shown in Fig. 3.3 with column buckling mode is almost independent of preceding history of thermal bending stress the frame has experienced. Rotational constraint at column ends must be appropriately evaluated for buckling of installed columns. In view of this, the effective buckling length factor 0.7 for columns of rigidly jointed frames may be a good but is a little arbitrary choice. Effects mentioned in the above items (iii) and (iv) on the problem seem negligible. Therefore the ultimate temperature of a frame with column buckling is almost identical to the theoretical buckling temperature of a simply supported and centrally compressed column with a suitable effective buckling length. Going back to Fig. 3.4, the vertical dotted line is the theoretical buckling temperature of the column used, which agrees very well with frames’ numerical ultimate temperatures. 1000

buckling temperature［⬚C］

predicted by Eq. 3.1. This is because, in numerical solutions, the beam displaces largely showing stable catenary action after it is bent plastically. Equation 3.1 has rather better correlation with Robertson’s criterion, for his criterion accords with the displacement which is several times the magnitude of the elastic limit of a simple beam under uniform load. It is now clear that the numerical ultimate temperature falls below the primary prediction if geometrically nonlinear post-buckling takes place unstably, while the former exceeds the latter conversely if geometrically nonlinear catenary action develops stably. Therefore the primary ultimate temperature can be regarded exactly as a primary ( first-order) approximation to the refined numerical one.

H. Suzuki

AIJ design formula theoretical

800 p = 0.1

600 p = 0.3

400

p = 0.5 p = 0.1 p = 0.3 p = 0.5 p = 0.7 for numerical

200 0

0

0.2

0.4

0.6

p = 0.7

0.8

1

λ of heated column Fig. 3.7 Buckling temperatures of uniformly heated steel columns

3

Ultimate Strength and Its Application to Post-Earthquake Fire Resistance of Steel. . .

3.2.3

Column Buckling and Redundancy of Frames in Fire

3.2.3.1 Value of Theoretical Buckling Temperature of Heated Columns [2] The discussion in the previous section is limited to installed columns when longitudinal constraint on columns is small. The column installed in an inner compartment of a multistory frame is subjected to increasing compressive force in fire. The increment in compression is of thermal stress, since they lengthen due to heat and are constrained strongly by the neighborhoods. There is a possibility in this case that earlier column buckling due to increased compression might cause collapse of the frame at a lower temperature. This is the most important problem on high temperature buckling of installed columns and is discussed as follows. Example solutions are given first using refined numerical analysis. Analyzed is a 12 story and 3 bay plane frame made of SN400 steel as shown in Fig. 3.8. Seismic design is given for member proportioning based on the domestic Building Standard. Strengths of the beams are so determined that frame’s base shear coefficient Cb is equal to 0.21. Twelve types of fire are considered, for each of which fire occurs in the central interior compartment of a different floor and only a beam and columns facing the fire are heated uniformly. Results analyzed are shown in Fig. 3.9a, in which, taking the number of floor fire occurs to the abscissa, three critical temperatures for every case are obtained and plotted by different dotted marks along the ordinate. In each case of fire analyzed, the heated columns buckle first, which is observed as appearance of bent buckled mode of displacement along the length of the columns and the

h 〃 〃

h lout

lin

lout

Fig. 3.8 Multistory and three-span frames used to study redundancy of frames in fire (Figure shows mechanism formed at ultimate state)

23

temperature at this instant is plotted by a hollow triangle in the figure. Since these columns are thermally lengthened and strongly constrained by their neighborhoods longitudinally, the increased compressive force thereby makes the columns buckle at a lower temperature than their theoretical buckling temperature under nominal compression. The latter is plotted by a solid circle on the same abscissa as shown in the figure. Nominal compression here is the magnitude of external load which the installed column carries initially at ambient temperature. This compression is a constant force and independent of any thermal stress. Therefore the obtained theoretical buckling temperature of the installed column is independent of thermal stress. The onset of buckling is not the ultimate state of the frame at all. Under continuing further heating, pre-buckling thermal lengthening of the column turns to post-buckling shortening regardless of continuing heating, since the post-buckling column behaves unstably and its strength deteriorates. However in parallel with this, constraining action of the neighborhood to the pre-buckling column turns to its lifting action in which the neighboring beams can hold the post-buckling column so as not to collapse downward. The latter action is also said to be stress redistribution mechanically in that a part of the vertical load which the pre-buckling column has carried is transferred to cooler and stable other columns by means of above beam’s lifting action. As previously noticed, this stress redistribution also synchronizes with release in thermal stress. In fact, strength reduction of the post-buckling column corresponds to reduction of mutual constraint between the column and the neighborhood, where the amount of strength reduced is said to be the released thermal stress itself. Since, in spite of column buckling, the frame can endure further increased temperatures owing to above stress redistribution, significantly higher ultimate temperature marked by a hollow circle in the figure is obtained for each case. Numerical solution diverges beyond this state, since such broadly unstable motion as shown in Fig. 3.8 begins to develop near this state. Different numerical solutions for another frame made of weaker beams with the base shear coefficient Cb of 0.125 is shown in Fig. 3.9b. Similar tendency is found for this case too as before. The following is found from comparison between Fig. 3.9a for frames with stronger beams and Fig. 3.9b. Individual columns start to buckle at lower temperatures for frames with stronger beams than those with weaker beams. This is because the stronger key beams bring the stronger constraint and the larger thermal compressive forces to pre-buckling columns. Conversely the same stronger key beams bring the higher ultimate temperatures of frames as seen in the same figures. This is because the stronger key beams have the stronger lifting capacities to post-buckling columns. The same key beams

24

H. Suzuki

Fig. 3.9 Three critical states of multi story frames in fire (Left (a) shows results of a frame with Cb ¼ 0.21 and right (b) shows that with Cb ¼ 0.125)

temperature[⬚C] 800

a

b

700

600

for a frame with 500

0

5

for a frame with

10

0

5

number of floor fire occurs

10

number of floor fire occurs

ulmate temperature theorecal buckling temperature temperature at onset of buckling

perform opposite mechanical roles between pre and post buckling behaviors of frames in fire. Knowing that the theoretical buckling temperature under nominal compression is always greater than the buckling onset temperature and less than the ultimate temperature of the frame as seen both in Fig. 3.9a, b, the first theoretical buckling temperature may be used for a conventional design critical temperature of a column. One reason is because this temperature can be simply calculated, since it is independent of thermal stress. Second is because this independency means that theoretical buckling temperature of a column is one of the structural characteristics independent of history of loading and heating and we can use this characteristic in practical design. Third that this temperature being significantly lower than the ultimate one gives us a safer criterion in view of redundancy and therefore ultimate strength performance of heated frames.

3.2.3.2 Ultimate Temperature of Overall Frames Taking Redundancy into Account [2, 4–8] The ultimate temperature of the previous multistory frame shown in Fig. 3.8 can be predicted by means of the same way with the simple plastic theory as the following. The ultimate state of the frame is the state when it forms a mechanism as shown in Fig. 3.8, which indicates that the beams can no more lift the buckled columns upward beyond this state. Since members forming the mechanism, which are buckled columns and all beams having active plastic hinges, play key role to estimate ultimate resistance of a frame, call them key members herein. When a frame is ultimate at temperature T, equilibrium of the forces acting on the key members and external load is written as N ðT Þ þ FðT Þ ¼ P

ð3:5Þ

where P is the sum of vertical load acting on all pre-buckling columns initially at ambient temperature. N(T) is the sum of post-buckling residual strengths of all buckled columns in their longitudinal direction and is decomposed in the form as N ðT Þ ¼ κ ðT Þ N res using function κ(T) defined by Eq. 3.3. F(T) is the vertically orienting maximum shear force all key beams can resist to lift the buckled columns upward, which represents the magnitude of redundancy of the frame in fire. First we assume that T in Eq. 3.5 is the ultimate temperature of the frame which is obtained from the previous refined numerical solution. Next we make a simple evaluation of F (T) and Nres at the above ultimate T as the following. X
 F* ð T Þ ¼ 2 κ ðT Þ MpBRT =l i ; i 2 key beams

N *res

¼

X j 2 buckled columns



qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi σ y A= 1 þ 4:5 Ey λ2

ð3:6Þ

j

where MpBRT is the full plastic moment of each key beam at ambient temperature. In Eq. 3.6, each term in the summation for Nres is known to be and used as an empirical postbuckling residual compressive strength of a bracing member in steel [2]. Based on the above assumption and evaluation, Fig. 3.10 shows the relationships between T and
 P  F* ðT Þ = N *res for a lot of frames in fire. From Eqs. 3.5 and 3.6 as well as from the relation between dotted marks and the curve κ(T ) shown in Fig. 3.10, the following observation can be made.

3

Ultimate Strength and Its Application to Post-Earthquake Fire Resistance of Steel. . .

0.5 0.4

k(T) curve

0.3 0.2 numerical solutions frame of span 6x6x6m frame of span 10x6x10m frame of span 14x6x14m

0.1 0

–0.1 200

300

400

500

600

700

800

T: ultimate temperature of frames [⬚C]
 Fig. 3.10 Figure to show the relation of calculated P  F* ðT Þ =N *res with numerically obtained ultimate temperature T is almost equal to the relation between κ and T PF* ðT Þ is N *res ðT Þ κðT Þ ¼ PF N res

nearly equal to or a little greater than

This shows the following two things. One is that F* ðT Þ  FðT Þ and N *res  N res . Second is that the ultimate state obtained from the numerical solution almost satisfies the equilibrium Eq. 3.5. This indicates that the mechanism shown in Fig. 3.8 may actually be active and Eq. 3.5 is appropriate to predict the ultimate temperature of the frame in fire. In fact, if the proportion of a frame is given, Eq. 3.5 can be solved with respect to T when assuming FðT Þ ¼ F* ðT Þ and N res ¼ N *res . If we interpret N(T ) as the pseud plastic compressive strength of post-buckling columns, we can say that Eq. 3.5 owes exactly to the simple plastic theory, since the quantities in the equation consist only of the plastic strengths of the key members and the external forces and they do not depend on any initial thermal strains. The above predicted ultimate temperature of a frame in fire is said to be one of the structural characteristics which takes structural overall redundancy into account. The above discussion indicates that the simple plastic theory can give simple means to evaluate this redundancy of a frame in fire. This may enable us to use the advantage of redundancy in practical fire resistance design of steel frames. Detailed discussion on the topic of this section is found in authors’ paper which focuses especially on fire resistance of seismically designed steel frames [8].

3.3

Post Earthquake Fire Resistances of Steel Frames in Fire [13]

We now treat post-earthquake fire resistance taking ordinary multi story office buildings in steel as examples. The following fire response analysis for these frames brings us not

25

only somewhat serious problems for post-earthquake technologies but an important general finding also. These are not independent of redundancy of frames which is discussed in the previous sections. Samples analyzed have all five stories and ten spans, which are steel plane frames with story height of 4 m and span length 7.2 m and designed seismically. Each of eight interior columns of the base floor carries initially vertical load as large as 0.3 of its own yield axial strength. Its normalized slenderness ratio is 0.33. It is assumed that they are subjected to earthquake damage of various permanent story drifts in advance before they undergo fire, in which the provided permanent drift angle ranges from 0 to 0.04 radians. Post-earthquake fire is assumed to occur inside a continuous space of the base floor over one or consecutive several spans of beams. Bearing in mind that the compartments may be damaged by pre-loading of earthquakes, fire over the entire floor is also considered as the most severe case. Before the second numerical fire response analysis, permanent story drift of the base floor is provided as earthquake damage, which is obtained also from the first numerical structural analysis at ambient temperature. Permanent story drift is always set to the right direction in the first analysis as shown later in Fig. 3.12. Under this pre-damage, all columns and beams facing the fire on the base floor are assumed to be heated uniformly in the second analysis. Total of 19 types of fire are considered, i.e. fires over the rightmost span, right two spans, . . . right nine spans and all spans as well as fires over the leftmost span, left two spans, . . . and left nine spans, respectively. This means that we are going to study also whether fire enlargement direction influences the post-earthquake fire resistance of the rightward pre-drifted frame. All results analyzed are summarized in Fig. 3.11. All heated interior columns have the same theoretical buckling temperature 614  C under their nominal axial forces, since their geometrical, loading and heating conditions are also identical respectively and this temperature is shown by a broken line in the same Fig. 3.11. Each connected dots in the figure exhibit how the ultimate temperature of the frame changes according to different magnitude of permanent pre-drift when it is subjected to an identical enlargement of fire. Analysis results can be classified into two groups in view of mode of collapse, i.e. sinking mode and side sway modes of collapse of the frame. Both modes are shown in Fig. 3.12. The sinking mode can be identified from the observation in Fig. 3.12a that, in this case, not only the individual heated columns exhibit bent buckled mode but the displaced roof level of the frame also sinks locally from the original horizontal line. All frames whose ultimate temperatures are greater than the theoretical 614  C and are close to 625  C exhibit sinking mode ultimately. This higher temperature tells us the fact

26

H. Suzuki

ultimate temperature in ⬚C

700 theoretical buckling temperature

600

fire over right 1 span right 7 spans right 8 spans right 9 spans all spans

500

0 0.01 0.02 0.03 0.04 permanent story pre-drift angle

ultimate temperature in ⬚C

700 theoretical buckling temperature fire over left 1 span left 8 spans 600 left 9 spans all spans

500

0 0.01 0.02 0.03 0.04 permanent story pre-drift angle

Fig. 3.11 Ultimate temperatures of a seismically damaged steel frame with various permanent story drifts to the right direction on the lowest floor (First (a) is for the cases fire spreads right to left, while second (b) left to right)

a

b

sinking mode of collapse

sway mode of collapse Fig. 3.12 Two typical modes of collapse of seismically pre-damaged frames in fire

that post-buckling lifting action of the whole frame discussed in the previous section is still active for these pre-damaged frames. This is because 625  C is almost

equal to the ultimate temperature of the intact frame without pre-drift under the same enlarged fire and such lifting action is always active for intact frames in fire with sinking mode as discussed previously. This fact holds even if these frames have suffered severe earthquake damage with pre-drift of 1/50 as shown in the figure. The reason why sinking mode is still active for largely pre-damaged frames is because, since the fire in these cases expands to somewhat limited space on the base floor, stiffness and strength of the remaining cooler columns on this floor overcome unstable P  Δ effect and prevent the frames from side sway movement, even if the pre-drift is going to accelerate P  Δ movement. However for the fires over consecutive right eight spans, right nine spans and all spans, i.e. when fire expands to very large or almost whole space on the floor, the frame collapses with side sway mode finally. The ultimate temperatures for these cases become lower than the theoretical one as shown in Fig. 3.11a. A typical mode of this failure is shown in Fig. 3.12b. It is important that the ultimate temperatures for these cases become significantly lower as the pre-drifts become larger. The lower ultimate temperature in these cases are caused by the mechanical weakness that, since most of the columns on the floor in fire are heated and there are few or no stiff elements which restrain the frame from unstable side sway movement, existing rightward pre-drift and horizontal lengthening of the heated beams to the same direction accelerate P  Δ effect. The cases of the fire over consecutive left nine spans seem meaningful. The undamaged frame under this fire is, of course, identical to the undamaged frame under right nine spans-fire and the frame in this case exhibits side sway instability. An interesting point is that, as shown in Fig. 3.11b, this fire brings more stability to the damaged frame with pre-drift angle of about 1/200–1/75 than to the undamaged frame. In fact, both the collapse mode and the ultimate temperature of these pre-damaged cases are those of sinking mode. The elongation of heated beams, in these cases, makes the heads of heated columns move leftward, since the stiff cooler right end column restrains the heads from moving rightward. This leftward movement cancels the unstable rightward sway due to combined rightward pre-drift and P–Δ effect. This is the reason sinking collapse takes place and the higher ultimate temperatures are attained for these cases. The cases of fire over consecutive left eight spans in Fig. 3.11b and those of right eight spans in Fig. 3.11a have different results from each other. Both the ultimate state and the ultimate temperatures for the former cases are those of sinking mode, while the latter of sway mode. The reason why so is similar to that of the above left-nine-spansfire cases in that leftward movement of the column heads cancels unstable rightward P  Δ movement for the former cases.

3

Ultimate Strength and Its Application to Post-Earthquake Fire Resistance of Steel. . .

It is now clear that, if fire is limited within a local or not so enlarged space on the floor or if any stiff structural element to prevent a frame from side sway movement exists on the floor in fire, fire resistance of a frame is not deteriorated in spite of preceding earthquake damage of permanent story drift. The ultimate temperatures of frames for these cases are not less than the theoretical buckling temperatures of the heated columns under nominal axial forces. Otherwise the ultimate temperature of a frame lowers depending on the magnitude of story pre-drift it has suffered. Lower ultimate temperatures in these cases are caused by unstable side sway of frames due to combined pre-drift damage and P  Δ effects. Detailed discussion for the latter cases is found elsewhere [9–13].

27

light steel framing

Top board Base board

3.4

Post-Earthquake Fire Resistances of Damaged Dry Partition Walls

Earthquake damage of permanent story drift may give some ill influences on not only structural but insulation performances of so damaged steel frames in fire. Since gypsum boards are widely used for fire resisting partition walls of buildings today, gypsum boards must first be picked up no matter what topic may be discussed on fire resistance of walls. In view of this, setting targets for dry partition walls of gypsum boards, their post-earthquake fire resistance is discussed herein. Such partition walls fitted in major members of a frame may be forced to displace as much as the members do when the frame is subjected to earthquakes. Therefore if the frame is then damaged in permanent story drift, such walls may also suffer identical in plane permanent shear deformation at the same time. Bearing in mind large permanent story drift angle 1/50 of frames after severe earthquakes, do identically damaged such walls remain resistible against subsequent fire? However these problems have never been solved yet. To answer this question the following experimental study has been conducted [14]. A 2.7  3.0 m gypsum board partition walls are used to experiment, which are constructed under the condition satisfying domestically licensed prescription. The wall is of one surface and two-ply system with light steel framing as shown in Fig. 3.13. Both top and base layers of the system are fire resistant gypsum boards of a thickness 21 mm. Both are attached with staples to give mutual adhesion and unify them. Way to construct the specimen never varies from the prescription. This wall specimen S2 is first fitted in the steel supporting frame without clearance and is subjected to alternately increasing repeated in plane shear deformation up to the angular amplitude of 1/50 which is provided by the loading

Fig. 3.13 Two-ply dry partition wall system constructed of light steel framing and top and base gypsum boards which are attached by staples

Fig. 3.14 Loading apparatus to give in plane shear deformation to walls

apparatus shown in Fig. 3.14. This test is done at ambient temperature and simulates seismic excitation. The test result of in plane shear force vs. according shear deformation of specimen S2 is shown in Fig. 3.16a, where nonlinear historical response of the specimen is observed and is caused mainly by in plane rocking motion of the top board relative to the base board of the plied wall. Sketch of rocking motion is shown in Fig. 3.17a, which is in plane rigid body rotation of each board resulting from local board to board prying action near the corner joints of boards. Mechanical prying action takes place and helps the ply system carry in plane shear force to compensate lack in interlayer shear strength between the top and base boards.

28

Fig. 3.15 A wall specimen set in a farness

This means that, under this in plane shear loading, the interlayer shear strength owing to staples is not large enough to unify the top and base boards and therefore the staples are necessarily pulled out to some extent. This results in above prying action and according rocking motion of the boards. The above loading test is terminated, decreasing and holding the permanent shear deformation of the specimen of 1/100 in angle. Setting so damaged specimen S2 in a furnace for full scale walls shown in Fig. 3.15, heating test is next conducted under the heating program in accordance with ISO 834 standard curve. The result of this test of averaged back side temperature response of the base board is shown in Fig. 3.18. The fire resistance duration indicating time length from the start of heating up to the ultimate state of this specimen is 50 min. The ultimate state is defined to be the state when the maximum of the distributing back side temperature attains 180  C. It is observed during heating test for S2 that a part of the top boards separates from the base boards and drops down when 22 min has passed after the start of heating. This accelerates increase of temperature of the base boards as shown in Fig. 3.18 and weakens the fire resistance of specimen S2. To know how the above separation weakens the fire resistance of walls, a bench mark test is also conducted as follows. Another specimen S1, which is a duplicate of S2, is subjected only to the same heating test without loading any pre-damage. The test result of S1 is shown also in Fig. 3.18. Any drop out of the top board has never been observed throughout the test of S1 and the resulting fire resistance duration of S1 is found to be 93 min. Obtained 93 min may be the uppermost of the fire resistance of identical walls, since this time length is seemed to be determined solely by

H. Suzuki

thermal conduction of an intact wall. Therefore any weakness may shorten the duration, if it exists or it is encountered during heating. Comparing the result of S1 and S2 tells us an important fact that such earthquake damage as specimen S2 suffers weakens fire resistance of these types of walls inevitably and significantly and the presently accepted public prescription to construct such walls may be insufficient in view of postearthquake fire resistance. In the following we try and examine improved construction of such walls in order to make them sufficiently resistible against post-earthquake fire. It does not seem difficult when we understand the cause why the performance of S2 weakens much. The main cause is separation of the top boards from the base boards which results from complete pull out of staples. As seen previously for S2 case, the pull out strength of staples is not large, so that the staples have been pulled out to some extent already under shear force loading before heating and this therefore results in rocking motion of the boards. Separation of the top boards during subsequent fire is caused by this weakness itself, because the staples are pulled out also during heating due to out of plane warping deformation of the unequally heated top boards. In view of this, an easy and practical improvement for the problem seems to hit more staples densely into boards and/or to make nails of staples longer to increase pull out strength. Improved specimens S3, S4, S5 and S6 are then constructed respectively and they all have so strengthened pull out characteristics. The resulting pull out strength of staples per unit board area is found in Fig. 3.19 for each specimen, which is obtained from pull out test for each case. Note that strengthening of S3 is small while S6 large. For improved specimens S3, 4, 5 and 6, shear loading test and subsequent heating test are then conducted individually and similarly as before. Relevant test results are found in Figs. 3.16b, 3.17b, 3.18, and 3.19. The shear loading test says that increase in pull out strength of staples makes the specimen stronger against shear force as shown in Fig. 3.16b for S6 case. This strengthening brings less premature rocking motion but more stable in plane shear deformation of the board system during shear force loading. In fact, slightly reduced rocking deformation is found for specimen S3, and more reduced ones for S4 and 5. Especially for specimen S6, rocking motion disappears and remarkable diagonal cracks develop instead due to increased shear force at later stages of loading as shown in Fig. 3.17b. Increase in pull out strength is reflected clearly in improvement of fire resistance of the walls, which is summarized in Fig. 3.19. In fact the following observation is obtained for subsequent heating test of the improved specimens. For specimen S3, the top boards separate and drop down during heating as for S2. But the separation time

Ultimate Strength and Its Application to Post-Earthquake Fire Resistance of Steel. . .

a

P(kN)

Fig. 3.16 (a) (left) in plane shear force vs. shear deformation relations of specimen S2 (right) (b) that of specimen S6

29

b

80 60

60

40

40

20

20

0 -20

-20 -40

-60

-60 20

40

60

-80 -80 -60 -40 -20 0 20 δ(mm)

80

40

60

80

b

a

Fig. 3.17 (a) (left) failure mode of gypsum boards of specimen S2 after shear force loading test (b) (right) that of specimen S6

250

100 S-2(ave)

200

S1, 4, 5 and 6

S-3(ave)

S2

S-4(ave)

150

S3

S-5(ave) S-6(ave)

100 50

ulmate state 0

10

20

30

40

50 60 70 Time(min)

80

90 100 110 120

Fig. 3.18 Time histories of back side averaged temperatures of all damaged wall specimens up to ultimate states including that of intact specimen S1

fire resistance duration [min.]

S-1(ave)

Temperature(⬚C)

0

-40 -80 -80 -60 -40 -20 0 δ(mm)

0

80

P(kN)

3

S1 e

80

S4 S3

60

S5

S6

f

d

S2 c

40

a b

: time top board is separated from base one

20

0

0

1000

2000

pull out strength of staples Fig. 3.19 Fire resistance durations of all tested pre-damaged wall specimens represented in terms of pull out strength of staples

in this case is lengthened to 31 min. For S4, 5 and 6, top board separation is not encountered throughout the heating test. The fire resistances of all these improved specimens become almost equal to that of intact S1. Therefore cracks and small opening of board to board joint lines may have no significant ill effect on post-earthquake fire resistance of walls. Slight reductions of the fire resistance durations of S4, 5 and 6 below that of S1 may probably be their effect. Figure 3.19 shows fire resistance duration (denoted by R) vs. pull out strength (denoted by P) relations of the tested specimens. In the figure the hollow triangles mark the state

of top board separation of S2 and S3 and the hollow circles the ultimate state of S2, 3, 4, 5 and 6. A broken horizontal line represents the uppermost of resistance duration of S1. A single x marked point indicates a predicted duration when back side of the same one-ply gypsum board of thickness 21 mm attains 180  C under the same heating as the test. Sketched curves in the figure help us interpret nature of the problem. Line 0a may represent staple pull out time vs. P relation. The curve smoothly extrapolated to point b from curve dc may represent R ~ P relation for the walls whose

30

top boards separate during fire, while curve ef for the walls whose two-ply system holds without layer separation throughout. Curves 0a and bd may not be smoothly connected to curve ef. The gypsum board used would yield under a certain pull out force during heating. Since yield strength of a gypsum board may decrease as board temperature increases, top board separation may not take place throughout the heating if their ply system has a certain pull out strength or more. Vertical line segment ae indicating discontinuous nature represents this anticipation.

References 1. Suzuki H (1995) Ultimate temperatures of steel frames subject to fire. J Struct Constr Eng Archit Inst Jpn (477):147–156 (in Japanese) 2. Architectural Institute of Japan (2008) Recommendation for fire resistant design of steel structures (in Japanese) 3. Ryan JV, Robertson AF (1959) Proposed criteria for defining load failure of beams, floors and roof construction during fire test. J Res Natl Bur Stand C Eng Instrum 63C(2):121–124 4. Suzuki H, Ruangtananurak N, Hujita H (2003) Stabilities of steel frames subjected to fire. J Struct Constr Eng Archit Inst Jpn (571):161–168 (in Japanese) 5. Suzuki H, Ruangtananurak N, Hujita H (2003) Overall stabilities of locally heated steel frames under fire. Int J Steel Struct 3 (3):227–233

H. Suzuki 6. Suzuki H (2005) Future work on fire resistance of building structural elements, The 21st century COE program, TUS 2nd international symposium, Tokyo University of Science, March, pp 137–146 7. Japanese society of steel construction (2005) Guidelines for collapse control design – II research 8. Suzuki J, Abe S, Suzuki H, Ohmiya Y, Wakamatsu T (2006) Ultimate temperature and structural redundancy of steel frames exposed to fire ~ effect of seismic design on fire resistance~. J Struct Constr Eng Archit Inst Jpn (608):157–164 (in Japanese) 9. Kondo S, Ikeda K, Suzuki H (2008) Reduction in ultimate temperature due to residual slope by relative story displacement after earthquake. J Struct Constr Eng Archit Inst Jpn 73 (630):1369–1376 (in Japanese) 10. Kondo S, Miyauchi T, Ikeda K, Suzuki H (2009) Reduction in ultimate temperature due to fracture at the end of the girder after earthquake. J Struct Constr Eng Archit Inst Jpn 74(638):385–392 (in Japanese) 11. Kondo S, Miyauchi T, Ohguma K, Ohmiya Y, Ikeda K, Suzuki H (2009) Post-earthquake ultimate temperatures of multi-story and multi-span frames subjected to whole floor fire. J Struct Constr Eng Archit Inst Jpn 74(645):2103–2109 (in Japanese) 12. Kondo S, Ohguma K, Miyauchi T, Ikeda K, Suzuki H (2009) Structural stability of frames damaged by earthquake at fire. Int J Fire Sci Technol Tokyo Univ Sci 28(1):33–50 13. Kondo S (2010) Post-earthquake ultimate temperature of steel frames in fire. Doctoral thesis, University of Tsukuba (in Japanese) 14. Ichihara T, Suzuki J, Suzuki H et al (2010) Experimental study on fire resistance of damaged partition walls – Part 1, 2, 3. Summaries of technical papers of annual meeting, Toyama, AIJ, pp 145–151 (in Japanese)

4

Integrating Wildland and Urban Fire Risks in Local Development Strategies in Indonesia Yulianto Sulistyo Nugroho

Abstract

This paper examines wildland and urban fires situations in Indonesia. On one hand, during the dry season when wetland dry out, large scale wildland/forest fires may occur on dry land and also on wetland such as peatland or peat-forest. Wildland fires, especially when involving peat land fires can be viewed as a regional and global disaster. Smouldering combustion of peatland fires produces thick haze that directly impact the life of the local peoples and can spread to other region and neighbouring countries. On the other hand, as consequences of economic development, rapid urbanization occurs. It is expected that larger portions of the populations will move to urban areas result in higher densities of people and complex highrise buildings and lines of utility infrastructures. These new build environments are prone with urban fires problems. Risk reduction efforts in wildland and urban areas becomes more significant than ever. Better understanding of fire phenomena in urban and wildland areas, and the needs of fire hazards response management, leads the efforts in integrating wildland and urban fire risks in national and local development strategies in Indonesia. # Springer 2015. Keywords

Wildland fire  Urban fire  Fire risk  Risk reduction  Development strategies

4.1

Introduction

Indonesia is one of the most disaster prone countries in the world. The country faces multiple hazards such as earthquake, tsunami, volcanic eruption, flood, landslide, drought, and forest fires. Of the numerous disaster that have happened in Indonesia during the past 30 years (1982–2012) there are 10,817 disaster events. The most frequently occurring is flood (4121 events), followed by landslide (1983 events), strong wind (1903 events), drought (1414 events), and other disaster (1397 events). During this period, the disasters have

Y.S. Nugroho (*) Department of Mechanical Engineering, Universitas Indonesia, Kampus UI, Depok 16424, Indonesia e-mail: [email protected]

claimed 225,509 lives, including the great earthquake and tsunami in Aceh in 2004 [1]. The increase in population and the limited paddy fields in Java, prompted the government in 1990s to convert peat and lowland swamp to rice cultivation and nature areas. However, the construction of canals often cut through the centre of peat domes resulting in excessive drainage, subsidence, irreversible drying, loss of habitat and increased risk, frequency and severity of fire [2, 3]. Some of these peatlands have become degraded with negative environmental and socioeconomic consequences [3]. Despite the growth in mining, manufacturing, agriculture and services sectors, forestry are very important to the Indonesian economy. However, practices that include forest clearing are connected to commercial activities, such as slashing leave behind dry fuel loads that are more susceptible to wildfire. These fires also cause noticeable secondary

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_4

31

32

Y.S. Nugroho

risks such as smoke, which contributes to local and regional degraded health and environmental conditions [4]. At the present time, fires remains the greatest environmental problems faced by Indonesia. Forest fires have destroyed million hectares of forest and land which cause economic loss, social problems including smoke related diseases and environmental disaster with long time consequences. Following long dry season, forest and peatland fires triggers local and trans-boundary haze problems due to their impact has spread to neighbouring countries and the gases they release into the atmosphere (such as CO2) are potential causes of global warming [4]. Besides facing the large scale wildland fires, as consequences of Indonesian economic developments in the national, provincial and local levels, rapid urbanization occurs [5]. Urbanized areas are typically have large populations, and with higher densities of people and buildings. Today, as more than half of the world’s population lives in urban areas, and coupling with the impacts of climate change, risk reduction in urban areas becomes more significant than ever. The increasing number and impact of natural disasters reveal themselves in statistics. The unprecedented rise in the number of natural disasters exposes a need to recognize global trends influencing this rise, and confront them through a larger policy framework. This paper presents the evolution of fire phenomena understanding, fire hazards response management and efforts in integrating wildland and urban fire risks in national and local development strategies in Indonesia.

4.2

Fire Situations in Indonesia

4.2.1

Wildland Fires

Left in their natural state, wildland fires in tropical forests would rarely occur. Nevertheless, many human activities in tropical forests are linked to wildfires. Opening forests for settlement, agriculture, plantation, or logging, can alter the humidity or moisture from the soil due to evaporation. During a long dry seasons, a much drier condition of the local ambient atmosphere close to the soil are expected. The sites that are filled with high amount of slash abandoned biomass are more susceptible to fire. The potential for fire differ from place to place depending on the nature of the fuels, climatic pattern, and sources of ignition [6]. Forest fires in Indonesia is one of the biggest environmental problems due to periodic fires in a large scale. As a result of these fires is the potential loss of timber, biodiversity and ecosystems. Annual average economic losses from forest and other land fires in the last decade or more, and how these have affected national gross product, are not fully

known. The 1997–1998 fires in Indonesia caused losses estimated at US$8.7–9.6 billion [7]. Forest fires usually occur due to ignition where the rate of heat released from the combustion process is greater than the rate of heat that is released into the environment. Ignition begins with the heat sources such as fire, lightning and so on. Once the flammable mixture of gases reaches its combustion temperature the process will continue despite the heat source has been removed. Records of wildland fires in Indonesia shows that fires occur not only on dry land but also on wetland such as peatland/peat-forest, particularly during the dry season when the wetlands dry out. Large scale clearing of peatlands with the digging of canals has further increased the risk of fire breaking out during the dry season, as the groundwater drains away through the canals leaving the peat excessively dry and easily combustible. As a result peat is no longer capable of absorbing nutrients or retaining water [8]. Fires in peatlands are dominated by smouldering combustion which is the self-sustained, slow, low temperature, flameless form of burning [9]. Excessive smoldering combustion of peatlands could endanger the environment as tropical peatlands have an important role in the ecosystem in many parts of the world. Early studies suggest that peatlands are largely (a) a major stock of carbon [10]; (b) enriched with biodiversity of various species [11]; (c) a storage of excess rainfall [12], and; (iv) a source of timber and non-timber forest products for local communities [11]. The large quantity of fuel burned on the surface of the land will produce thick smoke and extensive environmental damage. Smoke from peat and forest fires can be attributed to transboundary haze pollution. The haze from these fires affected visibility in some parts of Sumatra, Kalimantan and also the neighbouring countries in the region. It affected sea, land and air transportation due to low visibility caused by forest fires. The fires may also have led to a decrease in tourism, with potential visitors being concerned about both the effects of smoke haze on their health, and the safety of transportation. Fires also resulted in losses or destruction of natural habitat and forest biodiversity [6]. Given the vast nature of the geographical area of tropical forest and peatlands in Indonesia, satellite techniques are the primary means for wildland fire monitoring system. Along with on-ground and air monitoring, the use of satellite methods and products facilitate the expansion of the area capable of being monitored [4]. This increases both the efficiency of wildfire detection and the capacity to obtain detail information on wildfire, such as co-ordinates estimation of active burning area, total damaged areas, and temporal characteristics of fires. According to the official data recorded by Ministry of Forestry of Republic of Indonesia, i.e. Direktorat Pengendalian Kebakaran Hutan, Direktorat

4

Integrating Wildland and Urban Fire Risks in Local Development Strategies in Indonesia

Jenderal Perlindungan Hutan dan Konservasi Alam Kementerian Kehutanan, website: http://ditpkh-phka. dephut.go.id/, during the period of 2010–2014, there were about 20,000 hotspots events annually (Fig. 4.1) [13]. The hotspots in Indonesia mostly occurred between February and April, followed by July to September. Figures 4.2 and 4.3 suggest that fire seasons are related with dry seasons and also affected by hydrometeorogical conditions. The critical moisture content for initiating smoldering of various boreal peat has been measured in the range 40–150 % in dry basis [15, 16]. Drier than this threshold, peat becomes susceptible to smoldering. Thus, drainage, for example for oil palm plantations, drains off water from adjoining forested areas, and the general water table begins to fall. Whether accidental or deliberate, fires on peat can easily burn out of control, especially in periods of drought years. Because the fire spreads deep into the soil such fires can be hard to extinguish. Nevertheless, these fires are mostly caused by human factors [17]. Figure 4.3 shows that during the long dry seasons in September 2014, hotspots in peatlands are observed in Riau, Jambi, South Sumatra, West and Central Kalimantan (Fig. 4.4). Fire danger is the probability of a fire to start, the rate of spread and intensity of its burn. This probability is influenced by fuel type, fuel moisture, amount of fuel, and slope of the land area. Another important influence is weather, particularly wind and relative humidity. According to fuel moisture code characteristics, the mapping of weather-based Fire Danger Rating Systems (FDRS) – BMKG for ASEAN (Association of Southeast Asian Nations) regions on March, 19, 2015 was shown in Fig. 4.4 [18]. Most of the northern part of ASEAN region, i.e. Myanmar, Lao, Kambodia, the Philippines, Thailand, Malaysia and Northern Sumatra region were on red-alert that could be interpreted as grass fuels highly flammable, very high probability of fire starts. This condition was confirmed by the satellite image of hotspots and haze images (Fig. 4.5) published by the National Environment Agency of Fig. 4.1 Number of hotspots occurance through out the year [13]

33

Singapore [19]. The Southeast Asia Fire Danger Rating and the regional haze images are now available on national agencies for environment and disaster management in every ASEAN member countries.

4.2.2

Urban Fires

Indonesia is an archipelago consisting of 17,508 Islands. In 2014, Indonesia’s population is estimated about 250 million people. Around 57 % of Indonesia’s population live on the Island of Jawa, the fifth largest Island in the archipelago [5]. Figure 4.6 shows the distribution of Indonesian population according to the main Islands and regions (a) as well as the reported fire incidents (b). Although more than half of the Indonesian people live in Jawa Island, it contributes about a quarter of the total fire incident reported. The Islands of Sumatra and Kalimantan where most of the palm plantations are situated contribute almost half (44 %) of the fire incidents. Fire incidents in Jawa, and big cities across Indonesia are characterised as urban fire, i.e. fire hazard that involves areas where single family homes, multi-family occupancies and/or business facilities are clustered close together, increasing possibility of rapid spread to other structure. Most big cities of Indonesia are either located on seacoasts or directly linked with riverbeds. Potential hazards in coastal areas and cities built near rivers are coastal flooding, erosion of beaches, sedimentation in river floors, flooding, and landslides. In addition, remarkable changes in productivity and lifestyle have occurred as the Indonesian economy has grown. There are now expanding sites of settlements with dense population and complex high rise buildings, new industrial sites with locations for utilities, storage and production. As consequence, the fire risks of these new sites have increased significantly. Statistical data on urban fires occurance for the cities of Jakarta and Surabaya are presented in Fig. 4.7. Despite the

34

Y.S. Nugroho

Fig. 4.2 Rainfall forecast for Indonesia in August 2014 [14]

Fig. 4.3 Hotspots mapping in Sumatra, Kalimantan and Sulawesi, Indonesia on 12 September 2014 [14]

economic growth and development of new buildings, the number of fire incidents for the period of 2002–2008 were kept within a relatively constant value of about 370 incidents per annum for Surabaya and 860 incidents per annum for Jakarta. It is found that most fire incidents occurred in residential buildings. Similar to those of wildland fires, Fig. 4.8 suggests that urban fire incidents are peaking during the dry season in the months of July to September.

Due to poor installations and maintenance, substandard electrical products and ignorance of the users, as shown in Table 4.1, electricity related fires are still on the top causes of fire incidents. A study by Heru (2011) [21] suggested that the casualties and direct loss of urban fires in Jakarta and Surabaya are relatively low. Nevertheless, there is a latent problem related to a long emergency response time of fire brigade due to heavy traffic congestion and access difficulties.

4

Integrating Wildland and Urban Fire Risks in Local Development Strategies in Indonesia

Fig. 4.4 Characteristics of weather-based Fire Danger Rating Systems (FDRS) – BMKG for ASEAN regions on March, 19, 2015., website: http:// geospasial.bnpb.go.id/ monitoring/hotspot/ (Accessed on 20 March 2015) [18]

Berlaku untuk: 19 Maret 2015 90 E 30 N

100 E

110 E

35

Wilayah Association of Southeast Asian Nations 120 E

130 E

140 E 500

150 E 30 N

0 500 Km. N

BMKG Keterangan: 20 N

20 N Aman Tidak Mudah Mudah Sangat Mudah

10 N

10 N

EQ

10 S 90 E

EQ

100 E

110 E

120 E

130 E

140 E

10 S 150 E

Fig. 4.5 Regional haze map of Southeast Asia on March, 20, 2015. National Environment Agency of Singapore, website: http://www.haze.gov. sg/hotspot-satellite-images (Accessed on 21 March 2015) [19]

36

Y.S. Nugroho

a

Papua and Maluku 3%

Sumatra 21% Number of Incidents

Sulawesi Kalimantan 7% 6% Bali and Nusa Tenggara 6%

80

Surabaya

70

Jakarta

60 50 40 30 20 10 0

Jawa 57% Total populaon: 250 million (2014)

b Sulawesi 13%

Papua and Maluku 9%

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Fig. 4.8 Occurance of fire incidents through out the year [21]

Sumatra 32%

Kalimantan 12% Bali and Jawa Nusa 26% Tenggara 8% Total fire incidents : 44161 incidents, (2013)

Fig. 4.6 Number of population in the main islands of Indonesia (a), and reported fire incidents in 2013 (b) [20]

Number of incidents

Surabaya 1000 900 800 700 600 500 400 300 200 100 0

Jakarta

2002

2003

2004

2005 2006 Year

2007

2008

Fig. 4.7 Number of urban fire incidents in Jakarta and Surabaya, Indonesia (2002–2008) [21]

Improvement of fire situations in big cities like Jakarta, should be addressed in line with the development of basic physical and social infrastructure. In 1969, a governmentassisted self-help community planning was launched by Mr. H. Ali Sadikin, the Governor of Jakarta. The so called Kampung Improvement Program (KIP) components included the development of (i) access roads, bridges and footpaths, (ii) water supply and sanitation, public tap, drainage canals, and community sewage disposal system; (iii) social buildings, schools, health clinics and mosques. The design of the KIP was largely influenced by the need for an inexpensive method of rapidly providing basic infrastructure, using minimum of technical and administrative resources. The impact of the KIP program was significant in reducing problems of sanitation, flood, and health amongst millions of ordinary people who live in Jakarta. The KIP program was awarded the Aga Khan Award for Architecture in 1980 [http://www.akdn.org/architecture/pdf/ 0001_Ind.pdf] [22]. Figure 4.9 shows a photograph of a typical of kampung in Jakarta being improved by the KIP program. The access to the neighborhood was improved by creating a 120 cm width concrete footpaths and a 50 cm width drainage canals on both sides. Thus, it provided almost 3 m gap between the adjacent buildings. The replacement of wooden buildings with more fire resistant buildings were other factors that contributed to the reduction of urban fires. However, during the last 15 years, the fire situation worsen due to the increase demand of space in the individual houses. The old houses were replaced by two-storey accommodations. A numerical study by the author and co-workers as appeared in Fig. 4.10 suggested that a 3 m gap between a single story houses is sufficient to prevent fire spread (Fig. 4.9). However, for two-story houses a gap of at least 4.5 m was required to prevent fire spread. Thus, substantial

4

Integrating Wildland and Urban Fire Risks in Local Development Strategies in Indonesia

37

Table 4.1 Fire causes [21] Year 2002 2003 2004 2005 2006 2007 2008

Surabaya Stove 107 25 60 57 176 102 187

Light 12 15 4 10 23 8 6

Electricity 70 47 39 41 89 62 92

Cigar 9 1 13 6 0 1 1

Others 228 206 159 153 56 116 66

Jakarta Stove 89 82 83 68 92 94 93

Light 7 4 8 13 6 28 20

Electricity 397 463 456 458 519 469 475

Cigar 79 84 44 42 58 44 49

Others 297 255 214 161 227 220 193

and interpreted variously in practice [24]. The new initiative, at the same time, transformed the way of living for many people who previously lived in landed houses, i.e. from the 2D to 3D life style. Issues of public behavior, life safety, transportations, fire service performance and earthquake potentials are becoming more important to be considered during design, construction and operational stages of the new towers.

Fig. 4.9 Concrete footpaths and drainage canals of a typical kampung site in Jakarta

densely built urban areas still remain at risk of fire spread in the many cities of Indonesia. Since 1980s the modern real estate style developments were dominating the growth of new settlements in Jakarta and other cities in Indonesia. Most of the houses were concrete structures with better fire safety performance, including wider accesses to the houses. Due to land scarcity, in late 2000s, the real estate sector was flourishing by the introduction of Rusunami (Prosperous Ownership Flats) development program, characterised by the building highrise flats of 8 to around 20 stories. This program is supported by a government policy known as the ‘a thousand tower development program’, which started in 2007 with guided by the Regulation of Ministry of Public Works No. 05/PRT/ 2007 on, the Technical guidelines for the construction of the high-rise building flat and Multi-stories Building Act No. 20/2011. Problems arose on fire safety aspects, since the related technical guidelines were not written in details

4.3

Regulatory Framework on Fire Safety and Disaster Management

4.3.1

Legislations

Over the years, Indonesia has instituted a number of legislative provisions that directly and indirectly address the problems of fire safety, occupational health and disaster management in various sectors. Figure 4.11 illustrated the Acts related to Fire safety and disaster management. These legislations provide the legal framework for preventing, suppressing, mitigating, recovering and monitoring efforts to promote occupational and public safety, to control fire risks and disaster to the environment and released of pollutants related to fire incidents. The first Act was on occupational safety introduced in 1970. It rules the work safety requirements in planning, making, transporting, trading, installing, storing and utilising of goods and technical products and production instruments in order to guarantee the safety of the product themselves, safety at work and public safety. In response to large large forest and land fires in early 1980s and mid 1990s, the Forestry Act No. 41/1999 was enacted in 1999. Later in 2004, the legislation was amended by Act No. 19/2004 to minimise overlapping with existing mining permits or agreements in forest areas until the expiration of the license or agreement concerned. According to the Forestry Act, everybody has a responsibility to protect the environment and forests. To provide guidance and direction in wildfire control activities, the Government established the Government Regulation No. 45/2004 on

38

Y.S. Nugroho

Fig. 4.10 Effect of house to house distance on fire spread in urban fire conditions [23]

Fig. 4.11 Fire safety and disaster management are legislated in a number of Acts

Act No. 18/2004 on Agricultural Crop Plantaons

Act No. 24/2007 on Disaster Management

Act No. 1/2009 on Aviaon Act No. 4/2009 on Mineral Mining and Coal

Act No. 28/2002 on Building

Act No. 30/2009 on Electricity

Act No. 41/1999 On Forestry Fire Safety and Disaster Management Act No 1/1970 on Occupaonal Safety

Forest Protection and the regulation of the Minister of Forestry No. P. 12/Menhut-II/2009 on wildfire control. These regulations directed that forest fires control activities should include prevention, suppression and post-fire activities including fact finding and investigation, identification, evaluation, rehabilitation and law enforcement. In line with the Local Government Act No. 32/2004 and its successor, Act No. 23/2014, forest fire control measures are implemented at National, and Local (Provincial and District) levels as well as at forest management unit. As wildland fires were also occurred in the plantation areas outside the forest, thus, to strengthen efforts to cut the number of fires in this sector, the Plantation Act No. 18/2004 was enacted. The Act prohibits the use of fire in preparing land for planting. It specifies that plantation companies are responsible for controlling fires on land that

Act No. 23/2014 on Local Government

they manage, and must have adequate emergency response facilities to suppress fires at the earliest opportunity. Although Indonesia has enacted a number of regulations and established various institutions to tackle forest and other wildfires, but periodic wildland fires and haze continues to be a problem, with significant environmental and health crises. Regional haze map of Southeast Asia in March, 2015 (Figs. 4.4 and 4.5) revealed that besides Indonesia, other Southeast Asia nations also faced in-boundary and trans-boundary haze problems caused by wildland fires in their home country as well as their neighboring countries. In this issues, the Association of Southeast Asia Nations (ASEAN) plays an important role in addressing the regional forest fire problems and particularly trans-boundary effects. Ratification of the Agreement on Trans-boundary Haze Pollution by its member countries reaffirms the importance of

4

Integrating Wildland and Urban Fire Risks in Local Development Strategies in Indonesia

mitigating and combating wildland fires in the region, in particular forest and peatland fires. As mentioned earlier, the trend of urban population continues to increase. In 2010, the percentage of urban population has reached 49.8 % [5]. The increase in the proportion of the population living in urban areas reflects the process of urbanization that also reflected by transformation of rural areas into urban areas with very dense population. A thorough building legislation is required to safeguard people from injury or illness caused by natural disaster, fire, structural failure and infection and to protect property from physical damage caused by structural failure as triggered by earthquake, landslide, flood and other hazards. In 2002, Building Act No. 28/2002 was enacted to provide foundation for building life cycle and management. According to the Act, a building shall mean physical result of a construction work attached to its location, part or all situated above and/or under the ground and/or water, which functions as a place for humans to perform their activities, either as a settlement or home, place to perform religious activities, business, social, cultural, and other special activities. In general, the Building Act regulates the building functions, requirements, construction, including the rights and obligations of the owner and user of the building. Building administration is held with the principle of utility, safety, balance, also the harmony between buildings with its environment, during building life cycle. In more detail, the administration of building life cycle shall cover technical planning, construction, utilization, conservation, and demolition. In every stage of the process, it is important to make sure the functionality and the fulfilment of reliability requirements which ensures safety, health, comfort, and ease of use. As the Building Act regulates the principle and normative matters, the provisions on its implementation are provided by Government Regulation (Peraturan Pemerintah) No. 36/2005 and/or other statutory regulations, such as Ministry of Public Works Regulation No. 29/PRT/M/2006 on Guidelines on Technical Requirements of Building, No. 30/PRT/M/2006 on Technical Guidelines for Facilities and Accessibility in Buildings and Environment, No. 06/PRT/M/2007 on General Guidelines for Building Management and Environment, No. 24/PRT/M/2007 on Technical Guidelines for Building Permits, No. 25/PRT/M/ 2007 on Certificate on Function Feasibility, No. 26/PRT/M/ 2007 on Expert Teams on Building, and standards, e.g. Standard Nasional Indonesia (the SNIs), including Local Government Regulation, by continuously considering the provisions of other laws related to the implementation of this Act. With respect to life safety and fire risks in building, some regulation and technical guidelines are already in place, such as Ministry of Public Works Regulation No. 25/PRT/M/

39

2008 on Technical Guidelines for Preparation of Master Plan for Fire Protection Systems, No. 26/PRT/M/2008 on Technical Requirements for Fire Protection Systems in Buildings and Environment, No. 20/PRT/M/2009 on Technical Guidelines for Fire Safety Management in Urban areas. It is important to note that efforts are made to involve the community in taking significant role during the process of the planning, development and utilization of buildings, not only for their own benefit, but also in the fulfillment of building requirements and in achieving orderly construction of buildings in general. Implementation of the Ministerial guidelines at the National level throughout the autonomous regions requires the completion of the Local (Provincial and District) legislative drafting. The commitment of stakeholders such as building owners, architects, engineers, contractors and community who build and use the buildings are critical. Jakarta area is an example of the areas that already have local regulations regarding the building, both in terms of life safety and fire protection as well as aspects of energy and water conservation in buildings. In Jakarta, before building permits were issued, the design of the building is going through a process of technical assessment by a team of building experts, concerning aspects of urban architecture, building structures and building installations including aspects of life safety and fire protection.

4.3.2

Fire and Disaster Management Institutions

Fire can start in a single-family home, and if not properly controlled and suppressed, the domestic fire can spread to other houses leading to urban fires. Depend upon the scale of fires, a disaster stage may be declared by the authority as fire rapidly spread to the surrounding, affecting larger area with large scale economic and social disruptions and losses due to the haze problems. In a wildland-urban interface (WUI) situations the opposite scenario may take place. The ignition event and fire growth may initially start in the wildland, and due to the terrain, wind direction and thermal degradation of vegetation through radiative and convective heat fluxes, the fireline and fire plume can spread to the neighboring urban settlements. These process range from the small-scale (~1 mm) at ignition point, to the large-scale (~100 km) transport of smoke. Wildland fire modelling, considered in its entirety, is a very challenging task [25]. In urban areas, one of the most common type of public fire protection is the fire department. Although there are variety in the institutional levels, the Local (Provincial, District and Cities) Government in Indonesia have established their fire department, fire station or fire unit.

40

Initially, the main purpose of fire institution is for fire fighting/fire suppression purposes. Today, their roles expand to cover fire prevention, public education and rescue [26]. In Indonesia, Jakarta Fire Department is the oldest fire service institution, established 1919 or 96 years ago [26]. Along with the development in Jakarta, currently Jakarta Fire Department has been established into five administrative regions. Each administrative region are supported by divisions of fire fighting operation, fire prevention and code enforcement, community empowerment. In addition, Jakarta Fire Department is supported by urban fire training center, and laboratory unit. The distribution of the sub-fire department in five administrative regions of Jakarta, can effectively shorten the response time, during fire emergency situations. Meanwhile, during non-emergency situations, all administrative, management and business units are responsible for fire management in their own jurisdictions, including activities to prevent, monitor, suppress and mitigate fires. The responsibility of Local (Provincial, District and Cities) Government to provide fire protection services is clearly reinforced in the Local Government Act No. 23/2014. Besides, strengthening local fire service institutions, they also required to complete local regulations with fire safety and protection provisions with reference to the Government Regulations and guidelines given in the corresponding Ministerial regulations. The local government who issues the new and extension permits must ensure that all administrative and technical requirements are fulfilled, and the management and business of the entities are responsible for fire safety and protection management in their jurisdiction. In accordance with the regulations, therefore, each company such as oil palm, forest plantation and mining companies must equipped with adequate facilities for responding to fire emergencies. For fire suppression operation, in protected forests and conservation areas, the Ministry of Forestry established fire brigade units called Manggala Agni in 2002, and continuously enhanced their capacity to better control of fires in the state forests. Local governments also established wildland fire brigade units to enable a rapid response and a better fire control. Thus, local government is responsible for fires in the rest of the areas outside the state forest, protected forests and conservation areas. Learning from the great earthquake and tsunami in Aceh in 2004, earthquakes in Yogyakarta (2006), Padang (2007), volcanic eruption, floods and haze problems, the Indonesian Government and the National parliament marked a shift of paradigm from a previously response-oriented disaster management to disaster risk reduction, by enacted the Disaster Management Act No.24/2007 [1, 6, 27]. Later, Indonesian Government passed a number of ancillary regulations and numerous rules and procedures that regulate the many

Y.S. Nugroho

aspects of disaster management in the country, with the involvement of all stakeholders. In 2008 the Government establishment of the National Agency for Disaster Management (known as Badan Nasional Penanggulangan Bencana, or BNPB) through the Presidential Decree No. 8/2008. BNPB is a ministerial level independent body that has the authority to coordinates and implements disaster management programs and activities. The establishment of BNPB strengthens disaster management efforts, which are also handled by the National SAR Body (Basarnas), the Indonesian Red Cross (PMI), Indonesian Armed Forces (TNI), and other volunteers [6, 27]. At the local levels the establishment of BNPB (and BPBD at the local level) constituted a significant progress in the field of disaster management as it signified the departure from an “ad-hoc” and responsive approach to disaster management. The BNPB is responsible for coordinating disaster management including those that associated with wildland fires. Only in the event of a disaster or potentially disaster is the command line activated, nationally led by the head of the BNPB or, in the case of sub-national coordination, the head of Local (Provincial and District) Agency for Disaster Management. In the case of disasters related to forest fires, this national coordination will be supported by relevant departments and agencies in addition to the Ministry of Forestry, Ministry of Environment and Local Governments, such as: Aeronautics and Space National Agency; Meteorology, Climatology and Geophysics Agency; Police Department and Armed Forces. In terms of the Disaster Management Act No. 24/2007, the BNPB can also be involved in hazard prevention, and not focus only on coordination when disaster happens.

4.4

Integrating Wildland and Urban Fire Risks in Local Development Strategies

According to Local Government Act No. 23/2014, Government affairs in Indonesia consists of absolute government affairs by Central Government, concurrent government affairs by Central and Local Government, and general government affairs. Concurrent administration affairs under the authority of Local Government shall consist of mandatory government affairs and optional government affairs. Mandatory Government Affairs relating to basic services shall include (a) education, (b) health, (c) public works and spatial planning, (d) housing and residential areas, and (e) peace, public order, and the protection of society, and social. The division of government affairs for the field of peace, public order and the protection of society includes peace and public order affairs, disaster management affairs and fire service affairs.

4

Integrating Wildland and Urban Fire Risks in Local Development Strategies in Indonesia

Indonesian national commitment for disaster management that focuses on disaster risk reduction has been institutionalized into regulations and budgeting. The country has made disaster management an integral part of its national development priorities. Internalizing process of disaster management and fire risks into national development planning should go on from the policy making process at national level to the implementations at the local levels. It is important to integrate and sustain the risks reduction rational for the subsequent development planning cycles through effective management, transparency and accountability that are focused on achieving benefit and changes in the long run. Improvement in disaster management governance in Indonesian is started with the regulatory framework, and then planning and budgeting system. With regards to Fire service, the implementation of role and authority distribution from Central Government to Local Government are illustrated in Act No. 23/2014 as follows: (a) the central government has the roles to standardise fire service infrastructure, fire-fighter competency and certification, and to implement fire information system, (b) the Provincial Governments have the spatial role in mapping the fire prone areas, districts or cities under their jurisdiction, and (c) the District/Municipal authorities have specific duties on fire prevention, control, suppression, and rescue operations; inspection of fire protection equipment; investigation of fire incidents; and community empowerment for fire prevention. Central Government in conducting the concurrent administration affairs is authorized to establish norms, standards, procedures, and criteria in relation to the implementation of Government Affairs; and implement the guidance and supervision of implementation of Government Affairs under the authority of the Local Government. At present, Indonesia has 34 provinces (provinsi), 413 districts (kabupaten), 98 cities (kota), 6982 sub-districts (kecamatan), and 80,714 villages (desa) [5]. Within their jurisdiction, the Local Governments (Provinces, Districts and Cities) have the authority to issue permits for building design/planning, construction and operation in housing and residential areas, industrial, agricultural areas, transportation, business and trade sectors, etc. As the technical capacity of local governments to implement their authority are generally unequal, thus norms, standards, procedures, and criteria in relation to the implementation of Government Affairs are critical. It is anticipated that the local governments may take the opportunity to enact many local regulations to increase revenue through exploitation of natural resources. But, in return, they may face many unanticipated social and environmental problems. The Local Government Act No. 23/2014 gives central government better legal control over the exploitation of local natural resources including mining and forestry sectors. Responsibility of the central government in the

41

forestry sector, includes forest planning, forest management, conservation of natural resources and ecosystems, education and training, counseling and community development in the field of forestry, watershed management (DAS), forestry supervision. Meanwhile, in the mining and coal sectors, the authorities comprises allocations of mining area as part of a national spatial plan, which consists of the area of mining, artisanal mining area and reserve area, as well as concession area. Nonetheless, limited authorities are transferred to the local government at Provincial level for example the issuance of permits for direct use of geothermal energy source. The moves to give the central government better legal control on mining and forestry sectors are practical example of how fire risks and disaster risks reduction are integrated into the local development strategies. With better infrastructure, spatial considerations and funding at national levels, potential social conflicts over resource access and management that can increase wildfire risk, forest degradation and fragmentation can be reduced considerably. As the majority of Indonesian citizens will live in urban areas, and take into account the regions can have multiple functions as residential, business, industrial, agricultural land, and plantations, it is crusial for local government to impose provisions to undertake environmental impact analysis before issuing permits for major projects. Environmental impact analysis must be implemented properly to a major construction project that can generate undesirable impact to its direct surrounding or in the opposite, be affected by geographical or geological conditions of the sites. For example, it is undesirable to development a new airport close to peatlands with persistent smoldering fire problems. Regarding the wildland-urban interface (WUI), one of the fundamental issues driving the destruction of homes in urban fire or wildfire events is the very limited consideration of potential wildland fire and ember exposures in building codes and standards [25]. Integrating fire risks into local development strategy should also include the arrangement of buildings and the environment that could across the administrative boundaries, the human dimension of fire as a component of agricultural production systems, and the implementation of community development for disaster preparedness. In the field of fire emergency, disaster management and disaster risk reduction, particularly better response time during emergency call, early warning mechanisms and disaster preparedness, there is a growing need to improve community resilience, city/district coordination and national capacity. In this regard, the utilization of advanced information technology with correct geospatial and demographic data will support the emergency operations, as well as the decisionmaking process in planning and land use management, natural resources, environment, transportation, urban facilities, and other public services.

42

Y.S. Nugroho

In urban planning, emergency planning and measures, it is important to include the requirements for water supply necessary for extinguishing fires in residential, commercial and mixed-occupancy districts. For optimizing fire-fighting and rescue forces, it is necessary to establish the placement of urban fire stations and fire and rescue services based on the level of urban fire risk [28]. In addition, better understanding on human behavior in emergency situations [29] and the application of modelling tools for predicting smoke behavior in compartment fires [30] are urgently needed. Central and local government must work closely on measures to improve institutional capacity and people’s awareness in the effort of reducing disaster risks and impacts. At the same time, to be effective, the legislative provisions need good law enforcement. This should be integrated with measures to prevent and mitigate fires. The police department and prosecutor’s office play important roles in handling forest and land fire cases. The police have authority to investigate wildland and urban fire cases. Experiences from past incidents are important to improve the capacity and the preparedness of the parties involve in emergency situations. In the event following the Kobe Earthquake in 1995, fire spread could not be prevented by fire-fighting because the number of ignitions that occurred concurrently was greater than the number of fire engines. When the smooth operation of official fire services cannot be expected, due to the blockage of streets with collapsed buildings and structures and the cut of water supplies, firefighting activity by local residents, equipped with small fire pumps with firehoses is critical to minimize the damage and the risk of fire spread [31]. In the case of wildland and urban fires, every fire event is unique. Complex socioeconomic, ecological and governance factors are involved. It suggests that the solutions are multidimensional. In particular, the management of peatland fire hazard control must be based on the results of research such as predicting smoldering thresholds related to the critical moisture and inorganic contents for ignition [32], and developing canal the blocking system for a long-term rehabilitation measure [3]. Best practise of others may help, but should be adapted to the land conditions prevailing in Indonesia [8].

4.5

Concluding Remarks

Reducing fire risks in wildland areas including peatlands, and urban areas is challenging tasks that cover aspects of governance, socio-economic and ecological. Various efforts have been done to improve the existing regulations, to enforce laws, as well as through research and technical assistance and funding to local governments.

Taking into account the unique conditions attached to Indonesia as an archipelago, efforts to reduce the risk of fire needs to be integrated into development strategies with disaster risk reduction framework. Improved understanding of the phenomenon of fire, fire prevention and control still needs to be done especially for stakeholders in the conservation of forests and peatlands, as well as the development of the urban area. Through research and collaboration among local governments, at national and international levels are expected to increase the capacity and speed of the early warning system, fire prevention and fire management. At present tropical peatland fire management is still a big challenge in efforts to reduce pollution of smoke from wildland fires. Meanwhile, for urban areas, aspects of life safety, fire prevention and management, incident command for fires in tall buildings, deep basement, and residential fires are the crusial aspects in delivering safe and modern urban areas. Acknowledgments This work is supported by the Ministry of Research, Technology and Higher Education of Republic of Indonesia under National Strategic Research fund through DRPM UI. I would like to thank Ministry of Home Affairs, Ministry of Environment and Forestry Republic of Indonesia, Prof. Suprapto and Prof. Bambang Hero Saharjo for great discussion and their help.

References 1. Government of Indonesia National Agency for Disaster Management (2013) Executive summary National Assessment report on Disaster Risk Reduction 2013 Redefining Indonesian disaster management strategy, Directorate of Disaster Risk Reduction National Agency for Disaster Management, Jakarta 2. Page SE, Siegert F, Rieley JO, Boehm HDV, Jaya A, Limin S (2002) The amount of carbon released from peat and forest fires in Indonesia during 1997. Nature 420(6911):61–65 3. Ritzema H, Limin S, Kusin K, Jauhiainen J, Wo¨sten H (2014) Canal blocking strategies for hydrological restoration of degraded tropical peatlands in Central Kalimantan, Indonesia. Catena 114:11–20 4. Sagala S, Sitinjak E, Dodon Y (2015) Chapter 7: Fostering community participation to wildfire: experiences from Indonesia. In: Shroder JF, Paton D, Buergelt PT, McCaffrey S, Tedim F (eds) Wildfire hazards, risks, and disasters. Elsevier, Amsterdam 5. Statistik Indonesia (2014) Statistical yearbook of Indonesia 2014. BPS – Statistics Indonesia, Jakarta, pp. 1–634 6. Herawati H, Santoso H (2011) Tropical forest susceptibil-ity to and risk of fire under changing climate: a review of fire nature, policy and institutions in Indonesia. For Policy Econ 13:227–233 7. Applegate G, Chokkalingam U, Suyanto S (2001) The underlying causes and impact of fires in southeast Asia, final report. Center for International Forestry Research (CIFOR), Jakarta 8. Adinugroho WC, Nyoman N. Suryadiputra I, Saharjo BH, Siboro L (2005) Manual for the control of fire in peatlands and peatland forest, climate change, forests and peatlands in Indonesia project. Wetlands International – Indonesia Programme and Wildlife Habitat Canada, Bogor

4

Integrating Wildland and Urban Fire Risks in Local Development Strategies in Indonesia

9. Ohlemiller TJ (1985) Modeling of smoldering combustion propagation. Prog Energy Combust Sci 11:277–310 10. Page SE, Rieley JO, Banks CJ (2011) Global and regional importance of the tropical peatland carbon pool. Glob Chang Biol 17:798–818 11. Osaki M, Tsuji N (eds) (2016) Tropical Peatland Ecosystems. Springer, Tokyo, pp. 1–651 12. Wo¨sten JHM, van den Berg J, van Eijk P, Gevers GJM, Giesen WBJT, Hooijer A, Idris A, Leenman PH, Rais DS, Siderius C, Silvius MJ, Suryadiputra N, Wibisono IT (2006) Interrelationships between hydrology and ecology in fire degraded tropical peat swamp forests. Water Resour Dev 22:157–174 13. Presentation materials of Ministry of Forestry of Republic of Indonesia, i.e. Direktorat Pengendalian Kebakaran Hutan, Direktorat Jenderal Perlindungan Hutan dan Konservasi Alam Kementerian Kehutanan, Rapart Koordinasi Nasional Strategi Penguatan Kapasitas Pemerintah Daerah dalam Pengurangan Risiko Kebakaran, Banjarmasin, 28 Feb 2015 14. Ministry of Forestry of Republic of Indonesia, i.e. Direktorat Pengendalian Kebakaran Hutan, Direktorat Jenderal Perlindungan Hutan dan Konservasi Alam Kementerian Kehutanan, website: http://ditpkh-phka.dephut.go.id/ 15. Frandsen WH (1997) Ignition probability of organic soils. Can J Forest Res 27:1471–1477 16. Rein G, Fernandez-Pello AC, Urban DL (2007) Computational model of forward and opposed smoldering combustion in microgravity. Proc Combust Inst 31:2677–2684 17. Saharjo BH (2015) Pembuktian Ilmiah Terjadinya Kebakaran Hutan dan Lahan 18. Badan Nasional Penanggulangan Bencana, Sebaran Titik Panas Nusantara, website: http://geospasial.bnpb.go.id/monitoring/ hotspot/. Accessed on 20 Mar 2015 19. National Environment Agency of Singapore, website: http://www. haze.gov.sg/hotspot-satellite-images. Accessed on 21 Mar 2015 20. Kementerian Dalam Negeri Republik Indonesia, Direktorat Jendral Pemerintahan Umum, Profil Kebakaran 2013 21. Sufianto H, Green AR (2012) Urban fire situation in Indonesia. Fire Technol 48:367–387

43

22. Kampung Improvement Programme Jakarta, Indonesia, the Aga Khan Award for Architecture in 1980, website: http://www.akdn. org/architecture/pdf/0001_Ind.pdf 23. Nugroho YS, Istiyanto J, Erwandi D, Violita ES, Kencana DR, Renggana EA, Ridwan Santoso MA Influence of separation distance on house to house fire spread, submitted to International Journal of Technology (IJTech) 24. Wahyu Sujatmiko W, Dipojono HK, Soelami FXN, Hermawan K, Nugroho Soelamia FX, Soegijanto (2014) Performance-based fire safety evacuation in high-rise building flats in Indonesia – a case study in Bandung. In: 4th international conference on sustainable future for human security, SustaiN 2013, Procedia Environ Sci 20: 116–125 25. Mell WE, McDermott RJ, Forney GP (2010) Wildland fire behavior modeling: percpectives, new approaches and applications. Proceeding of 3rd fire behavior and fuels conference, 25–29 Oct 2010, Spokane 26. Dinas Penanggulangan Kebakaran dan Penyelamatan, website: http://www.jakartafire.net/profil/index.php 27. Indonesia’s Country Report on Disaster Response Management, 3rd AIPA Caucus Report, website: http://www.aipasecretariat.org/ wp-content/uploads/2011/07/Indonesia_Disaster-Response-Man agement.pdf 28. Guo TN, Fu ZM (2007) The fire situation and progress in fire safety science and technology in China. Fire Saf J 42:171–182 29. Leonita F, Sakti H, Nugroho YS, Study of the occupant characteristics during evacuation in medium and high-rise buildings in Indonesia, the 10th AOSFST 2015, Tsukuba (accepted) 30. Santoso MA, Bey Z, Nugroho YS (2015) CFD study on the ventilation system and shape configuration of underground car park in case of fire. Appl Mech Mater 758:143–151 31. Himoto K, Tanaka T (2012) A model for the fire-fighting activity of local residents in urban fires. Fire Saf J 54:154–166 32. Huang X, Rein G, Chen H (2015) Computational smoldering combustion: predicting the roles of moisture and inert contents in peat wildfires. Proc Combust Inst 35:2673–2681

5

Thermal Analysis and Flame Spread Behavior of Building-Used Thermal Insulation Materials Jinhua Sun and Lin Jiang

Abstract

To review the thermal analysis and flame spread hazards of thermal insulation materials, some research work during the past several years by SKLFS would be concluded in this paper. XPS, EPS, and PU were selected to undergo a series of thermal analysis experiments. We used TG-FTIR-GC/MS and explored the degradation mechanism of XPS and PU. The volatile gases during degradation process were analyzed. Then some studies on the radiation and thickness effects on XPS calorimetry were explored. Dr. Chen in SKLFS did some numerical studies on ignition hazards of PU and XPS. Then thermal insulation materials have conducted a series of bench-scale flame spread experiments with the effects of location altitude, sample width, thickness, and inclination angle. Keywords

XPS
EPS
PU
Degradation
Flame spread

5.1

Introduction

With rapid development of modern society, energy sources have become the most important metrical bases to human beings and fossil fuels, such as oil and natural gas, the lifeblood of contemporary national economy. However, contradictions between the nonrenewable character of energy sources and huge demands of human beings are gradually contributing to exhaust of fossil fuels. The crisis of energy shortage has become a global challenge. Many countries have paid high attention one after another to this issue and generally taken two measures, opening sources and reducing expenditure, to deal with it, such as seeking for new energy sources and controlling consumption and waste of existing energy. Under the big circumstance of worldwide energy shortage, problems in China are particularly severe.

J. Sun (*)
L. Jiang State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei, Anhui 230027, China e-mail: [email protected]

According to the current socioeconomic structure in China, consumption of building energy covers 27.6 % of total social energy consumption. In addition, with gradual improvement of people’s living standard, energy consumption by building will be larger and larger. There is no doubt that building energy conservation is important to fulfillment of the whole energy conservation goal. And insulation treatment to the external walls is just the key to effectively conserve building energy. Building industry especially the external wall insulation industry has gone through rapid development, but many issues and dangers are gradually exposed, of which the biggest flaw is the flammability character of thermal insulation material and the tendency to generate large amount of smog even toxic gases. The commonly used thermal insulation materials now are rigid polyurethane (PU), expandable polystyrene (EPS), and extruded polystyrene (XPS). The growing popularity of application has led to more and more fire accidents, causing not only huge losses of people’s life and property but serious social impacts. Typical accidents included the north annex fire of the new site of China Central Televisions that occurred in February 2009,

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_5

45

46

J. Sun and L. Jiang

caused by firework sparks with high temperature, which ignited the exposed thermal insulation material XPS on the external walls and caused direct economic loss of 163.83 million RMB. Besides, many other typical accidents such as the fire accident of Nanjing Zhonghuan International Square, which happened in April at the same year, and fire accidents in the Jinan Olympic Sports Center and Beijing Qiaobo skiing site were also attributed to the ignition of thermal insulation materials. There was one more severe high-rise residence fire that occurred at the Jing’an District of Shanghai in November 2011. According to the accident investigation and analysis, it was caused by unprofessional operation of electric welders. As they were carrying out the energy conservation projects, a mass of flammable materials including polyurethane foams were accumulated around the working site. The dropped sparks caused by electro welding occasionally ignited the thermal insulation materials and made fire spread rapidly. Finally, 58 people were killed and more than 70 were injured. From analysis of those typical fire accidents mentioned above, it can be concluded that after ignition, the fire of organic thermal insulation materials can rapidly spread from the ignition point to the whole building, form comprehensive combustion, and lead to big difficulties in firefighting. Therefore, researches on the fire spread characters of thermal insulation materials are extremely urgent. It is well known that pyrolytic reaction is the initial step for solid material combustion [1], of which the generated flammable gases can support combustion and intensify fire spread. So it is necessary to study the thermal degradation nature of organic thermal insulation materials as well as the danger of volatile products. After that, more fundamental data can be obtained to illustrate the development and spread rule of this type of fire, helping people to understand the main reasons for high-rise building fire and better prevent and control fire disaster. Also, our country has provided great support to this research [2] and provided us with the 973 program fund (973 Program, Grant. No. 2012CB719702).

instrument to record the mass loss and heat flow rate of PU, EPS, and XPS. The TG-DSC scans of three materials at heating rate of 10  C/min in nitrogen are presented in Fig. 5.1. It is obvious in Fig. 5.1a that EPS shows only one mass loss stage and heat absorption peak in a narrow temperature range from 377.4 to 417.3  C although DTG curve displays a small peak at around 120  C. From Fig. 5.1b, it can be seen that a gentle mass loss ratio of 7.10 % and a main mass loss rate of 87.55 % are presented sequentially in the temperature region of 228.0–298.0  C and 342.0–456.0  C. The complex scans in Fig. 5.1c indicate that PU undergoes at least three successive stages, consisting of the mass loss of 11.58 % in 71–234  C at first and then 50.07 % in 234.0–400.0  C and 10.32 % in 400.0–599.2  C. Figure 5.2 shows the active energy distribution of thermal degradation reaction of three materials under nitrogen atmosphere calculated by KAS method [3]. It can be seen from Fig. 5.2 that the active energy of PU is generally much lower than the others. This indicates that PU is easy to degrade. Meanwhile, the active energy curve of PU could be divided into three phases that first increase in the conversion range of α ¼ 0.05–0.25 and decrease when α ¼ 0.25–0.45 and then maintain an approximate constant from α ¼ 0.45 to 0.80. It is suggested that the thermal degradation process of PU can be regarded as a three-step reaction based on the model-fitting analysis. However, the active energy values of XPS and EPS almost keep constant during the entire thermal degradation procedure, which indicates that the thermal degradation process of PS in nitrogen can be approximated as a single reaction step. Whereas, the apparent active energy value of XPS stabilizes in the range of 220–230 kJ·mol1, which is a little lower than that of EPS, about 245 kJ·mol1, which indicates that XPS is easier to degrade under thermal stress in nitrogen, compared with EPS.

5.2

Degradation and Ignition Behavior of Building-Used Thermal Insulation Materials

5.2.1

Degradation, Characteristics, and Thermal Kinetics

To identify the volatile products of materials during thermal degradation is the essential procedure to study the detailed mechanism. Coupling TGA with an evolved gas analyzer such as FTIR or MS has been verified to be an efficient and powerful analytical technique, for the information of the thermal balance, and compositional information from spectrometer can be obtained simultaneously. Jiao et al. [4] used online test method to study the volatile products during thermal degradation process of rigid PU and XPS, respectively. The TG-DTG curves of PU foam and the total ion chromatogram and Gram-Schmidt reconstruction of the

It is known that pyrolysis is the initial stage which provides fuel and energy for combustion, and kinetics study is the fundamental method to investigate thermal degradation process of materials. Jiao et al. [3, 4] used TG and DSC

5.2.2

Volatile Products During Thermal Degradation

5

Thermal Analysis and Flame Spread Behavior of Building-Used Thermal Insulation Materials

a

DSC [mW/mg] DTG [%/⬚C] 28

100

47

b

DSC [mW/mg] DTG [%/⬚C] 0.5

TG

4

DTG

2

100 21

TG

0.0

14

80

80

0

DTG

-0.5

Mass [%]

Mass [%]

7

60 0

DSC

-1

-7

40

0

60

DSC -1.0 -2

40 -4

-14

20

-6

-21 -28

0 0

100

200

300

400

500

-1.5

20 -2.0

0

-2

-8

0

600

100

200

300

400

500

600

-2.5

700

Temperature [⬚C]

Temperature [⬚C]

DSC [mW/mg] DTG [%/⬚C]

c

1.0

100

TG 0.0

90 DTG

0.5

Mass [%]

80 70

-0.2 0.0

60

DSC

50 -0.4 -0.5

40 30 20

-1.0

100

200

300

400

500

-0.6

600

Temperature [⬚C]

Fig. 5.1 TG-DTG-DSC scans of EPS (a), XPS (b), and PU (c) degradations at 10  C/min in nitrogen atmosphere [3]

Fig. 5.2 Thermal degradation activation energies of PU, XPS, and EPS by KAS methods [3]

volatile products detected by TG-FTIR-MS are shown in Fig. 5.3, and the 3-D FTIR spectra of gaseous products at different temperature are shown in Fig. 5.4. The DTG curve presents that the maximum mass loss rate occurs at 350  C and the two dominant peaks of TIC and G-S reconstruction are in the same range of 300–420  C. Figure 5.4 clearly displays the change of volatile products in characteristic functional regions, such as hydroxyl group in 3700–3550 cm1, accumulated double bond stretch in 2500–2100 cm1, and carbonyl group in esters of 1800–1700 cm1. The strongest absorption of almost all evolved products appears at 360  C, and the corresponding absorption bands according to FTIR are summarized in Table 5.1. Furthermore, signals of some simple volatile products, blown agent HCFC-141b (m/z 81, 83, 61, 63), carbon dioxide (m/z 44), HCN (m/z 27), CH3OCH3 (m/z 46), and NO2

48

J. Sun and L. Jiang

Fig. 5.3 TG and DTG curves of PU foam. Total ion chromatogram (TIC) and GramSchmidt reconstruction of the volatile products [4]

Table 5.1 Summary of FTIR bands observed in the thermal degradation products of PU [4] Band position (cm1) 1100 1627 1042, 1269 2977, 2936, 2887 931, 1376, 1450 3700–3550 1514 2358, 2312 1724, 1750

Fig. 5.4 The FTIR spectra of volatile products of PU during the initial degradation stage [4]

(m/z 30, 46) at different temperature recorded by mass spectroscopy in SRM mode are given in Fig. 5.5. In accordance with the result of FTIR test, the release of carbon dioxide shows two obvious stages in the temperature regions of 250–410  C and 410–670  C.

Assignment Stretch of C-O-C in high polar ethers Vinyl ether Symmetric and asymmetric of C-O-C in aromatic and alkyl ether Stretch of CH3, CH2, and C-H -C-(CH3)3 -OH stretch (hydrogen bonded) Stretch of N-H in aromatic amine CO2 C¼O in ether

HCN, a kind of toxic gas resulting in the suffocation deaths with lethal dose of 0.035 % in 10 min, is released at 365  C. This adds the danger of evacuation for the maximum weight loss rate taking place at 350  C. W. D. Woolley [5] and Barbara C. Levin [6] experimentally studied the generation of hydrogen cyanide under nitrogen and air atmospheres, respectively. It is found that the HCN concentration was directly related to the amount of char formed,

5

Thermal Analysis and Flame Spread Behavior of Building-Used Thermal Insulation Materials

49

Fig. 5.5 Mass spectra of small molecule products during PU degradation [4]

and multiple compounds containing amine, amide, imine, and nitrile functional groups may be the sources of HCN. Combining the data collected by FTIR and MS analyzers, some other volatile products during the thermal degradation process can be identified, as listed in Table 5.2. Jiao et al. [7] also studied the volatile products of XPS under thermal stress in helium with TG-FTIR-GC/MS instrument and the TG-DTG curves of XPS foam; the total ion chromatogram and Gram-Schmidt reconstruction curve and 3-D FTIR spectrum of the volatile products as well as the gas chromatogram map of products evolved at 400  C are displayed in Figs. 5.6, 5.7, and 5.8. From Fig. 5.7, it can be observed that the absorption peaks of benzene ring skeleton (3075, 3029, 1600, 1496, 694) and the methylene (2862, 2938) are the main specific absorption peaks at entire temperature region. The four peaks with different retention time can be attributed to the products listed in Table 5.3. Meanwhile, the concentration distribution of main products, including 1,3-butadiene, styrene, toluene, and α-methyl styrene recorded by MS, indicates that thermal degradation of PS mainly takes place via random-chain scission and hydrogen transfer.

5.2.3

Cone Calorimetry Testing and Ignition by Hot Particle

In order to investigate the ignition mechanism of PU and PS under radiation, Wang [8] did some cone calorimetry testing on PU with different isocyanates, 1.05, 1.10, and 1.20, with a radiation of 35 kW/m2. An [9] did this work using XPS and EPS considering the effects of sample sickness and different radiation intensity. Both data under 35 kW/m2 are listed in Table 5.4. Time to ignition means the difficulty level of being ignited. From the above table, we can find that time to ignitions of three PU is much shorter than XPS and EPS. The time to ignition of PU is only several seconds, which indicates that PU is more easily ignited than XPS and EPS. The being ignited sequence is in the same order with degradation activation energy. The higher activation energy means higher degradation temperature. They have the same order, which indicates the difficulty level of degradation and combustion. For the reason that XPS and EPS would shrink when being heated, An [9] created a modified ignition model by analyzing the relation between ignition and heat flux in

50

J. Sun and L. Jiang

Table 5.2 Identification of volatile products during degradation of PU Figures of MS Figure 5.4

Tp( C) 165

Figures in [2]

230

Figures in [2]

370

Corresponding bands (cm1) 751.9 1006.5–1153.6 931 1376 1450

Evolving products CH3CCl2F OH

HO

OH O O

O

O

O

O O

HO

1100 1627 2887 2936 2977

O OH

Figures in [2]

485

1514

NH2 NH

CH3

H N

Fig. 5.6 TG curve and GramSchmidt reconstruction of the volatile products of XPS in helium atmosphere [7]

N H

5

Thermal Analysis and Flame Spread Behavior of Building-Used Thermal Insulation Materials

51

Fig. 5.7 FTIR spectra of volatile products of XPS during the initial degradation stage [7]

Fig. 5.8 Gas chromatography of XPS degradation volatile products at 400  C in helium atmosphere [7]

different thickness. The expression could be described as follows: tig ¼ FðdÞ2 tig ¼ ð1:0142  0:0674dÞ2 tig / 1=q_

00

2

ð5:1Þ

The results predicted by this modified model have a good linearity with experimental results, with only 7 % error. Pyrolytic reaction of external wall insulation materials can generate large amounts of flammable gases and release large heat, which will cumulate to ignite the materials easily. With isothermal analysis and model-fitting method, Wang et al. [10] studied the ignition danger of rigid PU by hot particles under open environment and pointed out that the thermal decomposition of PU under air atmosphere had three stages: loss of organic components due to low stability, oxidation of main organic components, and oxidation of residues. They also calculated the active energies of every stage with iso-conversional method, Friedman method, and OFW method, and the active energies were from 60.17 kJ/

mol to 260.39 kJ/mol. The released heat for the latter two stages was, respectively, as high as 5443.09 kJ/mol and 10,518.51 kJ/mol. Through combined calculation of the thermal kinetic parameters and dynamic parameters, it was found that under certain temperature, the rigid PU foam could not be ignited unless the diameter of hot particles reached the critical diameter rcrit, which could be obtained through the Frank-Kamenetskii parameter δcr. Therefore, with reasonable assumption on the physical property parameters of mental particles and PU, it is possible to know how the RPU is ignited by hot particles of aluminum oxide. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi     ffi r crit ¼ δcr λ0 = ρ0 AH P RT 2P =E exp E=RT 2P ð5:2Þ Furthermore, based on the special phenomenon that firework can ignite the external wall insulation, Song et al. [11] made detailed calculations on the motion trial of hot

52

J. Sun and L. Jiang

Table 5.3 Identification of volatile products during degradation of XPS m/z 39, 65, 91, 92

Temperature region ( C) 375–550

2

50, 51, 52, 63, 77, 78, 102, 103, 104, 105

350–500

3.03

3

77, 78, 91, 103, 115, 117, 118, 129

370–600

3.68

4

129, 130, 179, 180, 193, 194, 195

400–650

12.42

Peak 1

Assignments

Retention time (min) 2.5

Table 5.4 Cone calorimetry testing results of PU, XPS, and EPS with 35 kW/m2 radiation PU105 PU110 PU200 XPS EPS

Time to ignition/s 6 6 4 62 86

Total heat release/(MJ/m2) 53.07 44.18 22.09 45.86 25.94

aluminum particles under different initial velocities, initial temperatures, and diameters and found out that the influences of particle diameter, initial velocity, and initial temperature on trial were orderly weakened. With the same velocity, particles with higher initial temperature and smaller diameter could move further, but the temperature also decreased. As smaller particles carried limited quantity of heat, they did not necessarily have high danger. Meanwhile, the critical horizontal movement distance was found to be in direct proportion to the initial velocity, while the vertical movement distance was in inverse proportion. Conservatively estimating, the temperature for hot particles to ignite rigid polyurethane foam is at least 500  C and will be relatively safe when firework discharge point is 77.4 m far and 13.4 m high. These results have important practical significance.

Mean heat release rate/(kJ/m2) 145 133.74 97.46 240.13 180.49

Peak heat release rate/(kJ/m2) 335.38 344.10 290.50 359.03 446.43

5.3

Flame Spread Behavior of BuildingUsed Thermal Insulation Materials

5.3.1

PS Flame Spread Characteristics

Flame spread over combustible solid surface is actually the heat and mass transfer processes among burning flame, pyrolysis zone, and virgin materials. For material burning area, heat transferred from the flame makes combustible pool region pyrolyze, and then combustible solid and liquid in the pool zone are heated to generate combustible vapor, which would be ignited by flame immediately. This process makes the flame sustaining until combustible solid in the pool fire area runs out.

5

Thermal Analysis and Flame Spread Behavior of Building-Used Thermal Insulation Materials

Meanwhile, virgin materials in the front are also heated by flame, including heat transferred through solid phase, convective, and radiative heat flux above the materials surface. For being heated by flame, virgin materials could generate combustible gas. This process makes flame propagate from burned material zone to virgin zone continuously. XPS is a typical thermoplastic thermal insulation material. It becomes soft gradually and then shrinks and turns liquid phase eventually. Liquid XPS attaches to the wall or ground and ultimately congregates toward pool fire. What’s more, the liquid fuel would continuously pyrolyze and generate combustible gas to support liquid pool for burning. For materials’ thickness, flame would spread along both the XPS surface and broadside in the meantime. The flame spread rate along the broadside would be faster than the surface, which would result in the U-shaped structure at the flame front. During the XPS vertical flame spread process, for its melting property, there would be a dripping and flowing phenomenon. The dripping liquid XPS would form a plastic pool fire on the ground and then flow all around. The fire that occurred on CCTV building has its ignition position on the building roof. The reason why it is hard to control and spread so fast is that melting XPS could spread in all directions.

5.3.2

PU Flame Spread Characteristics

PU is a thermosetting material. It has little obvious shape change when being heated or undergoing combustion. It would pyrolyze and char with the effects of strong heat. The PU board would occur to bend for the carbon layer tension during the flame spread process. When the flame propagates across the PU surface, it would form a compact layer adhered to the surface tightly, which could inhibit heat and oxygen to material inner. So the combustion phenomenon only occurs at the material surface and cannot affect into the deeper inner. This phenomenon Fig. 5.9 Schematic representation of heat transfer for flame spread over thermoplastic materials’ surface [13]

53

was observed in our bench-scale experiments. However, in real building fire disaster, for the reason that there would be large radiation heat from surrounding, the carbon layer could not prevent flame spread over PU surface. What’s more, no pool fire forms for the property of charring. Only surface flame exists. The flame spread behavior of PU has the similar phenomenon with wood board, no melting, pyrolysis at the surface, and leaving carbide. When the flame spread from down to up, the carbon zone formed before the flame front could inhibit flame propagation. Then the flame could occur to extinguish. However, when the flame spread from up to down, it will not extinguish though the flame spread slowly. This is because at this time there is no carbon existing in the flame front. The heat transfer could be transferred to the virgin materials.

5.3.3

Experimental Conditions’ Effects on Flame Spread

In the past 3 years, we have studied a lot about horizontal flame spread over polymer surface [12–15], including the width effects on flame behavior, thickness, and environmental pressure. The mechanism of flame spread is actually determined by the heat transfer from the flame to the surface, as illustrated in Fig. 5.9. XPS and EPS are two kinds of polystyrene with different density, which both are thermoplastic. And PU has a different thermal behavior when being heated; it is thermosetting. The data we monitored during the testing process includes temperature variation in solid phase (pyrolysis temperature, melting temperature drawn in Fig. 5.10), flame temperature, flame height, flame spread rate, pool fire length, mass loss rate, and so on. Then we found that the flame spread rate would decrease first and then increase with the increase of sample width. What is more, the same phenomenon could also be found in XPS and wood board flame spread on plateau [14, 15], which was drawn in Fig. 5.11.

54

Fig. 5.10 Temperature variation in solid phase and gas phase: (a) EPS [12]; (b) XPS [12]; (c) PU [13]

Fig. 5.11 Flame spread rate of XPS and wood with different sample width and thickness [12– 15]

J. Sun and L. Jiang

5

Thermal Analysis and Flame Spread Behavior of Building-Used Thermal Insulation Materials

The heat transfer relationship between flame spread rate and heat flux by Williams [16] could be expressed as vρΔh ¼ qcond þ qconv þ qrad

ð5:3Þ

where v means flame spread rate, ρ is density of material,and Δh is enthalpy difference. qcond is heat conduction,qconv is heat convection,and qrad is heat radiation. The heat conduction could be written as qcond

∂T ¼k ∂x

ð5:4Þ

However, by calculating the conduction part, we found that this part could be ignored, compared with the convection and radiation parts [12, 13]. The convection of flame to unburned flame front nearly materials could be expressed as qconv8 9   14 > >  < g cos θβ T f  T P 3 = k 0:67 ¼ 0:68 þ h L 4 i 9 9 > L> av : ; 16 1 þ ð0:492=Pr Þ    Tf  TP

ð5:5Þ The radiation part is

qrad ¼ σ T 4f  T 4P ð1  expðks LÞÞ

Fig. 5.12 The curve of total heat flux against sample width [13]

ð5:6Þ

55

Combining the above equations, we could get the equation about flame spread and heat flux: V f /8q 9   14 > >  < gcos θβ T f  T P 3 = k 0:67 ¼ 0:68 þ h L 4 i 9 9 > L> av : ; 1 þ ð0:492=PrÞ16

   T f  T P þ σ T 4f  T 4P ð1  expðks LÞÞ

ð5:7Þ Some parameters in this equation could be obtained from our testing or previous reference, and the above equations could be simplified as, for example, for XPS [13]: Vf / q ¼ 13:13L1 þ 873:6L0:25   þ 41757:34 1  e0:37L

ð5:8Þ

From Eq. 5.8, we could find that the total heat flux would decrease first and then increase with the increase of sample width as drawn in Fig. 5.12, who has the same variation tendency with flame spread rate. The theoretical lowest point of flame spread rate is 5 cm. When the width is thinner than 5 cm, the flame spread is controlled by heat convection, and when the width is wider than 5 cm, it would be controlled by radiation. During the stable stage of flame spread, there would be a plastic pool fire that formed on the board. And the flame

56

J. Sun and L. Jiang

a

500

400 200

300

150

250

100

C

Melting zone

40 30

C

B

0 350 360 370 380 390 400 410 420

150

50

D

50

200

60

tig

20

dT/dt(⬚C/s)

350

B

80 Pyrolysis 70 zone

Melting zone

Preheated zone

T dT/dt

450

T(⬚C)

Fig. 5.13 Temperatures in solid phase and gas phase of XPS on plain and plateau [13]. (a) XPS on plain, (b) XPS on plateau

10 0

100

A

50

-10 -20 450

0 0

50

100

150

200

250

300

350

400

Time(s)

b

80

500

T dT/dt

450 400

Pyrolysis zone

zone

350

200

D'

40

100

250

30

50

200

0 170

60 50

tig

150

70

180

190

200

B'

210

20

C'

dT/dt(⬚C/s)

300 T(⬚C)

Preheated zone Melting

10

150

0

100

A'

50

-10 -20

0 0

40

80

120

160

200

240

Time(s)

height above the plastic pool has such a relationship with heat release rate as follows: For EPS: logðL þ 1:02DÞ ¼ logQ0:65  0:39

ð5:9Þ

For XPS: logðL þ 1:02DÞ ¼ logQ0:26  0:90

ð5:10Þ

logðL þ 1:02DÞ ¼ logQ0:60  0:83

ð5:11Þ

For PU:

Such equations are usually used in jet flame or simple fuel pool fire. However, we found that it could also be used in solid fuel surface flame spread with a high linear fitting.

Zhang [14] has conducted a series of experiments about horizontal flame spread over XPS surface with different sample width. A thermocouple tree was used to monitor the temperatures in solid phase and gas phase shown in Fig. 5.13. He found that it shows a difference between plain and plateau. And three stages in solid phase are separated out: preheating stage, melting stage, and pyrolysis stage. Then, An [17] and Gong [18] studied the downward flame spread on plateau and showed a reasonable explanation from heat transfer aspect. Then the relationship between mass loss rate and environmental pressure is obtained: m_ / Ch  P2n

ð5:12Þ

This equation indicates that the mass loss rate of flame spread has a positive correlation relationship with square of pressure. Then Gong conducted a series of experiments

5

Thermal Analysis and Flame Spread Behavior of Building-Used Thermal Insulation Materials

considering different altitude in Lhasa, Xining, and Hefei, respectively. And the exponential relation between mass loss rate and pressure altitude was modified as m_ / Ch  P1:8 . Then An [17] tested the XPS downward flame spread on plain and plateau, respectively, and found that for XPS, n ¼ 1.9–2.0. He explained that the higher value than Gong [18] is resulted from the melting dripping velocity on plateau, which would lead to the less mass loss rate. To study the sidewall hazardous effect on building outer wall, An [19] has done some studies on downward flame spread with and without sidewalls, respectively. And a mathematical modeling about the sidewall effect was built, with flame spread rate fitting. An [19] found that during the downward process, average flame height would increase first and then decrease with the raise of sample width, which is shown in Fig. 5.14. This increase would be a linear growth for condition without sidewall. H value for plain is larger than that for plateau. Above variation tendency is consistent with flame pulsation intensity. On plain area, for thinner sample, H value with sidewall is larger than that without sidewall. However, for wider sample, it shows an absolute opposite result. On plateau area, H value with sidewall is always larger than that without sidewall for all sample width. The flame spread rate could be expressed as vf ¼ Að1  expðCd ÞÞ. And the relationship between heat release rate and sample thickness could be written as Q_ * / expð0:3d Þ. Dimensionless flame height could be calculated pffiffiffi by H *f ¼ a=d þ bρ1 cp T 1 gd3=2 Q_ * =Δh. And the above equations were combined by An [17, 19], getting flame spread rate formula with adjusting heat flux and pressure as this,

  ∂T m  þ k1 a 2k2 W 5=4 P1=2 þ W 1=4 P1=2 k2 =δ  ∂x x¼xig  þ 2k3 P2 þ WP2 k3 =δ ð5:13Þ 

∂T m  ∂x x¼xig 5=4 1=2 1=4 1=2

where

means

heat

conduction,

P þW P k2 =δ means heat convection, 2k3 2k2 W 2 2 P þ WP k3 =δ is the radiation part. An [19] referred that when there is no sidewall, and downward flame spread mainly is controlled by gas heat flux. For thinner sample, the main gas heat flux is convection. The flame spread rate is mainly decided by convection part 2k2 W 5=4 P1=2 þ W 1=4 P1=2 k2 =δ. So the rate would decrease with the sample increase. For wider sample, it is mainly radiation part, 2k3 P2 þ WP2 k3 =δ. So this explained the experimental phenomenon in Fig. 5.15. Zhou [20] studied flame spread over inclined rigid polyurethane surface with different inclined angles (upward 0–90 , downward 0–90 ). The fitting relationship between flame length, xf, and pyrolysis zone length, xp, and flame 0 length and heat release rate per unit length, Q_ , during  m upward inclined surface, was obtained, xf ¼ a xp . Zhou [20] modified the Rayleigh number, which was then used to zone the different inclined PUR upward flame spread. According to the flow condition, each power exponent fitting relation has been obtained and separated; when 0 the inclined angle is 10–40 , n ¼ 1 until Q_ reached the stable value. When the inclined angle is 50–70 , n < 1. When the inclined angle is 80–85 , n > 1. And in laminar flow zone, the index n is equal to 1 between flame length and

plain,without sidewalls plateau,without sidewalls plain,with sidewalls plateau,with sidewalls Huang,XPS,plateau,surface flame Huang,XPS, plain, surface flame Huang,EPS,plateau,pool flame Huang,EPS, plain,pool flame

35 Average flame height/cm

Fig. 5.14 Flame height of downward flame spread of different sample width with and without sidewall [19]

vf ¼ k1 λm

57

30 25 20 15 10 5 2.0

2.5

3.0 3.5 4.0 4.5 Sample thickness/cm

5.0

58

Lhasa, without sidewalls Lhasa, with sidewalls Hefei, without sidewalls Hefei, with sidewalls

0.22 Flame spread rate/cm s–1

Fig. 5.15 Flame spread rate of downward flame spread with and without sidewall [19]

J. Sun and L. Jiang

0.20 0.18 0.16 0.14 0.12 0.10 4

6

8 10 12 Sample width/cm

14

16

Fig. 5.16 Linear correlation relation of flame length and pyrolysis length of inclined upward PU flame spread

heat release rate per unit width. In transition region, n < 1. And in turbulent zone, n > 1. Figure 5.16 is the relation between flame length and heat release rate during inclined upward flame spread. Then, by the analysis of mass loss rate, Zhou found that mass loss rate in transition region has such relation with modified Rayleigh number: X 00 M*P ¼ m_ xPr 1=4 ðu1 Þ ð5:14Þ In the range of sample inclined angle smaller than downward flame spread transition angle, the rate has a 1/3 power with stable mass loss rate shown in Fig. 5.17.

5.4

Conclusions

Based on our research review, the results could consider to be transferred into actual thermal insulation construction. For example, for the reason that the degradation temperatures of PS and PU are low, some flame retardants could be added, which could be used to improve its ignition temperature, or the reaction routine could be changed, resulting less poisonous gases in case that the fire disaster occur. As to XPS and EPS, these kinds of easily melting polymers, the melting and dropping polymers increase the intensity of flame spread. So some fire protection barriers

Thermal Analysis and Flame Spread Behavior of Building-Used Thermal Insulation Materials

Fig. 5.17 Fitting relationship of downward flame spread with (sinθ)1/3

59

Downward FSR of PUR foam / mm/s

1.8

Downward FSR of PUR foam Linear fit of downward FSR o o Chen(Paulownia,10 ~50 ) o o Zhang(Whitewood,10 ~40 ) o o Kashiwagi(Cellulose, 5 ~30 )

1.6

1.4

2.4

2.0

1.6 1.2 1.2

1.0

0.8

0.8

Downward FSR of other materials / mm/s

5

0.4

0.6 0.4

should be considered when the building insulation system is constructed. In the future, some new researches would be involved. Now, the flame spread experiments are all bench scale. So in the next step, some middle-scale flame spread, closer to the real building fire disaster, would be detected. And the models in our bench-scale experiments would be verified into large scale to verify its effectiveness. The relationship between polymer degradation and ignition, even flame spread, would be created. By controlling the degradation data, flame spread parameters wish to be predicted. Acknowledgment This study was supported by the National Basic Research Program of China (973 Program, No.2012CB719702) and Key Technologies R&D Program of China during the 12th Five-Year Plan Period (No.2013BAJ01B05). The authors greatly acknowledge these supports.

References 1. Quintiere JG (2006) Fundamentals of fire phenomena. Wiley, Chichester 2. Sun J, Hu L, Zhang Y (2013) A review on research of fire dynamics in high-rise buildings. Theor Appl Mech Lett 3(4):042001 3. Jiao L, Xu G, Wang Q et al (2012) Kinetics and volatile products of thermal degradation of building insulation materials. Thermochem Acta 547:120–125 4. Jiao L, Xiao H, Wang Q et al (2013) Thermal degradation characteristics of rigid polyurethane foam and the volatile products analysis with TG-FTIR-MS. Polym Degrad Stab 98 (12):2687–2696 5. Woolley WD, Fardell PJ (1982) Basic aspects of combustion toxicology. Fire Saf J 5(1):29–48 6. Levin BC, Paabo M, Fultz ML et al (1985) Generation of hydrogen cyanide from flexible polyurethane foam decomposed under different combustion conditions. Fire Mater 9(3):125–134

0.5

0.6

0.7 (sinq)1/3

0.8

0.9

7. Jiao L, Sun J (2014) A thermal degradation study of insulation materials extruded polystyrene. Procedia Eng 71:622–628 8. Wang H, Wang Q, He J et al (2013) Study on the pyrolytic behaviors and kinetics of rigid polyurethane foams. Procedia Eng 52:377–385 9. An W, Jiang L, Sun J et al (2015) Correlation analysis of sample thickness, heat flux, and cone calorimetry test data of polystyrene foam. J Therm Anal Calorim 119(1):229–238 10. Wang S, Chen H, Liu N (2015) Ignition of expandable polystyrene foam by a hot particle: an experimental and numerical study. J Hazard Mater 283:536–543 11. Song J, Wang S, Chen H (2014) Safety distance for preventing hot particle ignition of building insulation materials. Theor Appl Mech Lett 4(3):034005 12. Jiang L, Xiao H, Zhou Y et al (2014) Theoretical and experimental study of width effects on horizontal flame spread over extruded and expanded polystyrene foam surfaces. J Fire Sci 32(3):193–209 13. Jiang L, Xiao H, An W et al (2014) Correlation study between flammability and the width of organic thermal insulation materials for building exterior walls. Energy Build 82:243–249 14. Zhang Y, Huang X, Wang Q et al (2011) Experimental study on the characteristics of horizontal flame spread over XPS surface on plateau. J Hazard Mater 189(1):34–39 15. Zhang Y, Ji J, Huang XJ et al (2011) Effects of sample width on flame spread over horizontal charring solid surfaces on a plateau. Chin Sci Bull 56(9):919–924 16. Williams FA (1977) Mechanisms of fire spread[C]//Symposium (International) on Combustion. Elsevier. 16(1):1281–1294 17. An W, Huang X, Wang Q et al (2015) Effects of sample width and inclined angle on flame spread across expanded polystyrene surface in plateau and plain environments. J Thermoplast Compos Mater 28 (1):111–127 18. Gong J, Zhou X, Deng Z et al (2013) Influences of low atmospheric pressure on downward flame spread over thick PMMA slabs at different altitudes. Int J Heat Mass Transfer 61:191–200 19. An W, Wang Z, Xiao H et al (2014) Thermal and fire risk analysis of typical insulation material in a high elevation area: influence of sidewalls, dimension and pressure. Energy Convers Manag 88:516–524 20. Huang X, Zhao J, Zhang Y et al (2015) Effects of altitude and sample orientation on heat transfer for flame spread over polystyrene foams. J Therm Anal Calorim 121(2):641–650

6

A Discussion on Tall Building Fire Safety in the Asia-Oceania Regions Wan-Ki Chow

Abstract

Many new tall buildings of height over 300 m have been constructed in the Asia-Oceania regions over the past decade. Fire safety provisions were determined by following the prescriptive codes developed years ago for traditional buildings up to 120 m tall. Performance-based approach was applied when these buildings failed to satisfy the prescriptive codes. With the occurrence of many big fires in recent years, people are beginning to seriously question whether current fire safety provisions are adequate in these tall buildings with new architectural design features. Very few comprehensive studies reported on the dynamics of room fires in tall buildings. In this connection, a number of scenarios are of particular importance in scrutinizing new fire safety problems arising from tall buildings. In the first place, wind action on tall residential buildings with open windows will supply adequate oxygen to burn large amounts of combustibles present. Secondly, a big fire with high heat release rate only gives short ‘available safe egress time’. However, high occupant loading in tall buildings needs long ‘required safe egress time’. This discrepancy in safe egress times would cause serious safety problems. Thirdly, special designs such as open kitchens in small flats of very tall buildings would raise new problems in fire safety provisions that have not been thoroughly considered. This presentation will discuss the following general aspects of tall building fires: 1. Performance-based approach in determining fire safety provisions and problems resulting from using the approach adopted for traditional buildings 2. The effect of wind action on room fire at height 3. Residential building fire with an open kitchen and necessity of providing sprinkler system 4. Evacuation strategy with refuge floors, elevator and skybridge at height to reduce the egress time 5. The possibility of generation of internal fire whirl in a vertical shaft Keywords

Tall buildings
Fire safety
Performance-based design

W.-K. Chow (*) The Hong Kong Polytechnic University, Kowloon, Hong Kong e-mail: [email protected]; [email protected] # Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_6

61

62

W.-K. Chow

Nomenclature Q_ T m_ air Q_ vent Av Cd F Fw g Ho Hv n NT te Vf Vo Vw z ΔCp

Total heat release rate (in MW) Mass rate of air intake (in kgs1) Maximum heat release rate (in MW) Area of an opening (in m2) Discharge coefficient Evacuation flow capacity (in number of persons/s) Volumetric air flow rate prompted by wind (in m3s1) Gravitational force (in ms2) Reference height (in m) Height of an opening (in m) Dimensionless exponent Number of persons Evacuation time (in s) Ventilation factor for an opening (in m5/2) Wind speed at reference height Ho (in ms1) Wind speed at height z (in ms1) Height (in m) Dimensionless pressure coefficient

Abbreviation ASET CFD EPBFC FEA FLD IFW PBD PC RSET

6.1

Available safe egress time Computational fluid dynamics Engineering performance-based fire code Fire engineering approach Fire load density Internal fire whirl Performance-based design Prescriptive code Required safe egress time

Introduction

With the rapid growth of economics in the Asia-Oceania regions, many tall buildings with new architectural features are constructed [1] in cities and even in small towns as in Fig. 6.1. There are many buildings of height over 300 m, classified as ‘supertall buildings’ or ‘very tall buildings’ [2, 3]. Along with this new development, the number of tall building fires is observed to be increasing in many places in the Asia-Oceania regions. Taking Hong Kong as an example, the big Garley Building accidental fire killed 37 people in that commercial building in 1996 [4, 5]. Big fires also occurred in residential buildings [6–9], an example being the one started from burning polyurethane sofa [6] in 1998. Blue flame was reported in one case [8] in 2014, which was suspected to involve the burning of town gas. Another fire [9] started from the explosion of an old air conditioner

Fig. 6.1 A small town in Northeast China

due to the use of a new refrigerant following the implementation of an environmental protection scheme a decade ago. Flame, not just smoke, spread from the windows in these fires. Fire safety has drawn greater public attention after the occurrence of these big fires. The fire safety requirements in tall residential flats were then specifically revised. However, it is not easy to upgrade the fire safety provisions satisfactorily in the existing buildings constructed decades ago. Fire safety provisions required in both new and existing tall buildings should be studied carefully to provide adequate protection [10, 11]. There are various aspects to attend to in the fire safety provisions for tall buildings. As an example, it is important to note that even for the small fire that occurred in the supertall building Petronas Towers in Kuala Lumpur, Malaysia, in 2005, total evacuation took about 2 h [12, 13]. There are four possible building fire code systems [14– 16]. Prescriptive Code (PC) [17, 18] and Engineering Performance-Based Fire Code (EPBFC) [19–23] are at the two extremes. In between, the fire engineering approach (FEA) [1, 16, 18] and performance-based design (PBD) [24–27] can be applied for projects that fail to comply with the PC. Extensive research is required for developing EPBFC and so it is only available in very few countries.

6

A Discussion on Tall Building Fire Safety in the Asia-Oceania Regions

PC for tall buildings were developed many years ago with fire safety requirements specified [28] for traditional buildings normally of height up to 120 m in many places. Many supertall buildings [1, 11, 29, 30] with new architectural features fail to comply with the requirements specified in PC. In Hong Kong, FEA [1, 16, 28] similar to PBD [24–27] in other countries is allowed in determining fire safety provisions of buildings when there are difficulties in satisfying the code requirements. In fact, FEA is also specified in updated PC [18] after reviewing for over a decade in Hong Kong. With limited research support, the fire engineering profession is still not mature yet. There are always arguments in matters related to approval of building project submissions among officials, developers and fire engineers when they have different views on the fire safety designs and interpretation with reference to the PC requirements or FEA report. Natural ventilation provision through wind was promoted recently in Hong Kong [31]. Consequently, air ventilation assessment has to be carried out for new building projects. Similar to other green buildings, this poses challenges on fire safety [11, 16, 21, 22]. New residential buildings have openable windows and green balconies with glass doors, in addition to tallness [30]. High wind speed through room openings at upper levels would supply adequate oxygen to burn up large quantities of combustibles with high fire load density (FLD) [17, 29, 32]. Consequently, a big postflashover room fire is likely to break out [33]. More air is supplied to the higher levels of a supertall building for combustion to give a big fire. Smoke and fire can then spread [30] to adjacent areas and even to nearby buildings easily through openings, such as open windows, doors to green balconies and refuge floors. Strong airflow can bring firebrands emitted from the burning of modern synthetic materials to ignite the combustibles in the neighbourhood, resulting in an array of building fires. All fire incidents [7–9] have alerted the general public to the additional hazards of tall residential building fires. Note that the FLD was up to 1,400 MJm2 in residential buildings [32], which is much higher than 1,135 MJm2, the upper limit specified in the fire code [17]. This presentation will discuss the following general aspects of tall building fires: 1. Performance-based approach [1, 16, 28] including PBD or FEA in determining fire safety provisions when the building fails to comply with the PC and problems resulting from using the approach adopted for traditional buildings 2. The effect of wind action on room fire at height 3. Residential building fire with an open kitchen and necessity of providing sprinkler system in residential buildings

63

4. Evacuation strategy with refuge floors, elevator and skybridges at height to reduce the egress time 5. The possibility of generation of internal fire whirl (IFW) [34–37] in a tall vertical shaft

6.2

Performance-Based Approach

Prescriptive fire codes cannot be established quickly to cope with the rapid construction development with new architectural features. Taking Hong Kong as an example, the current fire safety codes [14, 15, 17, 18] were established after a long consultation period and hot debates. However, there are difficulties for some supertall buildings [11, 30] to satisfy the requirements stated in the current PC, such as taking less than 5 min for occupants to be evacuated to the protected staircase [38] for tall buildings with a large floor area. Performance-based approach, either PBD or FEA [19, 20, 24–27, 39, 40], in many places is commonly adopted in modern supertall buildings to determine the fire safety provisions. As very little research [41, 42] was carried out in these very tall buildings, design data, verification method and assessment criteria are not yet available in many AsiaOceania places. This is different from the more advanced countries with long-term systematic studies in PBD and EPBFC [20–22, 39] and regular reviews [40]. Therefore, engineers are required to justify their FEA-PBD recommended fire safety strategy. Mandatory hot smoke tests are required for big halls in tall buildings during the inspection of smoke extraction systems [43]. An important part in specifying possible hazards in FEA-PBD projects is to determine the design fire. The heat release rate of the design fire is the first parameter in fire hazard assessment to determine appropriate fire safety provisions. However, the design fire depends on the use of the building, the materials used and the amount of materials present. The heat release rate of an occupancy depends on the quantity of materials and how they are burnt. Fire development depends on many factors, including the type and quantity of combustible materials specified by the FLD, the orientation and position of the combustible materials present, the supply of oxygen and ventilation conditions and the availability of fire service installations. Therefore, the ‘design fire scenario’ [1, 16, 28] is one of the primary uncertainties in PBD. However, there is very little fire testing data on the above items in the literature, and no Hong Kong-specific data is available. Constant heat release rates were used for shopping malls or public transport terminals [44] with low values of 5 MW, 2.5 MW or even 2 MW for areas protected by sprinklers. Even for a small newsagent’s shop, a big fire with an 8 MW heat release rate was measured [45]. Note

64

W.-K. Chow

that burning potato crisps [46] would result in a fast t-squared fire up to 6 MW. All models that are used to calculate the heat release rates have to be verified. Factors to be considered include thermal radiation to the burning surface, intermediate combustion chemistry and turbulent mixing of air and fuel. It is almost impossible to develop such a model without using a large volume of empirical data. Very few studies had been conducted to examine the probable heat release rate in a residential flat, particularly for local buildings, to compile a heat release rate database, while implementing FEA-PBD two decades ago. That is why design fires are reviewed regularly in many other countries [47]. A framework for selecting design fire [47] has been proposed. The value of heat release rate is normally deduced from literature results on combustibles present inside a house. However, stack effect and wind action would supply more air, hence affecting the resultant building fire. For tall residential buildings with open windows, the heat release rate for an accidental fire can be very high due to high FLD and strong airflow induced by stack effect and wind action. These effects on post-flashover fires have been reported [33], with results useful for FEA-PBD projects. Note that offices, hotels and apartments are often housed together in many new tall buildings in the Asia-Oceania regions [11, 30]. Fire hazard assessment in buildings with multiple functions should be watched.

6.3

Wind Action

Survey studies on buildings indicated that the FLD [17] of buildings in Hong Kong is very high, due to the excessive presence of movable combustibles including clothing, books and plastics products. For tall residential buildings that are over 40 years old, domestic fuels such as kerosene and liquefied petroleum gas cylinders are stored. The highest FLD of old buildings was 1275 MJm2 as surveyed in 2007 [29], but up to 1400 MJm2 in 2010 [32], all exceeding the local allowable limit of 1135 MJm2 [17]. However, a high FLD does not necessarily give rise to a big fire in buildings of normal height, considering the lower air intake rate through openings. A fire started inside a room would have more unburnt fuel vapour. This is known as a ventilation-controlled fire. However, at the upper levels of supertall buildings, air flow rate through openings induced by wind can be very high. Adequate oxygen is then supplied to burn up all combustibles and liberate a higher heat release rate. Note that the materials will be burnt more vigorously under high heat fluxes to yield a big fire. Earlier studies reported that a 4 MW fire would grow into one of 80 MW in a very tall building [33, 48]. The fire dynamics of burning combustibles in supertall buildings with high air intake rates

under high radiative heat fluxes must be studied properly in developing PC, or determining acceptance criteria for FEA-PBD projects. In a normal room fire, there is more fuel than air after flashover. The burning rate of fuel depends on the airflow rate through openings. Burning 1 kg of air would generate 3 MJ of heat. The ventilation factor Vf (in m5/2) for an opening of height Hv (in m) and area Av (in m2) is a key parameter to determine the air intake rate: pffiffiffiffiffiffi V f ¼ Av H v ð6:1Þ The mass rate of air intake m_ air (in kgs1) induced by a fire through openings in a room at normal height is given by Vf: m_ air ¼ 0:5V f

ð6:2Þ

For a well-developed ventilation-controlled fire, the maximum heat release rate Q_ vent (in kW) can be estimated by: Q_ vent ¼ 3000 m_ air

ð6:3Þ

Q_ vent ¼ 1500 V f

ð6:4Þ

or

After breaking a glass balcony of 3 m wide and 2 m tall, Vf is 8.49 m5/2, giving Q_ vent up to 12.7 MW. The wind speed Vw of rooms at upper levels z taller than 300 m in a supertall building of height Ho can be very high. Breaking the window glass would result in a much higher air intake rate. Vw is related to z through parameters Vo, Ho and n as [49]:  n z Vw ¼ Vo ð6:5Þ Ho The volumetric air flow rate prompted by wind Fw (in m3s1) is calculated by:  0:5 Vw2 Fw ¼ Cd Av ΔCp 2

ð6:6Þ

Putting Eq. 6.5 into it gives:  Fw ¼ Cd Av

ΔCp 2

0:5 

 Vo n z H on

ð6:7Þ

The total maximum heat release rate Q_ T including wind effect [33] can be estimated by combining Eqs. 6.4 and 6.7: Q_ T ¼ 1500V f þ 3000ρair Fw

ð6:8Þ

6

A Discussion on Tall Building Fire Safety in the Asia-Oceania Regions

Strong wind gust of 10 ms1 at z ¼ 100 m gives (Vo/Hno ) of 1 s1. In the above example of breaking a big glass pane of 3 m by 2 m, taking the density of air as 1.1 kgm3, Q_ T is then 232 MW with a high value of n of 0.5 [33], about 18 times the heat release rate of 12.7 MW in a room fire at ground level without wind. Note that it will take a longer time [11] for the firemen to move up to the fire room located at higher level. The heat release rate of the fire can then be even bigger. Under such a big fire, air circulation [31, 50] resulting from the building groups under wind action would facilitate fire spread or would even generate fire whirls [34–37]. Further, clean refrigerants which are widely applied in new environmental-friendly air-conditioning systems can also be a possible ignition source [9, 51] in residential buildings, as some refrigerants are hydrocarbon propane. Fire safety provisions on air-conditioning systems with clean refrigerants should be enhanced at the moment. As pointed out earlier [51], it is too early to impose restrictions on clean refrigerants. The authorities might only be able to introduce some measures after more explosions induced by clean refrigerants occurred.

6.4

Residential Building Fire with an Open Kitchen

High FLD [16, 28, 32] is observed in small residential units in many dense cities of the Asia-Oceania regions such as Hong Kong. Many flats are subdivided [52] into smaller units as in Fig. 6.2, and such phenomenon has been in existence for several decades in Hong Kong. A small fire can lead to very serious consequences. Regular fire inspection schemes appear to be unsatisfactory as indicated in recent fires [7–9]. Water did not come out of the fire hydrant with adequate operating pressure, and the system was suspected to be short-circuited as reported [9]. Open kitchens [29, 53, 54] are very common in small units of tall residential buildings. It is very difficult to suppress a fire in an open kitchen in a small residential unit that has a FLD up to 1400 MJm2 [32]. However, these kitchens are not equipped with fire suppression systems such as sprinklers if these buildings were built before the implementation of the revised building fire safety code [18]. There has not been extensive fire research [18, 54] to work out the fire safety provisions for open kitchens. An IFW can be induced in a small flat fire with an open kitchen [35, 36]. As reported [55], even an electric induction cooker can ignite a pan of oil for preparing deep-fried food. These fire hazards suggest that sprinkler installation should be considered [56–60] in some residential buildings with special features such as open kitchens. At the moment, there is no requirement for the domestic buildings to install any automatic suppression system.

65

Fig. 6.2 A subdivided flat

Experiments [61] were carried out for evaluating the effectiveness and benefits of using automatic sprinkler system in domestic units. Full-scale burning tests were carried out in a burn room of FLD 1135 MJm2 [17], floor area 10 m2 and height 2.5 m on wood crib fires and furniture fires, with and without residential sprinkler setting following BS EN 12845 [60]. The development of fire and the associated phenomena inside a compact domestic building were studied. By analyzing the temperature profiles, water consumption and other relevant data, the effectiveness and benefits of using domestic sprinkler system in domestic units were evaluated. It was noted that damage to properties could be minimized as fire development was suppressed by sprinkler. Fire could be put out in a shorter time using less manpower and water when compared to tests without sprinkler protection. With sprinkler protection, only a small portion of the room furnishings and wood cribs was damaged. Smaller amounts of carbon monoxide and carbon dioxide were emitted during the combustion and less water was used for firefighting. If flashover occurred prior to fire service intervention, more combustibles would be burnt. A larger quantity of greenhouse gases would be emitted, and more burnt substances would be disposed of in the landfill. The use of sprinkler for protection of lives and properties was found to be effective and reliable in domestic units, which is also favourable to the environment. It can be concluded that using automatic sprinkler system may result in better fire safety for domestic buildings in protection of lives and properties. Further, indirect benefits include reduction in injuries and medical treatments of firefighters and in emission of greenhouse gases.

66

6.5

W.-K. Chow

Evacuation

There are over 200 buildings taller than 150 m in Hong Kong functioning as government agencies, commercial buildings, residential buildings and hotels [62]. Evacuation is the main concern in providing fire safety for tall buildings [63]. It is obvious that the evacuation situations would be very complicated and vary for different occupants as pointed out [64]. Different evacuation designs would affect the evacuation patterns and occupant load in the evacuation routes. Appropriate evacuation scheme must be designed to take occupants to safe places for tall buildings. Evacuation studies have been carried out for many years with focus on three parameters, namely, the number of occupants, evacuation plans and attitude of the owners and management. Personal protection equipment and emergency escape equipment for tall buildings proposed [65] are not regarded as fire protection systems in many places. Human factors should be considered but there are not yet systematic studies in the AsiaOceania regions. Note that human behaviour in evacuation will be different for different countries, cultures and styles of living. Two main evacuation designs [64] as in Fig. 6.3 have been proposed for tall commercial or residential buildings in Hong Kong. Design 1 requires separate evacuation routes for the upper portion and lower portion of the building as in Fig. 6.3a. Design 2 requires evacuation routes for the upper portion and lower portion linked by a transfer zone as in Fig. 6.3b. The staircases for the upper portion are separated from others and the evacuation routes are relatively simple in Design 1. The escape route is longer for people staying higher. The occupants on the upper floors can go down to the ground floor through a staircase isolated from the lower part without communication as shown in Fig. 6.3a. However, it would be more difficult to change the evacuation routes for people from the upper floors if some exits are blocked. The staircases for Design 2 are divided into two sections. The lower portion is usually a shopping mall, a club house or a public transport interchange. The upper portion is usually offices or apartments. In emergency evacuation, the occupants on the upper floors can discharge at a transfer level like the podium and join other people in the lower floors for subsequent evacuation. The total evacuation time [38] under agreed scenarios for all occupants to leave the building must be estimated. There are empirical equations [66, 67] available on calculating the evacuation time in buildings of different usage. Basically, these equations are on movement along a horizontal plane, movement through vertical paths and movement out of different openings such as doors. There are advanced studies such as using basic science like constructal theory to study evacuation [68]. However, simple hydraulic flow methods [69] can be applied to estimate the evacuation time te (in s)

Upper portion

Lower portion Isolated staircases for upper portion (a) Design 1

Upper portion

Transfer zone

Lower portion (b) Design 2

Fig. 6.3 Evacuation routes in an example supertall building (Pang et al. [64])

of tall buildings using the total evacuation flow capacity F (in number of persons/s), which sums up all staircases: te ¼

NT F

ð6:9Þ

Parameters for F in the flow Eq. 6.9 have to be obtained empirically. For large occupant loadings, travel time from a certain position to the exit will be much shorter than the ‘waiting time’. There are blocking effects while moving through protected corridors, lobbies, doors and staircases. Note that compartmentation requires fire doors to be closed normally. There might be two closed doors connecting some spaces in compliance to PC. A common example is the ‘protected lobbies’ linking the lift hall and the usable areas in Hong Kong [17, 18]. This will result in longer evacuation times, especially under crowded conditions with long ‘waiting time’. Some assumptions and parameters might not be applicable when the moving occupants are exposed to heat and smoke. People in different countries with different cultures might have different responses and travel speeds. However, te estimated from Eq. 6.9 is still useful while carrying out fire

6

A Discussion on Tall Building Fire Safety in the Asia-Oceania Regions

hazard assessment in the performance-based approach, either PBD or FEA, or in implementing EPBFC. Different evacuation designs can be compared in PBD-FEA projects. In most cases, fire safety provisions such as means of escape are compared with the requirements specified in the existing codes. It is difficult to set up PC quickly. Time is needed to work out the requirements through scientific investigations and to complete the legislative process. An immediate action is to study the evacuation pattern in these tall buildings and work out appropriate fire safety management under the existing protection systems, both passive and active. Refuge floors are required for tall buildings [17, 18] in many places including Hong Kong. Refuge floors are intended to serve multipurposes in fire safety. They can serve: 1. 2. 3. 4.

As a safe place for gathering in a very tall building [69] As a ‘command’ point in firefighting To stop or retard the upward spreading of flames To provide an area in which to change the vertical lift shafts to reduce the stack pressure on smoke movement, further facilitating the lift design in supertall buildings 5. To reduce the wind loading onto the building 6. To house certain special applications such as fire service plant rooms and new design features, e.g. communal sky gardens [70] However, fire hazards in such ‘sky gardens’ in supertall residential buildings should be assessed carefully. Numerical simulations suggested that refuge floors are safe under a normal fire of 2 MW and relatively safe for fires of 25 MW. However, for huge fires of magnitude comparable to those in the World Trade Center incident [71], occupants would be reluctant to stay on the refuge floor. It would be desirable to carry out some full-scale burning tests [72] to demonstrate whether protection is adequate for specific buildings and to educate occupants that it is safer to gather on designated refuge floors in case of fire. It should be noted that refuge floors and staged evacuation strategy [73] are not welcomed as nobody would like to wait at a refuge floor for a long time for rescue following the experience of the World Trade Center tragedy [71]. In this connection, education and fire drills should be regarded as important elements in fire safety.

6.6

Two Alternative Approaches

Two approaches on elevator evacuation and evacuation through skybridge at height can be considered. The evacuation process in tall buildings can be faster by using elevators [74–78]. The evacuation time at the Petronas Tower [12, 13]

67

could be reduced from 2 to 0.5 h if elevators in good order were used. However, elevators are not expected to be used [79] by occupants in fire emergencies. Fireman’s lifts [80, 81] are only expected to operate for a short time interval and are not designated as evacuation elevators. The biggest concern in using elevators in a fire emergency at the moment is the performance of the elevator system during a fire. The lack of fire protection of the lift shafts and the adjacent lobbies where occupants are waiting to use the elevators, the possibility of water damage, the thermal response of the elevator car and the cutting off of power supplies are the key points. There is no PC governing the use of elevators as an evacuation vehicle in tall buildings in many places such as Hong Kong. As pointed above, elevators are not yet designed for use in the case of fire. There might be fatal delivery of an elevator to the fire floor when a passenger pushes the car button for the fire floor, or when a corridor button is pushed on the fire floor. In addition, heat may melt or deform the corridor push button or the wiring at the fire floor, or may damage the control system of the elevator, leading to malfunctions. The office fire in Sa˜o Paulo on 1 February 1974 demonstrated that half of the 600 occupants were evacuated by elevators. That case was pointed out to be successful only by chance by Sumka [80]. In that incident, the elevators were set to run in the ‘express’ mode, and luckily the power supply to the elevators was also not affected. As only one successful case under a big fire was reported, risk assessment [82] should be considered for further analysis. Although elevators are included in the emergency evacuation strategy for several supertall buildings and even in a subway station 40 m deep underground [83] in Hong Kong, only part of the research behind was debated openly in local conferences [84]. Most of the earlier studies of elevator evacuation reported [74–78] were on human behaviour, evacuation modelling with elevators, smoke control in the lift shaft and the protected lobby for tall buildings of normal height. These works might not be applicable for supertall buildings. Designs of fire safety provisions should address specifically the concerns for evacuation using elevator systems. Note that for Taipei 101, it is mandated that elevator evacuation is not to be implemented when there is a fire [79]. It is observed that using elevators [12, 13] can accelerate evacuation in supertall buildings under fire. However, there is still no fire-safe elevator system at the moment [80]. Firesafe elevator systems must be developed for emergency evacuation of supertall buildings. Fire and smoke spread to the lift shafts and adjacent lobbies should be avoided by appropriate fire protection systems. An elevator system includes at least the lift shafts, elevator cars including their control and electronics systems and the protected lobbies

68

where occupants wait for the elevators. Apart from the problems of maintaining good order of occupants going to and staying at the ‘refuge place’ as claimed [83] during evacuation, there are many problems such as limited fire resistance and smoke spreading through leakages to the lift shafts, protection of the adjacent lobbies with waiting occupants, thermal insulation of the elevator car itself and the possibility of breaking the cables, water wetting the lift control electronics and thus causing electrical faults and steam generation and its hazardous effects on occupants. Note that some fire-resistant glass might generate smoke in case of fire [85]. Preliminary investigation in some supertall buildings suggested that [79] elevator evacuation should not be implemented under fire. This suggestion is different from other projects [84] with elevator proposed for evacuation of supertall buildings. Elevator evacuation in tall buildings is under careful review by the fire authority in Hong Kong. At least three key areas [86] should be satisfactorily addressed before elevators can be used for evacuation in a fire. These are the prevention of fire spread into the lift shaft, protection of the adjacent lobbies and prevention of water damage due to sprinkler discharge. Clusters of tall buildings have been constructed in public residential estates and commercial districts in big cities of the Asia-Oceania regions [87]. A new feature of these clusters is commonly observed. Single buildings are designed to be connected to each other through different linkage structures. One of the commonly used linking structures is the skybridge. From a structural point of view, skybridge can provide lateral reinforcement to the building towers, as it functions like a horizontal diaphragm and like a floor slab. Note that the floor systems made of slabs acting as horizontal diaphragms with very large stiffness in the horizontal direction can help to distribute the lateral load into the vertical structure elements such as walls and columns apart from resisting the vertical load. Skybridges at height are commonly proposed for linking tall buildings in clusters of public residential estates and commercial districts in big cities (Fig. 6.4). This would provide an additional evacuation path in case of fire, enhance structural stability under wind and maximize the commercial floor space. In addition, it can serve as a pedestrian crossing, a penthouse or other purposes, etc. It was proposed that skybridges would give safer tall buildings by reducing the total evacuation time [87]. However, wind effect [88] would pose problems on skybridges at height in tall buildings. Wind loading and wind-induced vibrations should be carefully taken into account for such skybridges. The challenges, especially the wind load problem in constructing skybridges to link buildings at height have been discussed in the literature [88].

W.-K. Chow

Fig. 6.4 Skybridge at height at University of Nottingham, UK

6.7

Internal Fire Whirls

Fire whirls are common phenomena in big forest fires [34]. The buoyancy-driven fire plume entrains ambient fluid when the bulk of the heated gases ascends. Vorticity associated with external effects such as wind shear might be concentrated and create a vortex core along the axis of the plume. All these forces will generate a whirling fire and flame elongation due to the decrease in turbulence intensity, air entrainment and the mixing of air with fuel prompted by centrifugal forces. The burning rate then increases substantially in comparison with the non-whirling case. For fires in vertical shafts [37] of a tall building, the flame surrounded by solid walls with an open ceiling and channels of flow caused by air entrainment through side openings can induce rotational airflow. This together with buoyancy would ‘push’ the hot gases up, which would mix with the air. A combustible mixture will be created and stretch the flame up. With appropriate sidewall ventilation, an IFW can be induced in a scale model easily. The buoyancy of hot fuel gases and swirl velocities are two key factors behind the generation of fire whirls. IFW can give a very hazardous environment in the vertical shaft due to the extended flame length and fast burning rate. In a small fire, the IFW is usually triggered in a vertical shaft with an open roof. Ventilation provision at the sidewall has been identified [37] as a key condition for the onset of an IFW. This is supported by both scale model experiments and full-scale tests that were conducted in a 12-m-high typical square vertical shaft structure. However, it is still not mandatory to inspect the IFW phenomena in the fire hazard assessment [37] of vertical shafts in very tall buildings. Experiments were carried out [35] to investigate the conditions for the onset of buoyant whirling flame. A test

6

A Discussion on Tall Building Fire Safety in the Asia-Oceania Regions

69

Fig. 6.5 Internal fire whirl in a room from Gao and Chow [36]. (a) Room geometry, (b) Velocity vectors

room of length 2.77 m with a door and a ceiling vent was constructed. A 0.6-m-diameter pool fire with commercial fuel gas was placed at the room centre. For scenarios without a ceiling vent and with the fire located at a corner, only small-scale whirls would be created by a small fire with low heat release rate. This point can be studied easily by investigating the scenarios leading to the onset of a fire whirl by computational fluid dynamics (CFD). Room-scale whirling flames were observed. A similar room geometry of length 3.6 m, width 2.4 m and height 2.4 m as in Fig. 6.5a was considered [36]. Effects of ventilation on the whirling speed and whirling paths were simulated. Numerical experiments were carried out under the two ventilation conditions in [35, 36]. The numerical input data used was based on the experimental studies reported [35]. The CFD software PHOENICS v3.5 [89] was used with a Pentium personal computer. Velocity vectors at different heights y at 0.2 m, 0.6 m, 1.0 m, 1.6 m, 2.2 m and 2.8 m for the bigger fire F1 at 5 s, 60 s, 120 s, 180 s and 240 s are plotted in Fig. 6.5b. Stronger whirling flow near the flame bottom was formed at 60 s as in Fig. 6.5b. Whirling motion was not induced about the flame axis at 120 s. Based on CFD simulation, an IFW appears as in the figure. It is interesting to note that an IFW can be induced in an open kitchen with appropriate ventilation provision [35, 36], and

the fire safety problem of open kitchens should not be overlooked.

6.8

Conclusion

Fire safety in tall buildings should be watched carefully in the Asia-Oceania regions, particularly for residential buildings that are over 300 m tall with openable windows. Large quantities of combustibles are present, and hence high FLD is found in these tall buildings as reported [29, 32, 48]. When a fire occurs, the high ambient wind speeds through open windows or green balconies at the upper levels of tall buildings can supply more air to burn up all these combustibles. Stack effect [30, 33] is found to be very important in places with extreme difference between the indoor and outdoor temperatures in winter, such as Seoul in Korea. The stack pressure can be high in very tall staircases and vertical shafts, and smoke and even fire can spread to all levels. Wind action is another critical force in inducing air movement through openings. The natural driving forces of air movement including both stack effect and wind action for supertall buildings in these areas should be considered carefully. The big post-flashover fire would emit very high radiative heat fluxes to consume more

70

combustibles. As a new feature in buildings, open kitchens could give hazardous fire scenarios. Installation of sprinkler is recommended in residential buildings with open kitchens. This point has been included in the updated fire code with PC and FEA [18]. An IFW can be created in vertical shafts [34–37] to give faster fire spread rate. Air, fire and smoke movements do not only follow the wind direction, but in all other directions following air movement in areas with many buildings [50]. Fire can spread quickly to neighbouring buildings. Evacuation is another concern as the Available Safe Egress Time (ASET) may not be sufficient for big fires. High occupant loading will lengthen the Required Safe Egress Time (RSET) so that ASET is shorter than RSET. Evacuation strategy with elevator and skybridge at height deserves careful investigation and research. The current PC on fire safety has not demonstrated capabilities to protect tall residential buildings against accidental fires, at least in Hong Kong. A performance-based approach, the PBD-FEA, has to be implemented [1, 16, 28] in determining the fire safety provisions. The heat release rate of the design fire is the first parameter in fire hazard assessment. The heat release rate of fires in supertall residential buildings with open windows under the stack effect and wind action can be very high. New approaches [47] should be applied to determine the design fire for new projects. However, fire safety provisions determined by PBD-FEA in existing buildings should be upgraded by referencing to these new approaches. Acknowledgement The work described in this paper was supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China, for the project “Wind action on supertall building fires and spread to adjacent areas” (PolyU 5135/12E) with account number B-Q31U.

References 1. Chow WK (2013) Experience on implementing performance-based design in Hong Kong, welcome speech, The 9th Asia-Oceania symposium on fire science and technology, 17–20 October 2012, Hefei, China; Appeared in Procedia Engineering, 62, pp 28–35 2. Council on Tall Buildings and Urban Habitat (CTBUH) (2011) Tall buildings in numbers. CTBUH J Int J Tall Build Urban Habitat, (II):46–47 3. Society of Fire Protection Engineers (2012) Guidelines for designing fire safety in very tall buildings – public review draft. Society of Fire Protection Engineers, Washington, DC 4. South China Morning Post (1996) Hong Kong, 21 Nov 1996 5. Chow WK (1998) Numerical studies on recent large high-rise building fire. ASCE J Archit Eng 4(2):65–74 6. South China Morning Post (1997) Mei Foo Sun Chuen fire, 17 Apr 1997

W.-K. Chow 7. South China Morning Post (2013) Immobile wife dies in fire after husband flees flat, Hong Kong. 16 Feb 2013 8. South China Morning Post (2013) Teen fighting for life after blaze guts 10th floor flat, Hong Kong. 20 Mar 2013 9. South China Morning Post (2013) Seven injured in Ap Lei Chau apartment blaze, Hong Kong. 14 Apr 2013 10. McElvaney J (2013) First responder challenges in very tall buildings. Fire Protect Eng Second Q (58):62–72 (http://c.ymcdn. com/sites/www.sfpe.org/resource/resmgr/FPE_Magazine_Archives/ 2013/2013_Q2.pdf) 11. Chow WK, Fong NK, Lui CH, Tam TK, Yue TK (2013) Fire safety strategies for supertall buildings in Hong Kong. CTBUH J – Int J Tall Build Urban Habitat (1):26–31 (http://global.ctbuh.org/ resources/papers/download/264-fire-safety-strategies-for-supertallbuildings-in-hongkong.pdf) 12. ABC News (Australian Broadcasting Corporation) (2005) Fire erupts at Malaysia’s twin towers (4 November 2005). http://www.abc.net. au/news/2005-11-05/fire-erupts-at-malaysias-twin-towers/2139184. Accessed 28 Jun 2013 13. Ding RJ (2002) Petronas at Malaysia. Fire Tech Prot Inf 4 (1):59–62, Retrieved from http://www.cnki.net (in Chinese) 14. Chow WK (2013) The four fire code systems, department of building services engineering, The Hong Kong Polytechnic University, Hong Kong, February 2013. Available at: http://www.bse.polyu. edu.hk/researchCentre/Fire_Engineering/Hot_Issues.html 15. Chow WK, Dong X (2014) Legislation, codes of practice and standards in Hong Kong and Mainland China. In Fire from first principles, a design guide to international building fire safety. P. Stollard, Editor. 4th Edition, Routledge, Chapter 10. New York 16. Chow WK (2014) Performance-based approach to determining fire safety provisions for new buildings in the Asia-Oceania regions, building and environment – 50-year golden issue, submitted for consideration to publish 17. Fire Services Department (2012) Codes of practice for minimum fire service installations and equipment and inspection testing and maintenance of installation and equipment. Fire Services Department, Hong Kong Special Administration Region, China 18. Buildings Department (2011) Code of practice for fire safety in buildings 2011, buildings department, The Hong Kong Special Administrative Region, Draft version released for public consultation in September 2011, Implemented April 2012 19. Hadjisophocleous GV, Benichou N, Tamim AS (1998) Literature review of performance-based fire codes and design environment. J Fire Prot Eng 9(1):12–40 20. Australian Building Codes Board (ABCB) (2005) International fire engineering guidelines. ABCB, Canberra 21. Fleischmann C (2012) Performance-based design in New Zealand and the new verification method, CPD lecture organized by Research Centre for Fire Engineering, Department of Building Services Engineering, The Hong Kong Polytechnic University, Hong Kong, 30 April and 2 May 2012 22. New Zealand Government (2013) Commentary for building code clauses C1–C6 and verification method C/VM2, Ministry of Business, Innovation and Employment, New Zealand Government, Dec 2013 23. New Zealand Government (2014) C/VM2 verification method: framework for fire safety design – for new zealand building code clauses C1–C6 protection from fire. Ministry of Business, Innovation and Employment, New Zealand Government, July 2014 24. Meacham BJ (1999) International experience in the development and use of performance-based fire safety design methods: evolution, current situation, and thoughts for the future, invited paper. In Proceedings of the international association for fire safety science, 6th international symposium, poitiers, France, July 1999, pp 59–76 25. Society of Fire Protection Engineers (2007) SFPE engineering guide to performance-based fire protection. SFPE, Bethesda

6

A Discussion on Tall Building Fire Safety in the Asia-Oceania Regions

26. Society of Fire Protection Engineers (2008) SFPE handbook of fire protection engineering, 4th ed. National Fire Protection Association, Quincy; Society of Fire Protection Engineers, Bethesda 27. CIBSE (2010) Guide E: fire safety engineering. The Chartered Institution of Building Services Engineers, London 28. Chow WK (2011) Performance-based design on fire safety provisions in the far east. Keynote presentation, 2011 SFPE annual meeting: professional development conference and exposition, engineering technology conference, 24–25 Oct 2011, Portland 29. Chow WK (2007) Fire safety of supertall buildings and necessity of upgrading active protection systems, international colloquium on fire science and technology, Seoul, Korea, 25 Oct 2007 30. Chow WK, Chan MY (2014) Fire safety concern for supertall buildings, ASHRAE 2014 Winter Conference, 18–22 Jan 2014, New York – Paper presented 31. Planning Department (2012) Urban climatic map and standards for wind environment – feasibility study executive summary. Planning Department, Hong Kong 32. Ove Arup Hong Kong (2010) Research in East Asia, Arup research brochure, Dec 2010 33. Chow CL, Chow WK (2010) Heat release rate of accidental fire in a supertall building residential flat. Build Environ 45(7):1632–1640 34. Meroney RN (2007) Fires in porous media: natural and urban canopies, In: Gayev YA, Hunt JCR (eds) Flow and transport processes with complex obstructions. Springer, Chapter 8, pp 271–310 35. Snegirev Y, Marsden JA, Francis J (2004) Numerical studies and experimental observations of whirling flames. Int J Heat Mass Transf 47(12–13):2523–2539 36. Gao Y, Chow WK (2010–201) Onsetting internal fire whirls in a room with ceiling vents. J Appl Fire Sci 20(2): 149–165 37. Chow WK, Han SS (2009) Experimental investigation on onsetting internal fire whirls in a vertical shaft. J Fire Sci 27(6):529–543 38. Chow WK (2004–2005) Evacuation in a supertall residential complex, J Appl Fire Sci 13(4):291–300 39. International Organization for Standardization (2009) ISO 13387 (1999), application of fire performance concepts to design objectives, its replacement, ISO 23932 (2009), fire safety engineering – general principles, and the associated standards and technical reports. International Organization for Standardization, Geneva 40. Alvarez A, Meacham BJ, Dembsey NA, Thomas JR (2013) 20 years of performance-based fire protection design: challenges faced and a look ahead. J Fire Prot Eng. doi:10.1177/ 1042391513484911 41. Torero J (2012) Structures in fire or fires in structures: what do we need to know to achieve innovative fire safety, international fire conference & exhibition Malaysia (IFCEM 2012), Kuala Lumpur, 20–21 Nov 2012 42. University of Science and Technology of China (2011–2015) The national 973 project on high-rise building fire safety, University of Science and Technology of China, Hefei 43. Chow WK (2009) Performance evaluation of atrium smoke exhaust systems with hot smoke tests. At website: http://www.scitopics. com/Performance_Evaluation_of_Atrium_Smoke_Exhaust_ Systems_with_Hot_Smoke_Tests.html 44. Chow WK (2005) Fire hazard assessment of combustibles in big terminals. Int J Risk Assess Manag 5(1):66–75 45. Chow WK (2005) Assessment on fire hazard in a small news agent in transport terminal halls. ASCE J Archit Eng 11(1):35–38 46. Arvidson M (2005) Potato crisps and cheese nibbles burn fiercely. Brand Posten 32:10–11 47. Borg A, Nja O, Torero J (2015) A framework for selecting design fires in performance based fire safety engineering. Fire Technol. doi:10.1007/s10694-014-0454-x

71

48. Chow WK (2010) Fuel load and peak heat release rate correlations in post-flashover room fires. Heat Trans Eng 31(3):250–254 49. Lamming SD, Salmon JR (1998) Wind data for design of smoke control systems. ASHRAE Trans 104(1):742–751 50. Chow WK (2001) Numerical studies of air flow induced by mechanical ventilation and air-conditioning (MVAC) systems. Appl Energy 68(2):135–159 51. Chow WK (2013) Lesson learnt from a recent incident in Ma On Shan: any explosion risk for environmentally friendly refrigerants? Department of Building Services Engineering, The Hong Kong Polytechnic University, Hong Kong, February 2013. Available at: http://www.bse.polyu.edu.hk/researchCentre/Fire_Engineering/ Hot_Issues.html 52. Buildings Department (2013) It’s dangerous living in subdivided flats in industrial buildings. Available: http://www.bd.gov.hk/ english/documents/pamphlet/SFIBe.pdf. Last accessed 23 Dec 2013 53. Chow WK (2011) Open kitchen fires in tall residential buildings, CPD lecture organized by Research Centre for Fire Engineering, Department of Building Services Engineering, The Hong Kong Polytechnic University, Hong Kong, 5 Mar 2011 54. Fire Engineering Discussion Forum – open kitchen, co-organized by fire division, The Hong Kong Institution of Engineers and Department of Civil and Environmental Engineering, The Hong Kong Polytechnic University, 23 Nov 2012 55. Wong AKK, Fong NK (2012) Experimental study of induction cooker fire hazard, 2012 International symposium on fire science and engineering, Guangzhou, 6–7 Dec 2012 56. Chow WK, Fong NK, Tam TK, Woo YK (2013 Mar) News from The Hong Kong Polytechnic University. In: Rein G (ed) International Association for Fire Safety Science Newsletter. No. 34, pp 12–13 57. Coleman RJ (1991) Residential sprinkler systems: protecting life and property. National Fire Protection Association, Quincy 58. Dubay C (2007) Automatic sprinkler systems for residential occupancies handbook. National Fire Protection Association, Quincy 59. British Standards Institution (2005) BS 9251, sprinkler systems for residential and domestic occupancies – code of practice, British Standards Institution, UK 60. British Standards Institution (2009) BS EN 12845, fixed firefighting systems – automatic sprinkler systems – design, installation and maintenance, British Standards Institution, UK 61. Woo YK, Tam TK, Chow WK, Fong NK (2015) Domestic sprinkler: it’s time to consider mandatory requirement in Hong Kong, The 10th Asia-Oceania symposium on fire science and technology, submitted for consideration to present 62. Emporis – building data and construction projects worldwide. http://www.emporis.com/en/. Accessed 26 Dec 2007 63. National Fire Protection Association (2003) NFPA 101 life safety code. National Fire Protection Association, Quincy 64. Pang ECL, Chow WK, Han SS (2008) Evacuation strategy for supertall buildings, The first international conference on building energy and environment (COBEE), 13–16 July 2008, Dalian, China, vol 1–3, pp 2094–2100 – Poster presentation 65. Clothier D (2005) EGRESS: personal building evacuation system, Elevator World, March, pp 63–66 66. British Standards Institute (2001) BS 7974 application of fire safety engineering principles to the design of buildings – code of practice, British Standards Institute, UK 67. Nelson HE, Maclenna HA (1995) Emergency movement. In SFPE handbook of fire protection engineering, 2nd ed., National Fire Protection Association, Quincy; Society of Fire Protection Engineers, Boston. Section 3/Chapter 14, pp 3-286–3-295

72 68. Lui CH, Fong NK, Lorente S, Bejan A, Chow WK (2015) Constructal design of evacuation from a three-dimensional living space. Physica A Stat Mech Appl 422:47–57 69. Chow CL, Chow WK (2009) Fire safety aspects of refuge floors in supertall buildings with computational fluid dynamics. J Civ Eng Manag 15(3):225–236 70. Niu JL, Tang YM, Mak CM (2005) The application of computational fluid dynamics to the assessment of green features in buildings, part 2 communal sky gardens. Archit Sci Rev 48:337–344 71. National Institute of Standards and Technology (2005) NIST NCSTAR 1, Federal building and fire investigation of the World Trade Center disaster – final report on the collapse of the World Trade Center Tower, technology administration, US Department of Commerce, Sept 2005 72. Chow WK (2008) Necessity of testing combustibles under welldeveloped fires. J Fire Sci 26(4):311–329 73. Klote JH (1993) Design of smoke control systems for areas of refuge, NISTIR 5132, National institute of standards and technology, US Department of Commerce, Mar 1993 74. Klote JH, Levin BM, Groner NE (1995) Emergency elevator evacuation systems, elevators, fire and accessibility, Proceedings, 2nd symposium, American Society of Mechanical Engineers, April 19–21, 1995, Baltimore, pp 131–149 75. Klote JH (2002) Elevators as a means of fire escape, National Bureau of Standards, NBSIR 82–2507; NIST SP 983, May 2002 76. Sekizawa, Nakahama S, Notake H, Ebihara M, Ikehata Y (2004) Feasibility study of use of elevators in fire evacuation in a high-rise building, 6th Asia-Oceania symposium on fire science and technology, 17–20 Mar 2004, Daegu, pp 235–246 77. Cook D (2004) Preplanning for elevator use during major fires by following seismic events, Elevator World, pp 71–75 78. Proulx G (2005) Evacuation by elevators: who goes first? Elevator World, pp 55–59

W.-K. Chow 79. Hsiung KH, Wen WJ, Chien SW, Shih BJ Hsiao WH (2006) A research of the elevator evacuation performance for Taipei 101 Financial Center, 6th international conference on performance-based codes and fire safety design methods, 14–16 Jun 2006, Tokyo 80. Sumka EH (1987) Presently, elevators are not safe in fire emergencies. ASHRAE Trans 93(Part 2):2229–2234 81. ASME/ANSI A17.1 (1987) Safety code for elevators and escalators 82. Thomas I, Bruck D (2008) Case study on tall apartment building, The 7th international conference on performance-based codes and fire safety design methods, 16–18 Apr 2008, Auckland 83. The Standard (2014) University station will be learning experience, 20 Nov 2014 84. Guo DG, Wong K, Kang L, Shi BB, Luo MC (2004) Lift evacuation of ultra-high rise building, In: Proceedings of the fire conference 2004 – total fire safety concept, 6–7 Dec 2004, Hong Kong, vol 1, pp 151–158 85. Wu M, Chow WK, Ni X (2015) Characterization and thermal degradation of protective layers in high rating fire-resistant glass. Fire Mater 39(1):26–40 86. Chow WK, Wood A, Li J (2013) Reduction in evacuation time for tall buildings through the use of skybridges. J Archit Plan Res 30(2):146–166 87. Cai N (2014) Fire safe elevator used for evacuation in supertall buildings, PhD thesis, Department of Building Services Engineering, The Hong Kong Polytechnic University 88. Chow WK, Li YF, Wood A (2008) Wind effects on skybridges at height in tall buildings, The 4th international conference on advances in wind and structures (AWAS’08), 29–31 May 2008, Jeju 89. PHOENICS Version (2007) Concentration heat and momentum limited (CHAM Ltd.), Wimbledon, Aug 2007

7

Toward a Risk-Informed Performance-Based Approach for Post-Earthquake Fire Protection Design of Buildings Brian J. Meacham

Abstract

Post-earthquake fires have extracted great tolls on society. While this has been acknowledged for some time and significant effort has been expended to develop tools to simulate post-earthquake fire spread at an urban scale, to develop performance-based seismic design methods for buildings, and to develop performance-based fire design methods for buildings, there has been little research dedicated to the development of performancebased design approaches which integrate earthquake and fire design of structures. In this paper, an overview of the post-earthquake fire problem and of post-earthquake building performance is provided, performance-based seismic and fire design approaches are briefly introduced, and developments toward risk-informed performance-based approach to seismic and fire design for buildings in earthquake-prone areas are presented. Keywords

Earthquake
Fire
Risk
Performance-based design

7.1

The Post-Earthquake Fire Problem

Earthquakes can cause significant damage to buildings, neighborhoods, cities, and regions. Unfortunately, postearthquake fire can be worse. The combination of earthquake-damaged buildings, post-earthquake ignitions, and resulting single-building, multi-building, and largearea fires can be devastating. In the case of the 1906 San Francisco earthquake (Fig. 7.1), it has been estimated that post-earthquake fire resulted in more damage than the earthquake itself, with a 3-day conflagration spread over an area of more than 4 square miles [2, 3]. According to Geschwind [3]: “Over the next 2.5 days (following the earthquake), the fire laid waste to 4 square miles (10.4 km2), including the central business district and numerous densely populated residential sections.

B.J. Meacham (*) Department of Fire Protection Engineering, Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609, USA e-mail: [email protected]

The total number of lives claimed by the earthquake and fire is unknown. At the time, about 700 deaths were attributed to the disaster, but more recent estimate placed the number at several thousand. The earthquake and fire rendered more than 200,000 people homeless, and property losses estimated at $500 Million in 1906 dollars ($13 Billion in 2015 dollars)” (note: comments in parentheses added for clarity). Likewise, the 1923 Great Kanto (Tokyo) earthquake and resulting fire led to widespread damage and loss of more than 140,000 lives [4, 5]. As explained by Usami [4]: “Immediately after the earthquake, fires arose at 163 points in Tokyo and about 3800 ha were burnt to ashes. About 316,000 houses, 70 % of all the houses in Tokyo, vanished in the fires. In Yokohama, fires arose at about 60 points and burnt about 950 ha and 60,000 houses, that is, 60 % were destroyed by fires. Sum of the dead and the missing was 142,807, about 80–90 % of whom were killed by fires.” Even in recent years, with advancements in building technology, firefighting technology, and increased resiliency of lifelines, more than 110 post-earthquake ignitions were reported as a result of the 1994 Northridge, California, USA,

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_7

73

74

B.J. Meacham “Southern California is unfortunately well situated for major fires to be generated following earthquakes. The number of ignitions that will create fires large enough to call the fire department can be extrapolated from previous earthquakes and depends upon the number of households at different levels of seismic shaking. This leads to an estimate of 1,600 ignitions of which 1,200 will be too large to be controlled by one fire engine company. . . The final level of fire damage is difficult to assess because it depends upon several unpredictable factors, especially the degree to which fires spread when the fire protection services lose water and are overwhelmed. We use the minimum value from the fire estimates at $40 billion in damage to buildings and $25 billion in damage to building contents” (italics added).

Fig. 7.1 Photograph by Arnold Genthe shows Sacramento Street and approaching fire [1]

earthquake [6, 7], and as many as 293 ignitions were reported and wide-scale fire damage occurred following the 1995 Hyogoken-Nanbu (Great Hanshin-Awaji) earthquake and fire in Kobe, Japan [8–10], where more than 5000 people perished. As noted by Chung [8]: “The January 17, 1995 Hyogoken-Nanbu earthquake of magnitude 7.2 in JMA scale (Mw ¼ 6.9), which struck Kobe, Japan and its surrounding area was the most severe earthquake to affect that region this century. The earthquake resulted in more than 6000 deaths and over 30,000 injuries. Fires following the earthquake incinerated the equivalent of 70 U.S. city blocks. They together destroyed over 150,000 buildings and left about 300,000 people homeless. The economic loss as a result of this earthquake is estimated to reach $200 billion.” More recently, the March 11, 2011, “off the Pacific coast of Tohoku Earthquake” (Tohoku earthquake or Great EastJapan earthquake) and tsunami resulted in the collapse of more than 128,000 houses and 96,000 other structures, missing persons and fatalities of more than 21,000, and as many as 330 ignitions, of which more than 190 were related to the earthquake and not the tsunami [11, 12]. As reported by Sekizawa and Sasaki [12]: “For the earthquake-induced fires, the majority of fire causes involved appliances with a heating source, such as space heaters, cooking stoves, and candles that overturned and fell by the shock of the earthquake. These cases accounted for 45.5 %, or nearly half of all causes. . . On the other hand, 38 cases (26.2 %) were caused by destruction and breakage of gas piping or electric wiring, accounting for about a quarter of all fires.” While the USA has not experienced a large earthquake in recent years, the potential is ever present, and concerns for significant post-earthquake fire persist as well. Recently, the extent of potential damage from fires following a magnitude 7.8 earthquake in Southern California was highlighted as a significant concern in the US Geological Survey (USGS) report on the 2008 ShakeOut Scenario [13] as well as by Scawthorn [14]. As stated in the USGS report [13]:

The challenge, as discussed above, is that numerous factors contribute to the potential for significant postearthquake fire damage: earthquake-induced damage to the buildings, post-earthquake ignitions, earthquake-damaged lifelines, and limits on emergency personnel and apparatus. This was highlighted in a recent review of post-earthquake fire ignitions by Lee et al. [15], where the authors note “fire following earthquake can be extremely destructive because there may be many simultaneous ignitions at the same time that damaged water supply systems impair fire suppression capabilities,. . . passive fire defenses are degraded (e.g., breached firewalls). . .” Although considerable research has been undertaken with respect to mechanisms to reduce structural damage to earthquakes, and various efforts have been conducted to enhance the prediction of post-earthquake fire ignitions (e.g., see [15–18]), there has been almost no research related to understanding and quantifying the performance of passive and active building fire safety systems during postearthquake fire – key components in any strategy to reduce the potential for large-scale post-earthquake fires, and essential data for any post-earthquake fire ignition model that operates at an individual building level, considering building and system performance [16], or which combines statistical analysis techniques with fire effects modeling to predict building-level fire spread potential [15, 17].

7.2

The Post-Earthquake Building Fire Performance Problem

Regardless of whether earthquake-induced ignitions occur, earthquakes can cause significant damage to key building fire and life safety components, both passive and active, such as fac¸ade components (glazing, curtain wall systems, and other exterior systems), interior partitions, ceiling systems, access and egress components (doors, stairs, lifts, and so forth), ventilation (including smoke control) systems, detection and alarm systems, suppression systems, lighting systems, electrical systems, and more. This is in addition to

7

Toward a Risk-Informed Performance-Based Approach for Post-Earthquake Fire. . .

causing damage to critical lifelines, such as water supply systems (for firefighting, as well as drinking water and sanitary water systems), fuel storage and distribution systems (fuel tanks, fuel distribution pipe networks), and roadways (for firefighter and emergency responder access). In the Northridge, CA, USA, and Hyogoken-Nanbu, Japan, earthquakes, for example, a large percentage of building fire protection systems experience failures, and the need to better understand and quantify the problem was reported [8–10, 19–23]. Numerous building fire sprinkler systems experienced failures, with damage reported in some cases of more than 40 % of the sprinkler systems and more than 30 % of the fire doors [23, 24]. Other fire protection systems also experienced significant damage, including passive systems [19, 20, 23, 24]. These data are consistent with other reports, including a Marine and Fire Insurance Association of Japan study on the reliability of fire protection systems in earthquakes, which reported that 34 % of sprinkler systems were damaged in the 1993 Kushiro-Oki earthquake and 41 % of sprinkler systems being damaged in the 1994 Sanriku-Haruka-Oki earthquake [24]. Recently, access and egress components, such as stairs, have been shown to perform poorly in some earthquakes. For example, precast stair units collapsed in at least four multistory buildings (e.g., see Fig. 7.2) and were severely damaged in several other cases, as a result of the February 22, 2011, Christchurch earthquake [25]. To a lesser degree, stair damage was also observed in steel frame structures as well [26]. Likewise, interior wall and ceiling systems [27] and exterior fac¸ade systems [28], which can protect against interior and exterior fire spread within buildings, were shown as being prone to significant damage, as illustrated in Fig. 7.3. Fac¸ade damage in particular was observed for a wide range of types and ages of buildings and fac¸ade systems [28]. In fact, nonstructural building systems, including ceiling, fac¸ade, and fire protection systems, are reported as contributing a significant percentage of the NZ$16 billion loss in the Christchurch earthquakes [28], a fact that unfortunately has been demonstrated in several other earthquake events as well [29, 30]. That is not to say that some aspects of post-earthquake performance of fire protection systems has not been studied but that it has not been extensive and has often been contained within broader research programs on nonstructural systems performance. For example, McMullin and Merrick [31] have studied damage thresholds for gypsum wallboard partition walls. In this particular series of experiments, eight modes of damage were observed, from cracking to buckling of large areas of the specimen wallboard panels. While nominally conducted to help develop fragility curves for damage estimates, this type of data is also helpful for assessing post-earthquake fire performance.

75

Fig. 7.2 Stair collapse in 2011 Christchurch earthquake [25]

Fig. 7.3 Fac¸ade damage in 2011 Christchurch earthquake [28]

However, there was no fire testing associated with this effort: a step which would have yielded helpful data for fire performance assessment. Similar studies have been conducted on the seismic performance on nonstructural systems, including wall assemblies, ceiling assemblies, and fire sprinkler systems and components [e.g., see 32–40]. Like the McMullin and Merrick experiments, there was no fire testing involved. However, many of these experiments have yielded useful data regarding the potential for damage to these systems given various accelerations and drift ratios, with fragility curves (discussed later in this article) being developed for wall assemblies [33, 36–38], ceiling assemblies [32, 39], and sprinkler systems [34, 35, 40]. Such data are extremely important for assessment of post-earthquake fire performance, as they provide a mechanism to assess reliability and availability of active and passive fire protection measures following an earthquake. At present, the seismic engineering community is using such data primarily for

76

B.J. Meacham

physical loss estimates, replacement costs, and downtime due to earthquake, but they can be helpful in estimating fire loss as well. Others, however, have looked at the fire performance of passive fire protection issue. Collier [41], for example, investigated the post-earthquake fire performance of passive fire protection systems, finding as much as a 50 % reduction in fire resistance of a 60-min plasterboard lined wall. A photo from this work is reproduced in Fig. 7.4. More recently, research has been conducted in Japan on the performance of compartment walls and structural steel systems exposed to post-earthquake fire [42], and a series of largescale earthquake and post-earthquake fire performance experiments were conducted on the Large Outdoor High Performance Shake Table (LHPOST) at the University of California San Diego [43–55], referred to as the Building Nonstructural Components and Systems (BNCS) tests, which included assessment of compartment wall (interior and exterior), ceiling performance, and sprinkler system performance, as well as other nonstructural systems (all project reports are available for download at the project website – http://bncs.ucsd.edu/index.html). The BNCS experiments were unique in that they involved construction of a five-story reinforced concrete structure on the LHPOST, conducting earthquake motion experiments for base-isolated and fixed-base configurations, collecting measurements during experiments and observations after each motion, and then conducting a series of in situ burn experiments to gain insight on the earthquake and postearthquake fire performance of the structure. Figure 7.5 shows how the stair at the floor 3 landing became disconnected, rendering the stair impassable. Figure 7.6 shows damage that occurred to the elevator door, rending the elevator unusable. Figure 7.7 shows ceiling damage which resulted from the motion tests. Details of the research program and outcomes are published elsewhere [43–55], but it is worthwhile to note that damage to fire protection systems, as observed in real earthquake events and previous experiments as outlined above, was observed, reinforcing that very real concerns exist with respect to post-earthquake building fire performance.

7.3

Performance-Based Design for Earthquakes and Fire: Similar But Independent

Performance-based design (PBD) has been identified by both the seismic and fire engineering communities as a means to better understand building performance in the face of hazard events, allowing for design solutions which

Fig. 7.4 Fire extension in openings created by ground motion [41]

Fig. 7.5 Damage to stairs during BNCS experiments

meet societal and economic targets in terms of life safety, property protection, and business continuity. PBD is rooted in the concept of delivering high-performing and resilient buildings that meet stakeholder risk tolerance objectives, with the concept of risk as a basis of performance becoming more central in recent years (e.g., see [56–68]). However, to date, the PBD approaches in each discipline have been developed independently and do not yet have a common understanding or framework. In addition, experimental data are lacking to support many of the target performance measures. Performance-based seismic design (PBSD) currently has a wide range of meanings. To many structural engineers in the USA, it means enhanced performance of the seismic force-resisting system that is demonstrated through displacement-based methods such as pushover analysis such as described in ASCE 41 [69] and FEMA 356 [70]. For others, it means developing structural systems

7

Toward a Risk-Informed Performance-Based Approach for Post-Earthquake Fire. . .

77

Select Performance Objectives Perform Preliminary Analysis Assess Performance Capability

DoesPerformance MeetObjectives?

No

Revise Design and/or Objectves

Yes Done Fig. 7.8 PBSD process outlined in ATC-58 guidelines [66] Fig. 7.6 Damage to elevator door during BNCS experiments

Fig. 7.7 Damage to ceiling on floor 1 during BNCS experiments

with seismic fuses and controlled inelastic behavior, where select key response parameters can be more narrowly predicted [71]. Yet another adaptation is to provide alternate design approaches other than those prescribed in the building codes to achieve the same or better performance than code-based construction using prescriptive requirements. Another approach is to satisfy the more stringent force and certification requirements for higher occupancy buildings (including nonstructural components and systems) such as those prescribed by the California Office of Statewide Health Planning and Development (OSHPD) for hospitals. Finally, for some, it means improving the overall performance of the building (including nonstructural components and systems) using enhanced protection methods such as seismic isolation and supplementary energy dissipation devices [71].

As a means to move toward some consistency, nextgeneration PBSD procedures have recently been developed for the FEMA guidelines series under the auspices of the Applied Technology Council [66]. As stated in the guidelines [66], performance-based seismic design is a formal process for seismic design of new buildings or design of seismic upgrades for existing buildings with the specific intent that the buildings will be able to achieve specified performance objectives in future earthquakes. Performance objectives relate to the amount of damage the building may experience and the consequences of this damage including potential casualties, loss of use or occupancy, and repair and reconstruction cost. In the performance-based process, owners, designers, and other stakeholders jointly identify the desired building performance characteristics at the project outset. Then as design decisions are made, the effect of these decisions on the building’s performance capability is evaluated to assure that the final structure will be capable of achieving the target performance. The process initiates with selection of one or more performance objectives. The process is illustrated in Fig. 7.8. Each performance objective is a statement of the acceptable risk of incurring damage or loss for identified earthquake hazards. Once the decision-makers select the performance objectives, design professionals must develop their design to a sufficient level of detail to allow determination of the building’s performance characteristics. For new buildings, this will include, as a minimum, identification of (1) the location and characteristics of the site; (2) building size, configuration, and occupancy; (3) type, location, and character of finishes and nonstructural systems; (4) selection of structural system type and configuration; and

78

(5) developing estimates of the strength, stiffness, and ductility of this system. In the case of existing buildings, these characteristics are already defined, and it is only necessary to determine what they are, and then define preliminary concepts for retrofit measures, if needed. Performance assessment is the primary subject of the ATC-58 methodology. In this step, engineers conduct a series of structural analyses to predict the building’s response when subjected to the earthquake hazards identified as part of the performance objectives and then use the information obtained from these analyses to assess the amount of damage that may occur and the probable consequences of this damage. Following performance assessment, the engineer compares the predicted performance with the desired performance. If the assessed performance matches or is superior to the stated performance objectives, the design is adequate and the project can be completed, assuming that the cost of completion is acceptable. If the assessed performance does not meet the performance objectives, the design team must either revise the design or alter the performance objectives in an iterative process, until the assessed performance meets acceptable objectives. The ATC-58 guidance [66] builds on the approach developed by the Pacific Earthquake Engineering Research (PEER) Center [e.g., see 60–63, 67], which is illustrated in Fig. 7.9. The PEER PBSD methodology integrates in a probabilistic approach to seismic hazard analysis, seismic demand analysis, capacity analysis (also called fragility analysis), damage analysis, and loss analysis. In brief, seismic hazard analysis (SHA) characterizes probabilistically the ground motion intensity measure in terms of a random variable called intensity measure (IM) typically taken as the spectral acceleration at the fundamental period of the structure.

Fig. 7.9 Loss analysis with PEER PBSD [72]

B.J. Meacham

The outcome of SHA is a seismic hazard curve for the site of the considered structure. The outcome of probabilistic seismic demand hazard analysis (DHA) is a seismic demand hazard curve for each of a number of response parameters (e.g., peak interstory drift, floor acceleration). These response parameters, called engineering demand parameters (EDPs), are obtained through nonlinear, finite element, time history analysis of the structure. Finally, the convolution of the seismic demand hazard curve for an EDP associated with a limit state of the structure with the corresponding fragility curve of the structure (characterizing probabilistically the capacity/resistance of the structure against this limit state) yields the probabilistic performance assessment results in terms of the mean annual frequency/rate (inverse of return period) of limit-state exceedance due to seismic action for each of several limit states. These results can be propagated further to decision variables (DVs) that relate to casualties, direct/indirect economic loss, and downtime and are of great interest to stakeholders. The objective of probabilistic seismic loss analysis is to assess DVs probabilistically (e.g., compute the mean annual rate of the total repair/replacement cost due to seismic damage exceeding a specified dollar amount) for a given structure at a given location. This PBSD methodology is attractive because it accounts consistently for the aleatory and epistemic sources of uncertainty associated with the estimation of the seismic hazard, the simulated structural response (effects of record-to-record variability for a given IM), the structural capacity, and the cost associated with the number of human casualties, the repair of individual structural components or replacement of the entire structure, and the downtime of the structure during repair, for a given damage level. This fundamental concept was the basis for the ATC-58 methodology [66]. As with the PBSD process, the PBFSD process [59] requires clearly specified goals and objectives (in terms of

7

Toward a Risk-Informed Performance-Based Approach for Post-Earthquake Fire. . .

79

Fig. 7.10 PBFSD process from SFPE [59]

function, performance, and/or loss), performance objectives (expectation of building performance under load), design loads (from normal use or hazard events expected to impact the building), and performance/design/failure criteria (metrics to judge successful performance). One representation of the process is illustrated in Fig. 7.10 [59]. At the present time, however, the PBFSD approach is predominantly deterministic in nature, with probabilistic methods used primarily for scenario development. In addition, existing guidance (e.g., [59, 73–75]) is largely process oriented, and it has been observed by many that design loads for fire are not well defined, that criteria are lacking, and that a comprehensive and widely accepted framework that links the components in a comprehensive risk-informed framework does not yet exist (e.g., [57, 58, 76–80]). While various efforts have been undertaken to address some of these shortcomings, either through “prescribed performance”-type approaches [79, 80] or risk-informed performance-based design approaches [68, 81, 82], the concepts are not fully developed, and it is not clear how well such approaches line up with specific PBSD representations of loads, criteria, and performance measure. Going forward, it would be helpful to move toward a performance-based design approach which

uses common approaches and integrates seismic damage and fire performance assessment.

7.4

Toward an Integrated, Risk-Informed Performance-Based Approach for Post-Earthquake Fire Protection Design of Buildings

While the approaches for PBSD and PBFSD have been developing along independent paths, it is possible to contemplate bringing them together for PBD of structures to earthquake and post-earthquake fire. Although this can be accomplished within a deterministic structure, a riskinformed framework, which utilizes probabilistic representations of loads and tolerable performance, would seem more appropriate. One approach to this problem is to adopt the PBSD approach for fire. Such an idea has been suggested by Deierlein and Hamilton [83, 84] with respect to structural fire engineering. In essence, Deierlein and Hamilton argue that SHA can be replaced with a fire hazard analysis, which results in design fire curves that can be deterministically or

80 Fig. 7.11 Structural fire engineering approach based on PEER PBSD approach [84]

B.J. Meacham

• Collapse & Casualties

Decision Decision Variable DecisionVariable Variable

• Direct Financial Loss • Downtime Damage Damage Measure DamageMeasure Measure

Engineering Engineering Demand EngineeringDemand Demand Parameter Parameter Parameter 1000 900 800 700 600 500 400 300 200 100 0

Gas Unportected Steel Insulated Steel

Intensity Intensity Measure IntensityMeasure Measure 0

30

60

90 120 150 180 210 240 Time [min]

COMPARTMENT TIME –TEMP. CURVES

probabilistically represented as IMs, the seismic EDMs with fire-related EDMs (e.g., temperatures, flashover, etc.), DMs that reflect fire-related damage (e.g., deflection, smoke spread, untenable conditions, etc.), and DVs that reflect tolerable levels of casualties, property damage, or economic impact and the thresholds for acceptability. One could see where this could be adapted into a common approach for performance-based earthquake and fire design. This is illustrated in Fig. 7.11. A different approach was suggested by Sekizawa et al. [24], who developed a framework for assessing the seismic-induced fire risk in a building along with the building fire safety performance level. This framework requires predicting the earthquake response of a building by taking into consideration the frequency and vibration characteristics of an earthquake and a building respectively. The structural response data is used to determine the damage levels of active and passive fire systems. A fault tree analysis is used to determine the dominant failure mode as well as the overall functional failure of a sprinkler system for active systems, while data from literature are used to reduce fire resistance time of compartments such as walls and fire doors due to lack of data. These failure probabilities of fire systems are used to predict transition probability of fire phases and burned area for a post-earthquake fire scenario. More recently, a conceptual framework has been suggested by Kim [55, 85], which illustrates how the ATC-58 approach could be applied to PBFSD, thus integrating earthquake and post-earthquake fire analysis and design of structures. The conceptual framework was developed based on factors that can affect the overall building performance levels and the building design objectives during fire and earthquake. The conceptual framework diagram is shown in Fig. 7.12. The main objective of the

diagram is to illustrate where the direct and indirect relationships lie between earthquakes and fires that occur in a building. The diagram shows a building and, given its characteristics, identifies what the possible consequences could be when an event of either fire or earthquake occurs. The relationship between the two events is mapped out to show what type of post-earthquake building fire conditions could be expected. The diagram is separated into three areas with the upper left, upper right, and bottom half sections dedicated to earthquake, fire, and post-earthquake fire, respectively, as separate events. The model is intended to show what the hazards, the consequences, and their effecting building attributes are for the separate events of an earthquake and a fire and also how these consequences combine to affect the building fire safety performance during a post-earthquake fire scenario [85]. The major hazards that can severely affect the building performance for both events are identified with red text in black-colored boxes, and some of the building attributes that can also influence the outcomes are identified with black text in black boxes. For both an earthquake and a fire, the intensity and duration of the event are considered to be the key hazards. What happens as a result of the event are termed “consequences” and are identified with red text in red boxes, and the building factors that may have a direct impact on certain consequences are mapped out by dotted gray arrow lines to show the connections. The red-dotted arrow lines are termed as secondary connections and are mapped to show which consequences could affect a post-earthquake building fire performance and to highlight the concerns in which a fire protection engineer should consider when developing fire safety designs for earthquake-prone buildings. The question is where does the fire protection engineer (FPE) get

7

Toward a Risk-Informed Performance-Based Approach for Post-Earthquake Fire. . .

Fig. 7.12 Conceptual model of integrated approach to performance-based seismic and fire design [85]

81

EARTHQUAKE PFA IDR

Intensity

Intensit y Duration

Height

FIRE

Duration

Geometry

Structural Smoke Production/Spread

Material Building Quality

Fire

EQ

Ventilation

NCS Temperature

Egress

Fuel Flame Spread

Physical Barriers

FP Systems

Rigidit

EQ + Fire

Structural Collapse ASET Decrease RSET Increase Limited Firefighting/Rescue

POST-EARTHQUAKE FIRE

Fig. 7.13 Steps in integrated approach to performance-based seismic and fire design [85]

earthquake damage data? As suggested by Kim [85], this could come from the ATC-58 approach. In the ATC-58 methodology [66], information on all of the building components that are subject to damage during earthquake motions is to be obtained. Such information could help FPEs identify the important building components that may affect post-earthquake building fire safety performance levels and set the tone for defining the project scope. When a seismic engineer performs building analysis using

the selected design earthquake motion, the results and building demand parameter values obtained can be passed on to the FPEs. Also, the fragility curves associated with each building component used to calculate damage states to calculate the overall building performance during an earthquake are essential information for the FPE. An FPE could use the building demand parameter and the fragility curves to determine damage conditions of specific building components which could help in setting up design fire scenarios for post-earthquake building fires. Combining the key elements of the ATC-58 PBSD approach and the general PBFSD approach, an integrated approach for considering post-earthquake fire performance is obtained. This is illustrated below, where text in the red boxes reflect the standard PBFSD approach, and text in the green boxes reflect input from the PBSD approach [85]. Since the PBSD and PBFSD processes are essentially the same (see Figs. 7.8 and 7.10), Fig. 7.13 simply integrates the seismic performance analysis into the fire performance analysis process. This works because goals, objectives, building characteristics, and related features are the same in each process and do not need to be duplicated. In the combined approach, the ATC-58 process [66] is used to develop the seismic hazard (Step 6), expected damage conditions of the building, which could have a significant impact on fire performance (Steps 7 and 8), and likelihood of damage (Step 9). This yields insight on availability of key systems and components, including interior compartmentation (walls, ceiling systems), stair

82

B.J. Meacham

Fig. 7.14 Comparison of building fire/life safety performance with and without earthquake damage consideration [85]

integrity, elevator integrity, exterior wall/fac¸ade systems, sprinkler systems, mechanical (smoke management) systems, and lighting and electrical systems, which in turn impact both the potentials for fire development and spread, as well as safe evacuation of occupants. Whereas a typical PBFSD approach does not consider earthquake or other precursor events as damaging fire protection systems, this approach provides damage states for performance assessment. The concept was applied to a fictitious building (shopping mall) in which available safe egress times (ASET) were compared to required safe egress times (RSET) in the PBFSD process with and without consideration of earthquake-induced damage. It should be noted that at the time of this analysis, some of the fragility data, such as for sprinkler systems which have only recently been published, were not available [34, 35] and were not in the ATC-58 Performance Assessment Calculation Tool (PACT). Additionally, damage assessment of key systems, such as compartment walls and ceilings, is coded in PACT for replacement cost estimates, more so than fire performance. As such, factors such as actual size (percent area) of openings developed due to cracking, joint separation, and similar openings in ceilings and walls due to earthquake motion are not estimated. The same is true of percentage of glazing which falls out of its frame. Since these values are important for estimation of fire size and spread (ventilation and compartmentation), estimates had to be made for the fire analysis. Likewise, distribution of fuel load was not anticipated in the PACT, so assumptions on this were made as well. Finally, a combination of PACT assessments and assumptions was used to estimate injuries to people, which could impact ability to escape and time of movement. In some cases, such as damage to stairs, fragility curves are available in PACT. Given the assumptions and limitations as outlined above, computational analysis of fire and smoke development and

spread and of time to evacuate the building – in the undamaged and earthquake-damaged conditions – was simulated. For the fire effects analysis, PyroSim and Fire Dynamics Simulator (FDS) were used [86, 87], and for evacuation analysis, the tool Pathfinder was used [88]. Based on the analysis, ASET and RSET estimates were developed for the undamaged and the earthquake-damaged mall building. A comparison of one set of ASET and RSET values is presented in Fig. 7.14. While the outcomes should not be taken as absolute, especially given the limits on data availability and number of assumptions made, the magnitude of the difference in outcomes is significant. This suggests at the very least that earthquake damage should be considered when developing performance-based fire solutions to buildings in earthquake-prone areas.

7.5

Conclusions

Post-earthquake fire in buildings is a significant, yet underresearched, area. It has been shown that movement toward an integrated approach for risk-informed performance-based analysis of post-earthquake fire performance of buildings is needed and can be achieved. Although some data are available for building component performance under seismic loads, much more data is required to perform detailed and accurate fire performance analysis and design of buildings in earthquake-prone area. While different resources can be utilized to find needed data, the use of data from a single resource, or at a minimum in a common format, is needed, especially since data collected from different earthquake events and experiments would have been developed under differing seismic loads. This creates a significant source of uncertainty in current data. In addition, since several of the current damage estimates are made by seismic engineering experts based on their judgment and experiences, this contributes to the uncertainty as well.

7

Toward a Risk-Informed Performance-Based Approach for Post-Earthquake Fire. . .

Going forward, more interaction between the fire and seismic engineering communities is encouraged in attempts to better understand approaches, data, and analysis needs and to identify the connecting links between earthquake and fire. This interaction can help seismic engineers better understand how different building components can affect building fire safety performance and can help fire safety engineers better understand building component performance under seismic loads to better identify post-earthquake building fire hazards. For now, the conceptual framework can be used to help provide context to the problem and to provide general steps and procedures required to conduct post-earthquake building fire analysis, identify connections that link building earthquake and fire events, and instill awareness of post-earthquake fire hazards to seismic and fire engineers. Acknowledgment The author gratefully acknowledges the invitation to present this paper and the support which made it possible. The author also acknowledges the contributions of Jin-Kyung Kim and of all the project participants in the referenced BNCS experimental program.

References 1. Photo from Steinbrugge Collection of the UC Berkeley Earthquake Engineering Research Center, as available from USGS http://earth quake.usgs.gov/regional/nca/1906/18april/index.php 2. NOAA (1972) A study of earthquake losses in the San Francisco bay area – data and analysis. A report prepared for the Office of Emergency Preparedness: U.S. Department of Commerce, National Oceanic and Atmospheric Administration (NOAA), 220 p 3. Geschwind CH (2001) California earthquakes: science, risk and the politics of hazard mitigation. The Johns Hopkins University Press, Baltimore 4. Usami T (2006) Earthquake studies and the earthquake prediction system in Japan. J Disaster Res 1(3):416–433 5. SPA (2008) Enhancements in HAZUS-MH fire following earthquake. SPA Project No. 10010-01-07-01, SPA Risk LLC, Berkeley CA, Prepared for PBS&J and the National Institute of Building Sciences, Berkeley 6. Scawthorn C, Cowell AD, Borden F (1998) Fire-related aspects of the Northridge earthquake. NIST GCR 98-743. National Institute of Standards and Technology Building and Fire Research Laboratory, Gaithersburg, 165pp 7. Scawthorn C, Eidinger JM, Schiff AJ (2005) Fire following earthquake, Technical council on lifeline earthquake engineering monograph no. 26. American Society of Civil Engineers, Reston, 345pp 8. Chung RM (1996) January 17, 1995 Hyogoken-Nanbu (Kobe) earthquake: performance of structures, lifelines, and fire protection systems, NIST special report SP 901. NIST, Gaithersburg, Available at http://www.nist.gov/manuscript-publication-search.cfm? pub_id¼908748 9. Ohnishi K (1997) Causes of the seismic fires following the Great Hanshin-Awaji earthquake survey. In: Beall KA (ed) NISTIR 6030, U.S./Japan Government Cooperative Program on Natural Resources (UJNR), fire research and safety, 13th joint panel meeting, vol 2, 13–20 Mar 1996 Gaithersburg, pp 337–344

83

10. Fire and Disaster Management Agency (2006) Report of the 1995 Hyogo-ken Nanbu earthquake. Fire and Disaster Management Agency, Japan 11. Himoto K, Yamada M, Nishino T (2014) Analysis of ignitions following 2011 Tohoku earthquake using Kawasumi model. In: Proceedings of the International Association for Fire Safety Science, 11th international symposium. International Association for Fire Safety Science. Available at http://www.iafss.org/publications/ fss/11/24/view 12. Sekizawa A, Sasaki K (2014) Study on fires following the 2011 Great East-Japan earthquake based on the questionnaire survey to fire departments in affected areas. In: Proceedings of the International Association for Fire Safety Science, 11th international symposium. Available at http://www.iafss.org/publications/fss/11/100/ view 13. Jones LM, Bernknopf R, Cox D, Goltz J, Hudnut K, Mileti D, Perry S, Ponti D, Porter K, Reichle M, Seligson H, Shoaf K, Treiman J, Wein A (2008) The ShakeOut scenario. USGS open file report 2008-1150, Reston. Available at http://pubs.usgs.gov/of/ 2008/1150/of2008-1150small.pdf 14. Scawthorn CR (2011) Fire following earthquake aspects of the southern San Andreas fault Mw 7.8 earthquake scenario. Earthq Spectra 27(2):419–441, http://dx.doi.org/10.1193/1.3574013 15. Lee S, Davidson R, Ohnishi N, Scawthorn C (2008) Fire following earthquake – reviewing the state-of-the-art of modeling. Earthquake Spectra 24(4):933–967 16. Zolfaghari MR, Peyghaleh E, Nasirzadeh G (2009) Fire following earthquake, intra-structure ignition modeling. J Fire Sci 27:45–79 17. Lee SW, Davidson RA (2010) Physics-based simulation model of post-earthquake fire spread. J Earthq Eng 14(5):670–687. doi:10. 1080/13632460903336928 18. Thomas G, Heron D, Cousins J, de Roiste M (2012) Modeling and estimating post-earthquake fire spread. Earthquake Spectra 28 (2):795–810, http://dx.doi.org/10.1193/1.4000009 19. NIST (1994) 1994 Northridge earthquake: performance of structures, lifelines and fire protection systems, NISTIR 5396. NIST, Gaithersburg 20. Fleming R (1998) Analysis of fire sprinkler system performance in the Northridge earthquake, NIST-GCR-98–736. NIST, Gaithersburg 21. Scawthorn C (1996) Fire following the Northridge and Kobe earthquakes. U.S./Japan government cooperative program on natural resources (UJNR). In: Beall KA (ed) Fire research and safety. 13th joint panel meeting, vol 2, 1997. National Inst. Of Standards and Technology, Gaithersburg, pp 325–335. Available at http://fire. nist.gov/bfrlpubs/fire97/PDF/f97147.pdf 22. LeGrone P (2004) An analysis of fire sprinkler system failures during the Northridge earthquake and comparison with the seismic design standard for these systems. In: Proceedings, 13th world conference on earthquake engineering, Vancouver. Available at http://www.iitk.ac.in/nicee/wcee/article/13_2136.pdf 23. Chen S, Lee GC, Masanobu S (2004) Hazard mitigation for earthquake and subsequent fire. http://mceer.buffalo.edu/research/Inter national_Research/ANCER/Activities/2004/chen_sw_mceer.pdf. Accessed on 20 Aug 2007 24. Sekizawa A, Ebihara M, Notake H (2003) Development of seismicinduced fire risk assessment method for a building. In: Fire safety science – proceedings of the seventh international symposium, International Association for Fire Safety Science, pp 309–320 25. Elwood KJ, Pampanin S, Kam WY (2012) 22 February 2011 Christchurch earthquake and implications for the design of concrete structures. In: Proceedings of the international symposium on engineering lessons learned from the 2011 Great East Japan earthquake, 1–4 Mar 2012, Tokyo. Available at http://www.jaee.gr.jp/event/ seminar2012/eqsympo/pdf/papers/178.pdf

84 26. Clifton C, Bruneau M, MacRae G, Leon R, Fussell A (2011) Steel structures damage from the Christchurch earthquake series of 2010 and 2011. Bull N Z Soc Earthq Eng 44(4):297–318, Available at http://www.nzsee.org.nz/db/SpecialIssue/44(4)0297.pdf 27. Dhakal RP, MacRae GA, Hogg K (2011) Performance of ceilings in the February 2011 Christchurch earthquake. Bull N Z Soc Earthq Eng 44(4):377–387, Available at http://www.nzsee.org.nz/db/ SpecialIssue/44(4)0377.pdf 28. Baird A, Palmero A, Pampanin S (2011) Fac¸ade damage assessment of multi-storey buildings in the 2011 Christchurch earthquake. Bull N Z Soc Earthq Eng 44(4):368–376, Available at http://www.nzsee.org.nz/db/SpecialIssue/44(4)0368.pdf 29. Villaverde R (1997) Method to improve seismic provisions for nonstructural components in buildings. J Struct Eng 123:432–439 30. Taghavi S, Miranda E (2003) Response assessment of non-structural building elements, Report No. 2003/05. Pacific Earthquake Engineering Research Center, Richmond 31. McMullin KM, Merrick DS (2007) Seismic damage thresholds for gypsum wallboard partition walls. J Archit Eng ASCE 13 (207):22–29 32. Reinhorn AM, Ryu KP, Maddaloni G (2010) Modeling and seismic evaluation of nonstructural components: testing frame for experimental evaluation of suspended ceiling systems, Report MCEER10-0004, June 2010, 182 pages 33. Davies R, Retamales R, Mosqueda G, Filiatrault A (2011) Experimental seismic evaluation, model parameterization and effects of cold-formed steel-framed gypsum partition walls on the seismic performance of an essential facility, report MCEER-11-0005, Oct 2011, 218 pages. Available at http://mceer.buffalo.edu/pdf/report/ 11-0005.pdf) 34. Tian Y, Filiatrault A Mosqueda G (2013) Experimental seismic study of pressurized fire sprinkler piping systems, report MCEER13-0001, Apr 2013, 404 pages 35. Soroushian S, Maragakis EM, Zaghi AE, Echevarria A, Tian Y, Filiatrault A (2014) Comprehensive analytical seismic fragility of fire sprinkler piping systems, Technical report MCEER-14-02, 26 Aug 2014. Available at http://mceer.buffalo.edu/pdf/report/140002.pdf) 36. Bersofsky A (2004) A seismic performance evaluation of gypsum wallboard partitions. MS thesis. University of California, San Diego 37. Lang A, Restrepo J (2006) Seismic performance evaluation of gypsum wallboard partitions. In: The proceedings of the national conference on earthquake engineering. San Francisco 38. Kanvinde AM, Deierlein GG (2006) Analytical models for the seismic performance of gypsum drywall partitions. Earthquake Spectra 22(2):391–411, http://dx.doi.org/10.1193/1.2191927 39. Badillo-Almaraz H, Whittaker AS, Reinhorn AM, Cimellaro GP (2006) Seismic fragility of suspended ceiling systems, MCEER-060001, 4 Feb 2006. Available at http://mceer.buffalo.edu/pdf/report/ 06-0001.pdf 40. Malhotra PK, Senseny PE, Braga AC, Allard RI (2003) Testing sprinkler-pipe seismic-brace components. Earthquake Spectra 19 (1):87–109, http://dx.doi.org/10.1193/1.1543160 41. Collier, P.C.R. 2005. Post-earthquake performance of passive fire protection systems. BRANZ Study, Report 147, Judgeford, Wellington 42. Naruse T, Suzuki J (2015) Experimental study on fire resistance of mid-rise steel construction exposed to post-earthquake fire. Presented at joint NIST-BRI workshop operation Tomadachi, NIST, Gaithersburg, March 2015 43. Pantoli E, Chen MC, Astroza R, Ebrahimian H, Mintz S, Wang X, Hutchinson T, Conte J, Restrepo J, Meacham B, Kim J, Park H (2013) BNCS report #2: full-scale structural and nonstructural building system performance during earthquakes and postearthquake fire – test results. Structural systems research project

B.J. Meacham report series, SSRP 13/10. University of California, San Diego (available at) 44. Chen M, Pantoli E, Astroza R, Ebrahimian H, Mintz S, Wang X, Hutchinson T, Conte J, Restrepo J, Meacham B, Kim J-K, Park H (2013) BNCS report #1: full-scale structural and nonstructural building system performance during earthquakes and postearthquake fire – specimen design, construction, and test protocol. University of California, San Diego 45. Pantoli E, Chen M, Wang X, Astroza R, Ebrahimian H, Mintz S, Hutchinson T, Conte J, Restrepo J, Meacham B, Kim J-K, Park H (2013) BNCS report #2: full-scale structural and nonstructural building system performance during earthquakes and postearthquake fire – test results. University of California, San Diego 46. Pantoli E, Chen M, Hutchinson T, Restrepo J (2013) BNCS report #3: full-scale structural and nonstructural building system performance during earthquakes and post-earthquake fire – camera and analog sensor details. University of California, San Diego 47. Chen M, Pantoli E, Wang X, Mintz S, Hutchinson T, Hutchinson T, Restrepo J (2013) BNCS report #4: full-scale structural and nonstructural building system performance during earthquakes and post-earthquake fire – construction details and technical specifications of specific subsystems. University of California, San Diego 48. Kim JK, Meacham BJ, Park H (2013) Full-scale structural and nonstructural building systems performance during earthquakes and post-earthquake fire: fire test program and preliminary outcomes. Worcester Polytechnic Institute, Worcester 49. Hutchinson T, Restrepo J, Conte J, Meacham BJ (2013) Overview of the building nonstructural components and systems (BNCS) project. In: Proceedings, 2013 ASCE structures congress. ASCE, Reston 50. Pantoli E, Chen M, Wang X, Hutchinson T, Meacham B, Park H (2013) Shake table testing of a full-scale five-story building: seismic performance of the major nonstructural components: egress systems, facades. In: Proceedings, 2013 ASCE structures congress, ASCE, Reston 51. Meacham BJ, Kim JK, Park H (2013) Shake table testing of a fullscale five-story building: post-earthquake fire performance. In: Proceedings, 2013 structures congress, ASCE, Reston 52. Park H, Meacham BJ, Kim JK (2014) Fire performance of full-scale building subjected to earthquake ground motion: fire test program and outcomes. In: Proceedings, 11th international symposium on fire safety science, International Association for Fire Safety Science. Available at http://www.iafss.org/publications/fss/11/71/ view 53. Kim JK, Meacham BJ, Park H, Hutchinson T, Pantoli E (2014) Fire performance of full-scale building subjected to earthquake ground motion: test specimen, ground motions and seismic performance of fire protection systems. In: Proceedings, 11th international symposium on fire safety science, International Association for Fire Safety Science. Available at http://www.iafss.org/publications/fss/11/90/ view 54. Wang X, Pantoli E, Hutchinson TC, Restrepo J, Wood RL, Hoehler MS, Grzesik P, Sesma FH (2015) Seismic performance of coldformed steel wall systems in a full-scale building. J Struct Eng. doi:10.1061/(ASCE)ST.1943-541X.0001245, ASCE, Reston, VA, USA 55. Meacham BJ (2015) Post-earthquake fire performance of buildings: summary of a large-scale experiment and conceptual framework for integrated performance-based seismic and fire design. Fire Technol. doi:10.1007/s10694-015-0523-9, published on-line 30 July 2015 56. Hamburger RO, Court AB, Soulages JR (1995) Vision 2000: a framework for performance-based engineering of buildings. In: Proceedings of the 64th annual convention, Structural Engineers Association of California, 19–21, pp 127–146

7

Toward a Risk-Informed Performance-Based Approach for Post-Earthquake Fire. . .

57. Meacham BJ (1998) Assessment of the technological requirements for realization of performance-based fire safety design in the United States: final report, NIST GCR 98-763. NIST, Gaithersburg 58. Meacham BJ (1999) International experience in the development and use of performance-based fire safety design methods: evolution, current situation, and thoughts for the future. In: Proceedings of the international association for fire safety science, 6th international symposium, pp 59–76 59. SFPE (2000) SFPE engineering guide to performance – based fire protection analysis and design of buildings. NFPA, Quincy, MA. 60. Cornell CA, Krawinkler H (2000) Progress and challenges in seismic performance assessment. PEER Center News, 3(2). http://peer. berkeley.edu/news/2000spring/index.html 61. Jalayer F, Cornell CA (2003) A technical framework for probability-based demand and capacity factor design (DCFD) seismic formats, Report no. PEER 2003/08, Pacific Earthquake Engineering Research Center, Berkeley, Nov 2003 62. Porter KA (2003) An overview of PEER’s performance-based earthquake engineering methodology. In: Proceedings of the ninth international conference on applications of statistics and probability in civil engineering (ICA SP9) 6–9 July 2003, San Francisco 63. Moehle J, Deierlein GG (2004) A framework methodology for performance-based earthquake engineering. In: Proc. 13th conf. on earthquake eng., Vancouvera, 1–6 Aug 64. Meacham BJ (2004) Performance-based building regulatory systems: structure, hierarchy and linkages. J Struct Eng Soc N Z 17(1):37–51 65. Meacham BJ Editor, Johann M Associate Editor (2006). Extreme event mitigation in buildings: analysis and design. National Fire Protection Association, Quincy 66. ATC-58 (2012) Guidelines for seismic performance assessment of buildings (100% Draft). Applied Technology Council, Redwood City 67. Gunay S, Mosalam KM (2013) PEER performance-based earthquake engineering methodology, revisited. J Earthq Eng 17 (6):829–858. doi:10.1080/13632469.2013.787377 68. Alvarez A, Meacham BJ, Dembsey NA, Thomas JR (2014) A framework for risk-informed performance-based fire protection design for the built environment. Fire Technol 50:161–181. doi:10.1007/s10694-013-0366-1 69. ASCE 41–13 (2013) Seismic evaluation and retrofit of existing buildings. ASCE, Reston 70. FEMA 356 (2000) Prestandard and commentary for the seismic rehabilitation of buildings. Federal Emergency Management Agency, Washington, DC 71. Personal communication, Robert Bachman, Sept 2008 72. Deierlein GG (2006) Quantifying risk by performance-based earthquake engineering. In: Presentation at the IRCC workshop on use of risk in regulation, San Francisco, Oct 2006

85

73. ISO/TR 13387-1 (1999) Fire safety engineering – Part 1: Application of fire performance concepts to design objectives. International Organization for Standardization, Geneva 74. BS7974 (2001) Application of fire safety engineering principles to the design of buildings. British Standard Institute 75. International Fire Engineering Guidelines (2005) Australian Building Codes Board (ABCB), Canberra, ACT, Australia 76. Alvarez A, Meacham BJ, Dembsey NA, Thomas JR (2013) 20 years of performance-based fire protection design: challenges faced and a look ahead. J Fire Prot Eng 23(4):249–276. doi:10. 1177/1042391513484911 77. Meacham BJ (n.d.) Fire safety engineering at a crossroad. Case studies in fire safety, Elsevier. http://dx.doi.org/10.1016/j.csfs. 2013.11.001 78. Moore AE, Albano LD, Fitzgerald RW, Meacham BJ (2005) Defining design fires for structural performance. In: Proceedings of the 2005 ASCE structures congress. ASCE, Reston 79. Fleischmann CM (2011) Is prescription the future of performancebased design? In: Proceedings, 10th international symposium on fire safety science. International Association for Fire Safety. Available at http://www.iafss.org/publications/fss/10/77/view 80. C/VM2 (2013) Verification method: framework for fire safety design. Ministry of Business, Innovation and Employment, Wellington. Available at http://www.dbh.govt.nz/UserFiles/File/ Publications/Building/Compliance-documents/c-vm2-protectionfrom-fire-amendment-3.pdf 81. Albrecht C (2014) Quantifying life safety – Part 1: Scenario-based quantification. Fire Saf J 64:87–94 82. Albrecht C (2014) Quantifying life safety – Part 2: Quantification of fire protection systems. Fire Saf J 64:81–86 83. Deierlein GG, Hamilton S (2003) Framework for structural fire engineering and design methods. In: Proceedings of the NIST/ SFPE workshop to develop a national R&D roadmap for fire safety design and retrofit of structures. SFPE and NIST, Baltimore 84. Hamilton SR, Deierlein GG (2004) Probabilistic methodology for performance-based fire engineering. In: Proceedings, 5th international conference on performance-based codes and fire safety design methods. SFPE, Bethesda, pp 327–341 85. Kim J-K (2014) A conceptual framework for assessing postearthquake fire performance of buildings, MS Thesis, Department of Fire Protection Engineering. Worcester Polytechnic Institute, Worcester 86. PyroSim User Manual, Thunderhead Engineering, Manhattan KS (2014) Available at https://www.thunderheadeng.com/wp-content/ uploads/dlm_uploads/2014/02/PyroSimManual.pdf 87. Fire Dynamics Simulator (Version 6) – User’s Guide (2012) U.S. Dept. of Commerce, Technology Administration, National Institute of Standards and Technology, Gaithersburg 88. Pathfinder User Manual (2014) Thunderhead engineering, Manhattan. Available at http://www.thunderheadeng.com/wp-content/ uploads/downloads/2014/10/users_guide.pdf

Part III Compartment Fire

8

Interaction of a Pool Fire in a Compartment with Negative Pressure Generated by Mechanical Ventilation Yasuo Hattori, Ken Matsuyama, Hitoshi Suto, Seiji Okinaga, and Eiji Onuma

Abstract

We have experimentally investigated a medium-size pool fire in a compartment, the dimensions of which corresponded to the ISO 9705 room. Airflow rates in ducts, pressure, mole fraction, and temperature of air in the compartment and a mass loss rate of the fuel were measured. The liquid fuel and pool diameter were ethanol and 600 mm, respectively, which rapidly increased the compartment pressure just after ignition. The compartment was ventilated at inlet and outlet ducts with natural and mechanical ventilation systems, which initially gave negative compartment pressures in the range of
2 to
85 Pa. The negative pressure was much weaker than the pressure increase, which restrained the air supply with natural ventilation and resulted in extinction due to lack of oxygen. On the other hand, the negative pressure with the stronger mechanical ventilation sustained the air supply and yielded the transition to a ventilation-controlled fire without extinction. The ventilationcontrolled fire led to two kinds of oscillating flame: one was caused by poor oxygen supply, which is similar to that reported by previous studies, and the other was caused by repetition of ignition and extinction, which was attributed to the change in the flow rate and direction of fresh air at the inlet duct. This oscillation generated large pressure fluctuations but did not yield thermal energy with combustion. Keywords

Fire experiment  Limited ventilation  Oscillating fire  Pressure fluctuation  Ventilation controlled fire

Nomenclature D m P Q

Pool diameter Mass loss rate Difference pressure Ventilation flow rate

T x Y z

Temperature Radius distance from the centerline of plume Concentration Vertical distance from fuel surface

Greek Symbols τ

Time (ignition times are τ ¼ 0)

Y. Hattori (*)  H. Suto Central Research Institute of Electric Power Industry, 1646 Abiko, Abiko-shi 270-1194, Japan e-mail: [email protected]

Subscripts

K. Matsuyama  S. Okinaga  E. Onuma Tokyo University of Science, 2641 Yamazaki, Noda-shi, Chiba 278-8510, Japan

0 e

Values of initial condition Values of ethanol

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_8

89

90

i O2 s

8.1

Y. Hattori et al.

Values for ignition period Values of oxygen Values for steady period

Introduction

Accurate description of a pool fire in a compartment with confined ventilation is of practical interest in fire protection engineering [1]. Thus, such pool fires have been extensively investigated through theoretical, experimental, and numerical studies, which revealed peculiar fire behavior, e.g., ghosting and oscillating flames [2, 3], flame ejection [4], and spill plume [5]. Nevertheless, research on fires in a mechanically ventilated compartment is seldom found. This is because previous studies have mainly dealt with naturally ventilation compartments. In fact, the fire behavior with natural ventilation is generally classified by using the opening (ventilation) factor, which gives the flow rates only with natural ventilations [1, 3, 6]. Contrary to this, mechanical ventilation must superpose negative pressures on flow fields in compartments. The details of the interactions of a ventilation-controlled fire with such flow fields are, however, still unknown. Fires with mechanical ventilation are typical fire scenarios in offshore ships [7], nuclear power plants [8], and thus better understanding for such fire behavior is eagerly awaited. In this study a pool fire in a mechanically ventilated compartment was examined. Fire experiments were performed with a medium-size pool fire in a compartment, the dimensions of which correspond to the ISO 97051 room. The liquid fuel and pool diameter were ethanol and 600 mm, respectively, which rapidly increased the compartment pressure just after the ignition. The compartment was ventilated at inlet and outlet ducts with natural and mechanical ventilation systems, which gave negative values of initial pressure in the range of
2 to
85 Pa. The effects of flow fields with negative pressure due to mechanical ventilation on the transition to a ventilation-controlled fire were explored through the measurements of pressure, mole fraction, and temperature of air in the compartment.

8.2

Experimental Setup

Figure 8.1 shows a schematic drawing of the fire compartment, which was designed in accordance with Sano et al. [9]; they have minutely examined a compartment fire under

1

International Organization for Standardization, ISO 9705:1993 Fire tests – Full-scale room test for surface products.

limited ventilation conditions. The details of apparatus were presented in our previous report [10, 11] and thus the outline of specifications is described as follows. A pool fire was set on a circular pan located at 50 mm above the floor and in the center of a fire compartment. The fuel was continuously supplied through a siphon tube from a storage tank located outside the compartment. The fire compartment was comprised of autoclaved lightweight concrete (ALC) panels, steel plates, and support frames. The ALC panels were painted black and had a thickness of 50 mm. The compartment had internal dimensions of 3.6 m in length, 2.4 m in width, and 2.4 m in height. These dimensions correspond to the ISO 9705 room for standard fire tests. The outside of the panels was overlaid with steel plates and the joints between the steel plates were sealed using arc-welding and heat-resistant sealant to insure no unintended leaks. The compartment was ventilated by natural and mechanical systems, which consisted of an inlet and an outlet duct and a blower. The ducts were situated on the side walls of the fire compartment along the center plane. The inlet and outlet ducts were located near the floor and the ceiling and had a cross section of 0.008 m2 and 0.013 m2, respectively. The inlet duct was equipped with a baffle plate to reduce interference with the fire plume from the inflow. The blower (Muto Denki MIH-8/6), which has a maximum pressure head of 3.43 kPa and flow rate of 12.0 m3min
1, was installed downstream the outlet duct and produced a negative pressure in the fire compartment. Airflow rate in the ducts, differential pressure, mole fraction, and temperature of air in the compartment, heat flux through the walls of the compartment, and the mass loss rate of the liquid fuel were measured. The airflow rates within the ventilation ducts were measured with bidirectional probes [12], which are similar to pitot-static probes, connected to a differential pressure transducer. The pressure in the fire compartment was measured with a differential pressure transducer. Species concentrations were measured continuously at two locations near the pool in the compartment (0.35 m above the floor) and in the outlet duct. The time lag of these measurements was approximately 10 s. Gas temperature distributions in the compartment were measured with a thermocouple grid placed within the compartment. The grid consisted of 50 thermocouples (K-type, 1 mm diameter). Fuel mass loss was measured with an electronic balance installed under the storage tank. In the present study, the liquid fuel and the pan diameter, D, were ethanol and 600 mm, respectively, which is sufficient to cause a pressure increase in the compartment just after ignition. Before the ignition of the pool fire, the blower output was adjusted to get a negative pressure in the range of
2 to
85 Pa, as shown in Table 8.1, which also indicates the corresponding ventilation flow rate. Notice that no

8

Interaction of a Pool Fire in a Compartment with Negative Pressure Generated by. . .

Fig. 8.1 Schematics of experimental apparatus, pool, fire compartment, and measurement instrument

ALC panel

91

steel plate

support frame

gas analyzer

blower bi-direction probe and TC

z

baffle plate pan

O

heat flux sensor x pressure transducer

Table 8.1 Initial ventilation conditions, pressure in compartment, P0, and ventilation flow rate, Q0 P0 [Pa]
2
5
20
40
85

Q0 [m s ] 0.006 0.018 0.04 0.06 0.09

600 400

p [Pa]

200 0 Po –2 –5 –20 –40 –85

–200 –400 –600

8.3

Results

8.3.1

Effects of Initial Negative Pressure on Fire Transition

3
1

0

300

600

900

1200

t [sec]

Fig. 8.2 Time series of pressure in compartment, P, for various initial compartment pressures, Po

adjustment was done on the ventilation system during the duration of fire, and thus the negative pressure in the compartment gradually decreased during the experiments. This is due to the increase in pressure loss with combustion products and also to changes in the combustion reactions of fire, especially just after the ignition and extinguishing of the pool fire. The measurements were carried out for τ ¼ 0–1200 s, with data acquisition and A/D conversions having sampling frequencies of 2 Hz and 10 Hz, which is sufficiently large compared with the time responses of the measurement instruments.

The time series of pressure in the compartment, P, for various initial compartment pressures, Po, are presented in Fig. 8.2. For all Po, the pressures, P, initially show rapid increases due to ignition and have positive peak values at τ ffi 10 s. These positive pressures gradually decrease but are maintained for τ ¼ 0 to about 102 s. With the change in sign from positive to negative values, the pressures depict remarkable oscillations. In particular, the oscillations of pressures for Po ¼
40 and
85 Pa have large amplitudes of 200–400 Pa. The time series of air temperature, T, at the inlet and outlet ducts for various initial compartment pressures, Po, are shown in Fig. 8.3. At the inlet duct, for all Po, the temperatures initially increase and have maximum values. The temperature increase coincides with the compartment pressure increase presented in Fig. 8.2. This suggests that the pressure increase is due to ignition, which causes a sudden increase in air temperature in the compartment, which led to reverse flows from the compartment to the outside through the inlet duct. The temperature fluctuations observed after temperature increase period must be related to the oscillations of pressure, as the temperature fluctuations for Po ¼
40 and
85 have large amplitudes. At the outlet duct, the temperatures initially increase for all Po and then gradually decrease with the change in gradient (∂T/∂τ) for Po ¼
2,
5, and
20 Pa, which implies the occurrence of extinction. The temperatures increase slightly with large fluctuations for Po ¼
40 and
85 Pa. The periods for large fluctuations of T at the outlet duct for Po ¼
40 and
85 Pa agree well with those for large temperature fluctuations at the inlet duct and those for large pressure oscillations.

92

Y. Hattori et al.

a 25 Po -2 -5 -20 -40 -85

T [⬚C]

300

200

20 Yo2 [%]

a 400

0

300

600 t [sec]

900

5

1200

0

300

600 t [sec]

900

1200

b 30 Po -2 -5 -20 -40 -85

500 400 300

Po –2 –5 –20 –40 –85

25

Yo2 [%]

b 600

T [⬚C]

Po –2 –5 –20 –40 –85

10

100

0

15

20 15

200

10

100

0

300

600 t [sec]

900

5

1200

Fig. 8.3 Time series of air temperature, T, at inlet (a) and outlet ducts (b) for various initial compartment pressures, Po

The time series of oxygen concentration, Yo2, in the compartment and at the outlet duct for various initial compartment pressures, Po, are shown in Fig. 8.4. These concentrations also show the temporal variations that correspond to those in compartment pressure and inlet and outlet ducts. The fluctuations of concentrations at the outlet duct appear for Po ¼
40 and
85 Pa and τ > about 600 s and 750 s, respectively. In addition, the concentrations decrease with the changes in gradient (∂Yo2/∂τ), and the time of the changes in the gradient agree with the time of changes in the sign of compartment pressure. Notice that the concentrations have initially the same gradients against time in the compartment and at the outlet duct for all Po, and then the concentrations at the outlet become significantly smaller than those in the compartment. They indicate the transition to ventilation-controlled fires with the change of the vertical (zone) structures in the compartment from one layer to two layers. The time series of mass loss of the ethanol,
m, for various initial compartment pressures, Po, are presented in Fig. 8.5. We have also checked the time series of

0

300

600 t [sec]

900

1200

Fig. 8.4 Time series of oxygen concentration, Yo2, in compartment (a) and at outlet duct (b) for various initial compartment pressures, Po

3000 Po –2 –5 –20 –40 –85

2500 2000 –m [g]

0

1500 1000 500 0 –500

0

300

600 τ [sec]

900

1200

Fig. 8.5 Time series of mass loss of ethanol,
m, for various initial compartment pressures, Po

temperature of ethanol in the pan through measurement with thermocouples, which shows the temperature is quite low compared with the boiling point of ethanol.

8

Interaction of a Pool Fire in a Compartment with Negative Pressure Generated by. . .

Table 8.2 Measured fire parameters: peak pressure due to ignition, Pi, start time of pressure oscillation, to, extinction time, te, and mass loss rate of ethanol with steady flame, ms, for various initial compartment pressures, P0 P0 [Pa]
2
5
20
40
85

Pi [Pa] 276 240 170 140 66

τo [s] 211 225 351 610 729

τe [s] 307 292 531 cont cont

ms [gm
2s
1] 7.0 8.3 9.1 10.9 10.9

Here, we should stress that the measurement of the heat release rate was quite difficult in these experiments because oxygen consumption calorimetry is not easily used for a pool fire in a poorly ventilated compartment. Also, the mass loss measured with the electronic balance installed beneath the fuel storage tank gives a spurious peak just after ignition (τ ffi 0 s), which is related to the increase in compartment pressure. The heat release rate, which is roughly estimated by the changes in mass loss,
m, with time, clearly shows the effects of initial compartment pressure on the mass loss rate due to the ventilation-controlled fire. The initial compartment pressures of Po ¼
2,
5, and
20 Pa, which cause extinction due to the lack of oxygen in the compartment, produce a reduction in heat release rate, whereas the fires with Po ¼
40 and
80 Pa preserve the same level of heat release rate during the experiments. The measured fire parameters, peak pressure due to ignition, start time of pressure oscillation, and extinction time for various initial compartment pressures are summarized in Table 8.2. The time-averaged values of the mass loss rate of ethanol for ventilation-controlled fires under steady conditions, which appear before the oscillating flame periods, are also shown in this table. All initial compartment pressures in the present study, P0 ¼
2 to
85 Pa, yield the positive peak pressures due to the ignition that causes the reverse flows at the inlet duct. The positive values decrease with increasing mechanical ventilation. The increasing ventilation simultaneously leads to increases in the mass loss rate of ethanol for ventilation-controlled fires and also to the delay of extinction, implying that the negative compartment pressures with mechanical ventilation alleviate the restriction on air supply at the inlet duct and also delay the extinction due to the lack of oxygen in the compartment. Before the extinction, the flames show the oscillation behavior; the beginnings of the oscillations also depend on the initial compartment pressures, i.e., strong mechanical ventilation causes the late occurrence of oscillating flames. We should stress that two kinds of oscillation processes are observed in this study, as shown in the figures; they have different frequencies and amplitudes, i.e., the large pressures with mechanical ventilation, P0 ¼
40 and
85 Pa, product

93

the oscillating flames with low frequency and the large amplitude. Photos of the transition to a ventilationcontrolled fire with normal (Po ¼
20 Pa) and strong (P0 ¼
85 Pa) oscillation are shown in Fig. 8.6. For Po ¼
20 Pa, initial phase, ventilation-controlled, and oscillating fires are shown at τ ¼ 120, 300, and 360 s; for Po ¼
85 Pa, initial phase, ventilation-controlled, and oscillating fires are shown at τ ¼ 180, 600, and 900 s. Both initial negative pressures yield similar geometry and color of flames during the transition to ventilation-controlled and oscillating fire; thus, these photos give no clue about the transition processes. For example, regardless of the value of P0, all the initial phase fires, which burn with sufficient fresh air, have orange flames; on the other hand, the ventilation-controlled and oscillation fires also have blue flames, indicating a poor oxygen supply.

8.3.2

Temporal Variation of Strong Flame Oscillation

The measurements at τ ¼ 600–1200 s for the initial compartment pressure of
85 Pa are discussed in this section to explore the details of the dynamic process of strong flame oscillation. Temporal variation of differential pressure in the compartment and air temperature in the inlet and outlet ducts is shown in Figs. 8.7 and 8.8. The variations of oxygen concentration in the compartment and the outlet duct are also shown in Fig. 8.9. At τ ffi 700 s, the pressure shows an abrupt change with the peak value of 400 Pa, and, simultaneously, the air temperature at the inlet duct has a peak with the decreasing air temperature at the outlet duct. Also, the oxygen concentration decreases in the compartment2 and increases dramatically at the outlet duct. Similar behavior of pressures, temperatures, and oxygen concentration is also observed at τ ffi 1000 and 1100 s. They indicate that the origin of the oscillating flames is the repetition of ignition and extinction, which was attributed to the change in the flow rate and direction of fresh air at the inlet duct; the ignitions and extinctions also must cause the rapid increases and decreases of compartment pressure, which might produce the pressure oscillation with large amplitudes. Temporal variations in the vertical profile of air temperature along the center line of the fire plume (x ¼ 0) are presented in Fig. 8.10. The temperatures of the fire plume clearly drop for the periods of oscillating flames. These drops generate the vertical temperature profiles under stable stratifications, i.e., the temperature gradually increases with

2

Measurement height from the floor is 0.35 m, which is quite high compared with the height of the ethanol surface at 0.075 m.

94

Y. Hattori et al.

Fig. 8.6 Pictures of initial phase (a), ventilation-controlled (b), and oscillating (c) fire above the pool with normal, Po ¼
20 Pa, (left side), and strong, Po ¼
85 Pa, (right side) oscillation. For Po ¼
20 Pa, initial phase, ventilationcontrolled, and oscillation fires are at τ ffi 120, 300, and 360 s; for Po ¼
85 Pa, initial phase, ventilation-controlled, and oscillation fires are at τ ffi 180, 600, and 900 s

400

600

Location inlet outlet

400 300 T [C]

p [Pa]

200 0

200

–200 100

–400 –600 600

900 t [sec]

1200

Fig. 8.7 Temporal variation of differential pressure, P, in compartment with strong oscillation pressure (Po ¼
85 Pa, τ ¼ 600–1200 s)

0 600

900 t [sec]

1200

Fig. 8.8 Temporal variation of air temperature, T, in inlet and outlet ducts with strong oscillation pressure (Po ¼
85 Pa, τ ¼ 600–1200 s)

8

Interaction of a Pool Fire in a Compartment with Negative Pressure Generated by. . .

95

25 Location compartment outlet duct

Yo2 [%]

20

15

10

5 600

900

1200

t [sec] Fig. 8.9 Temporal variation of oxygen concentration, Yo2, in compartment and outlet duct with strong oscillation pressure (Po ¼
85 Pa, τ ¼ 600–1200 s)

800

T [⬚C]

600

400

200

0 600

z [m] 1.64 1.34 1.04 0.74 0.44 0.14

900 t [sec]

1200

Fig. 8.10 Temporal variation of vertical profile of air temperature, T, along the center line of fire plume (x ¼ 0) with strong oscillation pressure (Po ¼
85 Pa, τ ¼ 600–1200 s)

z, the vertical distance from the ethanol surface. These profiles, which give the opposite sign of vertical temperature gradient compared with a normal fire plume, imply that the oscillating fire plumes with large pressure fluctuations generate no thermal energy with combustion. A typical example of the oscillating sequence of flames just above the pool is presented in Fig. 8.11. The time sequence shows the expansion of blue flames; this dynamic process is significantly different from that of normal oscillating flames. In this study, moving flames away from the pool, which are reported in previous studies, such as [2– 5], were not observed. The flames do not change their location; they do change their shape for oscillation periods. We will discuss the details of the differences in structures between the oscillating flames and moving flames with the measurement of spatial distributions of oxygen concentrations in future studies.

Fig. 8.11 Sequence photos of flames above the pool with strong oscillation pressure (Po ¼
85 Pa, τ ¼ 600) captured at 10 frame s
1

8.4

Discussion and Conclusion

This study examined a pool fire in a mechanically ventilated compartment. Fire experiments were performed with a medium-sized pool fire in a compartment, the dimensions of which correspond to the ISO 9705 room. The liquid fuel and pool diameter were ethanol and 600 mm, respectively, which rapidly increased the compartment pressure just after ignition. The compartment was ventilated at the inlet and outlet ducts with natural and mechanical ventilation systems, which gave negative values of initial pressure in the range of
2 to
85 Pa. The measurements of pressure, mole fraction, and temperature of air in the compartment reveal that the initial pressure strongly affects the transition to a ventilationcontrolled fire. The negative pressure with the ventilation that was much weaker than the pressure increase restrained the air supply with natural ventilation and resulted in extinction due to lack of oxygen in the compartment. As the mechanical ventilation becomes stronger, the negative

96

pressures sustained the air supply and produced the transition to a ventilation-controlled fire without extinction. With the ventilation-controlled fire, two kinds of oscillating flame processes were observed: one was caused by poor oxygen supply, and the other was caused by repetition of ignition and extinction, which was attributed to the change in the flow rate and direction of fresh air at the inlet duct. The latter yielded only pressure fluctuations, but did not generate thermal energy with combustion. This is due to the oscillation being closely related to the dynamic process with the expansion of the flames, which provide the energy in the pressure field. Transition processes observed in the present study partly correspond to those observed in previous studies. The transition to a ventilation-controlled fire with extinction agrees with Utiskul et al. [3] (region 1) and Sano et al. [9]. The oscillating flame with poor oxygen supply might be related to the ghosting fire phenomena [2–5], whereas the flame in the present study stays above the pool during experiments, implying that the present study does not reproduce the flame in region 2 of Utiskul et al. [3]; this might be due to the baffle plate equipped at inlet duct to reduce the interference of inflows or the size of compartment, which is much larger than that of Utiskul et al. [3] and gives the larger distance between the inlet duct and the pool. As for fire modeling, the oxygen concentration yields one-zone structures above the pool for several hundred seconds after the ignition, which is attributed to the wellmixed combustion air without inflow of fresh air. Indeed, the estimated values of mass loss rate with measurements (ffi10 g
2s
1) agree with those with the theoretical conservation equations [1] under the condition of no supply of fresh air. These values support the applicability of theoretical analysis [1] with flammability criteria [13] for predicting the occurrence of extinction due to lack of oxygen in the compartment. The transition to a ventilation-controlled fire creates two-layer structures with both fresh and combustion air in the compartment. For oscillating fires, fire models, such as zone models, must have a capability to represent the temporal variances of vertical profiles of fire parameters in a compartment [3]. Contrary to this, the prediction for oscillating fires caused by repetition of ignition and extinction must be quite difficult because the prediction requires representation for the conversion process from chemical to thermal and physical energies. Also, this process essentially depends on the type of fire source. The present study has focused on a pool fire with a liquid fuel, which is a typical

Y. Hattori et al.

fire event in a plant. The development of a fire model with accumulation of experimental measurements on the flame physics will be of interest in future works. Acknowledgment The authors wish to express their gratitude to Mr. Tadashi Sano of Hitachi, Ltd., for helpful comments on the design of the fire compartment and Mr. Yoshimasa Terada and Takayoshi Mizuno of Ceres, Ltd., for implementing careful measurement in the experiment. The authors are deeply indebted to fruitful discussions on this study by Dr. Koji Shirai and Dr. Yuzuru Eguchi of CRIEPI.

References 1. Quintiere JG (2002) Fire behavior in building compartments. Proc Combust Inst 29:18fa1–18fa193 2. Sugawa O, Kawagoe K, Oka Y, Ogahara I (1989) Burning behavior in a poorly-ventilated compartment fire. Fire Sci Tech 9:5–14 3. Utiskul Y, Quintiere JG, Rangwala AS, Ringwelski BA, Wakatsuki K, Naruse T (2005) Compartment fire phenomena under limited ventilation. Fire Safety J 40:367–390 4. Hu L, Lu K, Delichatsios M, He L, Tang F (2012) An experimental investigation and statistical characterization of intermittent flame ejection behavior of enclosure fires with an opening. Combust Flame 159:1178–1184 5. Tang F, Hu LH, Delichatsios MA, Lu KH, Zhu W (2012) Experimental study on flame height and temperature profile of buoyant window spill plume from an under-ventilated compartment fire. Int J Heat Mass Transfer 55:93–101 6. Delichatsios MA, Silcock GWH, Liu X, Delichatsios M, Lee Y-P (2004) Mass pyrolysis rates and excess pyrolysate in fully developed enclosure fires. Fire Safety J 39:1–21 7. Floyd JE, Hunt SP, Williams FW, Tatem PA (2005) A network fire model for the simulation of fire growth and smoke spread in multiple compartments with complex ventilation. J Fire Protect Eng 15:199–229 8. Torero JL (2013) Special issue on PRISME-Fire safety in nuclear facilities. Fire Safety J 62:79 9. Sano T, Shirai K, Hattori Y, Suto H, Tshichino S (2013) Comprehension of limited ventilated fire behavior and study on fire prediction method in an enclosed space. Trans AESJ 12:32–42. (in Japanese) 10. Hattori Y, Matsuyama K, Suto H, Onuma E, Okinaga S (2014) Turbulence measurements in a ventilation-controlled poof fire. Proc 16th Int Symp Flow Vis 11. Hattori Y, Matsuyama K, Suto H, Onuma E, Okinaga S (2014) Entrainment process in the vicinity of pool fire under ventilation condition. Proc 15th Int Heat Transf Conf 09377 12. Bryant RA (2011) Evaluating practical measurements of fireinduced vent flows with stereoscopic PIV. Proc Combust Inst 33:2481–2487 13. Floyd JE, McGrattan KB (2009) Extending the mixture fraction concept to address under-ventilated fires. Fire Safety J 44:291–300

9

Pool Fire Behavior in a Small and Mechanically Ventilated Compartment Alexis Coppalle, Alvin Loo, and Philippe Aine´

Abstract

Several fire tests were conducted with heptane and in a chamber of 1 m3, in which the smoke exhaust was done by a fan. The aim was to analyze the rate of mass loss of fuel and the behavior of the flame in the case of under-ventilated fires. For all tests, the oxygen concentration in the flow which feeds the base of the flame decreases continuously. However the mass loss rate increases or decreases depending on the ventilation in the compartment. This can be explained by the respective contribution of the flame and of the smoke layer in the heat transfer toward the liquid fuel. If the ventilation of the room is sufficient, a periodic phenomenon of propagation/expansion of the flame is also observed, with a characteristic frequency that has been determined by an analysis of the flame images and which is close to 1 Hz. Keywords

Pool fire
under-ventilated flame
Smoke
Radiation

Nomenclature

Greek Symbols

D Fs-F

αF eF

L MLR MLRfree T Yo Yo,/

Pool fire diameter View factor between the fuel surface and the smoke layer Heat of gasification (kJ/g) Mass loss rate (g/m2/s) Mass loss rate with free-burning conditions Temperature Oxygen concentration (%volume) Oxygen concentration in ambient air (%volume)

Subscripts F s

9.1

A. Coppalle (*) National Institute of Applied Sciences, UMR 6614 CORIA, rue de l’Universite, St Etienne du Rouvray 76801, France e-mail: [email protected] A. Loo
P. Aine´ AREVA, Paris, France

Absorptivity of the liquid fuel Emissivity of the liquid fuel

Fuel Smoke

Introduction

The compartmented and under-ventilated fires are now an important topic, as they can occur in habitat or in industry. They may be encountered when the ventilation of compartments is strongly controlled, as in the new buildings or in the confined spaces of some industries, in particular in the nuclear industry. These conditions can lead to scenarios during which the fire consumes quickly the air available in the compartment. The combustion is then made with a lack of air and

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_9

97

98

A. Coppalle et al.

the flame extinction may also occur. In these cases, it can be emitted a significant quantity of combustible vapors, mainly after extinction because of the thermal inertia of heated combustibles and also due to the convective and radiative fluxes that remain high on the surfaces of materials. The excess of combustible vapors may trigger a deflagration if a fresh air is introduced suddenly into the room. This thermal accident represents a serious hazard that people in the vicinity might be exposed as well as the emergency services. It is important to identify the key phenomena that control such scenarios, in order to determine the mass loss rate of combustibles and the flame behavior inside the compartment. The conditions that lead to the extinction of the flame must also be known. Several tests with heptane were carried out at small scales in a compartment of 1 m3, ventilated by a fan installed upon the smoke exit, which allows to control and to adjust the air inlet flow rate. The experimental setup will be described. Then the measured values of the mass loss rate, the temperature at different heights, and the oxygen concentration will be given and discussed. A simple model for the mass loss rate is presented in order to estimate respective contribution of the flame and of the smoke layer in the heat transfer toward the liquid fuel. For high air inlets, a periodic phenomenon of propagation/expansion of the flame is also observed, with a frequency that has been determined by a flame image analysis, which will be presented.

9.2

Experimental Setup

The compartment is a 1 m3 chamber of equal sides. Its interior walls, ceiling, and floor are covered with 25 mm thick refractive ceramic fibers for thermal insulation. Through the air inlet on a sidewall near the floor, air can enter the compartment in one of two ways (see Fig. 9.1):

Fig. 9.1 Experimental setup

– A 1 cm diameter duct with a hot-wire probe to measure airflow – An iris diaphragm across which pressure difference measurements allow the determination of airflow Smoke exits through a 4.5 cm diameter vent at the ceiling of the compartment on the side opposite to the air inlet. The ventilation of the chamber can be controlled using an electric fan fixed upon the smoke exit. The fuel mass, vertical gas temperature profile at a corner of the compartment and below the smoke exit, pressure inside the compartment, oxygen concentration near the base of the flame, and the airflow into the compartment were measured. Two fuel pans of 111 mm inner diameter made of borosilicate glass were used for all tests, and these were installed side by side. The fuel mass is measured in real time by a Mettler Toledo weighing scale with a resolution of 0.1 g. The weighing scale was placed at the center, under the fire compartment to protect it from heat. A ceramic cylindrical shaft and water seals were used to connect the fuel pan to the scale. This sealing system is similar to the one used in the study by Utiskul et al. [1]. The vertical gas temperature profiles are measured by two arrays of five type-K thermocouples. Oxygen concentration at approximately 3 cm away from the flame base was measured by a Testo 350 gas analyzer. A pressure gauge tube was inserted into the compartment at floor level to measure pressure variations inside the compartment. Data from the precision scale, the thermocouples, and the airflow by the iris diaphragm were taken at a scan rate of 300 ms. The gas analyzers and the hot-wire probe provided values every 1 s. The varying fire power in some tests with high ventilation rates produced some pressure variations in the compartment. Since the force felt by the cylindrical shaft can be transmitted downward to the precision scale, such pressure variations

9

Pool Fire Behavior in a Small and Mechanically Ventilated Compartment

within the compartment can affect fuel mass loss measurements. Corrections have therefore been made to the fuel mass measurements.

9.3

Results

Fig. 9.2 Experimental results for oxygen concentration Yo at flame base. Vertical bars indicate the extinction time

oxygen concetration at the base of the flame (%)

Tests of n-heptane pool fires have been carried out for different conditions of ventilation. This one is defined by the airflow rate at the inlet at the ignition time. It was equal to 2.6, 1.3, 0.075, and 6 l/s for tests 1–4, respectively. The mass loss rate is expressed per unit area of the fuel pan. The oxygen concentrations, in dry mixture, at flame base and close the pan is reported in Fig. 9.2 and the fuel mass loss rates in Fig. 9.3, as a function of time. The first figure shows that oxygen concentration is continuously decreasing close to the flame, because the air inlet flux is not sufficient to compensate the oxygen consumption by combustion. For tests 1 and 2, the flame left the fuel pan and has been

99

attached again close to the floor between the vertical wall and the liquid pan (to the left in Fig. 9.1). So the oxygen probe was no more in the vicinity of the flame, and this is why oxygen concentration dropped so low, at about 10 %. The mass loss rate MLR behavior is quite different. Our experiments show that the MLR variations depend on the compartment ventilation rate. For the test 3 with the lowest ventilation rate, it decreases with time, for the test 2 it remains constants, and in contrary, for the high ventilation rates (tests 1 and 4), it increases. These variations of the mass loss rate will be analyzed in the next section according to the respective heat fluxes of the flame and the smoke layer on the combustible. For the two fuel pans of 111 mm inner diameter, the “free-burning” mass loss rate is equal to 11.6 g/m2/s [1]. This value would be observed if the pool fire was not enclosed in a compartment. This could have been observed at the beginning of each test when the effect of radiation of the smoke layer is small. This is not the case during the tests performed in this

22 Test1

20

Test2 Test3

18

Test4

16 14 12 me (s) 10

0

100

200

300

400

500

20

mass loss rate (g/m2/s)

Fig. 9.3 Experimental results for fuel mass loss rates. Vertical bars indicate the extinction time

15

test1 test2 test3

10

test4

5

0

time (s) 0

100

200

300

400

500

100

A. Coppalle et al.

100 90 80 70 60 50 40 30 20 10 0

Height (cm)

Fig. 9.4 Vertical temperature profiles under the smoke exhaust (up) and at the corner (down) for several times after ignition and for test 2

Height (cm)

study as shown in Fig. 9.3. With the “free-burning” value of 11.6 g/m2/s, the stoichiometric airflow rate can be calculated from the global combustion reaction of the heptane in air, and the result is 3.3 l/s. By comparing this theoretical value to those used for each experiment, it is clear that the tests 2 and 3 are strongly under-ventilated, test 1 slightly ventilated, and test 4 over-ventilated. However, Fig. 9.2 shows that the flames have not been fed by fresh air during all tests. It should be noted that, for all these tests, the flames have been extinguished due to oxygen depletion inside the chamber, and this occurred before the fuel has been totally consumed. The time of extinction is 250 s, 400 s, 350 s, and 380 s, respectively, for tests 1–4. For test 4, the MLR parameter shows oscillations; this point will be discussed later. Figure 9.4 shows, for the test 2, the vertical temperature profiles at two locations in the chamber, under the smoke exit and at one corner (right), and at different times after ignition. There is stratification inside the compartment, with a temperature increase from the floor up to about 30 cm height and above a hot-gas homogeneous layer, which fills the major part of the chamber, between the flame and the ceiling. At each height above the floor, the temperature

100 90 80 70 60 50 40 30 20 10 0

difference between the two thermocouple locations is small, less than 30  C. For the other tests, the profiles are similar, indicating that the hot-gas layer begins always at the same height above the floor of the chamber (at about 30 cm height). However, the temperature of the hot gases varies between the different tests. This can be seen on Fig. 9.5 which represents the average temperature between 0.3 m above the floor and the ceiling. It is important to remember that the temperature of the smoke layer is higher when the ventilation rate is increased. During a test, the oxygen is continuously decreasing into the compartment; the behavior and the shape of the flame are changed. For under-ventilated regimes (tests 2–3), flames look like the examples shown in Fig. 9.6. At the beginning of the test (left picture), the flame attachment is stable and conical; pulsations are observed in the flame zone at a frequency of about 3–4 Hz. This process of “puffing” is well known and is due to instability of the fluid inside the flame zone [2]. Then (middle picture) the flame becomes unstructured; its height and luminosity are reduced. Before the extinction (right picture), the flame can move from the pan (only in test 2), persisting for a few seconds giving what is called a “ghosting flame.”

under the smoke exhaust, Test 2 50s 100s 150s 200s

0

50

100

150 200 250 Temperature (°C)

300

350

Box corner, Test 2 50s 100s 150s 200s

0

50

100

150 200 250 Temperature (°C)

300

350

Pool Fire Behavior in a Small and Mechanically Ventilated Compartment

Fig. 9.5 Average temperature between 0.3 m above the floor and the ceiling, as a function of time

101

350 Average Temperature of the smoke layer (°C)

9

300 250 200

Test1

150

Test2

100

Test3

50

Test5

0 0

50

100

150

200

250

300

350

400

Time (s)

Fig. 9.6 Images of the flame during the test 2, left at 50 s, middle at 150 s, and right at 230 s just before extinction

For the slightly under-ventilated regime and the wellventilated regime (tests 1 and 4), the behavior and the shape of the flame are different during the period before extinction. Some pulsations become more abrupt, leading to a rapid expansion of the flame zone as seen in Fig. 9.7. Following this pulsation, the flame becomes more luminous and is like a fire ball which rises driven by the effects of buoyancy.

9.4

Discussion

The compartmented fires fed by natural ventilation (without a fan) is well known. In particular, the under-ventilated cases and the flame behaviors have been analyzed in several studies at reduced scales [3–8]. In these ones, the ventilation rate is not measured, but it is determined with the ventilation parameter, defining the product A√H, where A is the aperture surface and H is its height. The present study is also performed at small scales; however, the ventilation is

mechanical. So it is difficult to compare our ventilation conditions to those used in the previous works. Another important difference is that the present study has been focused on the cases for which there has been extinction. However, some studies [3–5] show a number of similar features to those observed in this study: There exists a lower limit for the ventilation (in volume per second) below which there is no sustainable combustion and a flame extinction occurs; near this limit, oscillating flames are observed; there is a range of values of the ventilation for which the mass flow rate is greater than the “free-burning” value. In addition, we have also observed the phenomenon called “ghosting flame” as in the works of Sugawa et al. [9] and Pearson [10]. However, in these studies and unlike our experiments, the fuel pan was put in the smoke layer, and the fuels were different. They observe large horizontal structures of flame which meander in the compartment, while we see several vertical flames detached from the pans and near the floor, as shown in Fig. 9.6. Figure 9.3 shows that, after ignition, the fuel mass loss rate increases for the high ventilation rates (tests 1 and 4), becoming greater than the “free-burning” value (11.6 g/m2/ s). It remains constant for the moderate ventilation rate (test 2) and decreases for the lowest value (test 3). The mass loss rate depends on the heat flux of the flame toward the liquid pool. This flux is a function of the flame temperature and so of the heat release rate and consequently of the oxygen concentration available in the flow which fed the flame. This oxygen concentration near the flame is continuously decreasing for all tests. This can explain the MLR decrease during test 3, but the trends observed for the other tests 1, 2,

102

A. Coppalle et al.

Fig. 9.7 Images showing an expanding flame during a well-ventilated test; the time between images is 33 ms

and 4 are not fully explained by the observed variation of the oxygen concentration close to the flame. Another phenomenon also influences the mass loss of the combustible. The liquid is heated and vaporized due to the received fluxes originating from the flame but also from the smoke layer (and also from the wall) [11]; these two fluxes may have the same order of magnitude [12]. By taking into account the temperature of the smoke layer, as reported in Fig. 9.5, it can be expected that the radiative flux originating from the smoke layer is greater for tests 1 and 4, than for tests 2 and 3. This may explain the observed trends for the mass loss that are different during tests 1–4 and 2–3. The mass loss rate for under-ventilated and confined fires can be approximated by the sum of two terms [5]:
  σ aF FsF T 4s  eF T 4F Yo ð9:1Þ MLR ¼ kMLRfree þ L Y o, / where MLRfree is the free-burning rate per unit area, Yo the local oxygen concentration at the flame base, Yo,1 the oxygen concentration of ambient air, σ the Stefan-Boltzmann constant, Ts the smoke (and wall) temperature, and TF and L the temperature and the heat of gasification of the fuel. The parameter k is a correction factor, αF and eF the absorptivity and the emissivity of the fuel, and FsF the view factor between the fuel surface and the smoke layer. These parameters can be determined and optimized by comparing the MLR-calculated values to the experimental results. In (Eq. 9.1), it is assumed that the flame flux (convective and radiative) is proportional to the oxygen concentration near the flame base, and the smoke emissivity is equal to one. Utiskul [5] has assumed that the parameters k, FsF, αF, and eF are also equal to one. In that case and using the O2 and temperature experimental values given in Figs. 9.2 and 9.4, the previous equation gives results that exceed much the measured values. Therefore, the previous adjustable parameters have to be changed. It is assumed that the absorptivity αF and emissivity eF factors are equal. So the

comparison between experimental and calculated values of MLR shows that the best agreement is obtained for k, αF, and FsF equal 0.55, 0.8, and 0.7, respectively. The last value is close to the theoretical one corresponding to the view factor between a small surface and a 1 m2 square surface located at 35 cm above. These results have to be considered as a first attempt to explain the different trends of MLR observed in our compartment, and further works are needed to generalize them. However, they show that, in the case of a stratification inside a compartment, the two parameters k and FsF are not equal to one, and they have to be considered. Oscillating flames have been observed in previous studies [3–5]; however, the expanding flames, as the one in Fig. 9.7, have not been extensively studied up to now. The main features of this phenomenon are as follows: it becomes stronger when the ventilation is increased; it occurs in the lower part of the flame, close to the pan (see Fig. 9.7); it begins with a short deflagration which is followed by a longer period of latency; and it is cyclic with a period of about one second. These flames which expand quickly outside the fuel pan seem to correspond to the propagation of a premixed flame. However this occurs in the vicinity of the diffusion flame located just above the fuel pan. The juxtaposition of a diffusion flame and of a premixed flame represents a complex problem. In some cases, this has been called “triple flames.” It has been shown that they intervene in the stabilization of lifted non-premixed flames [13] or in turbulent combustion with large strain rates, when extinction holes develop in the flame surface [14]. Close to the diffusion flame edge, the competition between the flow velocity and the premixed flame propagation with respect to the unburned mixture can lead to extinction, propagating or “standing triple flame” regimes [15]. In the present study, the speeds above the fuel pan and throughout the test chamber are low, and the strain rate inside the diffusion flame is not as high as in the previous studies. However, it is suspected that some similarities exist. One important issue is to know how some premixed gas can be produced in the vicinity of the pan and of the diffusion

9

Pool Fire Behavior in a Small and Mechanically Ventilated Compartment

flame. An explanation is proposed and it is illustrated by the scheme reported in Fig. 9.8. In the central zone of the test chamber, the hot plume of the flame drives an upward flow and a recirculation of smokes laterally and downward. This induces a vitiated flow toward the center, which is mixed with fresh air coming from the inlet and which feeds the flame at its

Fig. 9.8 Schematic view of the flow inside the chamber Fig. 9.9 Frequency spectrum of the flame pulsations versus time for test 4. At each time, normalized values on the bottom figure are obtained by dividing the values given in the upper picture by their maximum

103

base. This is suggested by the oxygen concentration near the pan which is continuously decreasing during the test. Due to this lack of oxygen, all the fuel vapors, emitted by the liquid vaporization, are not consumed in the flame zone and are accumulated close or under the flame base. So in the mixing which feeds the flame, represented by the hatched zone in Fig. 9.8, there are simultaneously air and combustible vapors, which can ignite if the combustible concentrations are greater than the minimum of inflammability. The ignition source is the diffusion flame, and the propagation occurs mainly in the horizontal direction, from the center of the chamber to the sidewalls. An examination of the images on Fig. 9.7 can give a rough value of the speed of the propagation. It is found equal to about 0.2 m/s, which is the same order of magnitude as a premixed flame speed for a poor richness mixture [16]. The flame pictures, as in Fig. 9.7, have been taken by a camera that records at 30 frames/s. Each image has been converted to black and white with a resolution of eight bit, and values of all pixels on the image have been added. The Fourier transform of these summed values is performed over a period of 20 s. By shifting this period of 1 s, one obtains the frequency spectrum of the flame pulsations versus time. An example of such spectrum is shown in Fig. 9.9 for test 4. Normalized values at the right are obtained by dividing, at each time t, the values given at the left by their maximum;

104

A. Coppalle et al.

this procedure is allowing to highlight the most important frequency. There clearly exists for test 4 a beat frequency, which is about 0.8 Hz. This frequency is varying at the beginning and the end of the test, but remains constant during the major part of the test. One can suspect that this periodic pulsation corresponds to the so-called “puffing” phenomenon which has been described in detail for pool fires [17, 18]. For pool fires burning in an open environment, it has been noted that there are characteristic pulsations of the flame near the burner corresponding to formation of alternative “necks” and “bulges.” The pulsation frequency ( f ) is well correlated to the pool diameter (D); a power law fits the experimental data: f / D0.5. With the equivalent diameter of the present study (15.7 cm), f is equal to 4 Hz, much higher than the value shown in Fig. 9.9. That means the phenomena are different. For the non-compartmented flames, the “puffing” is generated by instabilities occurring inside the flame zone. For the flames observed in the ventilated compartment, the oscillations are due to a coupling between the heat release rate, the pressure, and air inlet inside the compartment. Once the propagating flame is ignited, as shown in Fig. 9.7, there is a new and fast heat release inside the compartment, and the pressure increases quickly. Then the mechanical evacuation brings back more slowly to their initial values the temperatures, the pressure, and the composition of the mixtures inside the compartment, in particular the amount of fuel vapor and oxygen near the base of the flame. Other experiments, not described here, have shown that the frequency is slightly increased if the ventilation rate increases.

9.5

Conclusion

Several tests were conducted with heptane and at a reduced scale, in a chamber of 1 m3 where the ventilation was mechanically controlled. The different ventilation rates were chosen in order to analyze cases of under-ventilated fires with extinction. Some similarities have been observed with other studies carried out at reduced scales but with natural ventilation. The oxygen concentration in the gas flow which feeds the base of the flame decreases continuously, oscillating flames are observed, and the mass flow rate increases and can be greater than the “free-burning” value for the cases slightly underventilated or well ventilated. However, the mass loss rate of the combustible remains constant or decreases if the ventilation in the room is strongly reduced. A phenomenon of propagation and expansion of the flame is also observed. It suggests the presence of a premixed zone near the fuel pan, which can ignite at regular periods. This phenomenon is strengthened for the high ventilation rates. Its characteristic frequency is equal to about 0.8 Hz for the best ventilated test. This value is smaller than the one

observed for the “puffing” pulsations. Further studies are necessary in order to better determine, around the flame and the pan, the conditions of occurrence of the phenomenon and to better characterize this premixed gas deflagration. Acknowledgment This work is supported through a Phd sponsorship by AREVA NC. The experimental apparatus is financed by the research group MRT of the Haute-Normandie region.

References 1. Babrauskas V (1983) Estimating large pool fire burning rates. Fire Technol 19–4:251–261 2. Wu Y-C, Wu X-C, Cen K-F, Lu S-X, Zhang J-Q (2012) Novel methods for flame pulsation frequency measurement with image analysis. Fire Technol 48:289–403 3. Hisahiro T, Kazuo A (1981) Critical phenomenon in compartment fires with liquid fuels. Eighteenth Symposium (International) on Comb, pp 519–527 4. Kwang I, Hideo O, Yoichi U (1993) Experimental study on oscillating behaviour in small-scale compartment fire. Fire Saf J 20(4):377–384 5. Utiskul Y, Quintiere JG, Rangwala AS, Ringwelski BA, Wakatsuki K, Naruse T (2005) Compartment fire phenomena under limited ventilation. Fire Saf J 40:367–390 6. Snegirev AY, Makhviladz GM, Talalov VA, Shamshin AV (2003) Turbulent diffusion combustion under conditions of limited ventilation: flame projection through an opening. Combust Explosion Shock Waves 39:1–10 7. Chamichine AV, Makhviladze GM, Oleszczak P, Yakush SE (2007) Flame exhaust from pool fires in a small-scale compartment with a single opening. In: Proceedings of the 5th international seminar on Fire and Explosion Hazards, pp 338–348 8. Makhviladze GM, Chamichine AV, Yakush SE, Oleszczak PT (2010) Experiments and simulations of burner and pool fires in under-ventilated compartments. In: Proceedings of the 6th international seminar on Fire and Explosion Hazards, pp 161–172 9. Sugawa (1989) NIST – burning behavior in a poorly-ventilated compartment fire: ghosting fire 10. Pearson A, Most JM, Drysdale D (2007) Behaviour of a confined fire located in an underventilated zone. Proc Combust Inst 31:2529–2536 11. Quintiere JG (2006) Fundamentals of fire phenomena. Wiley, New York. ISBN: 978-0-470-09113-5 12. Nasr A, Suard S, El-Rabii H, Gay L, Garo JP (2011) Experimental study on pyrolysis of a heptane pool fire in a reduced-scale compartment. In; 7th mediterranean combustion symposium 13. Takahashi F, Schmoll F, Katta VR (1998) Attachment mechanisms of diffusion flames. Proc Combust Inst 27:675–684 14. Vervisch L, Poinsot T (1998) Attachment mechanisms of diffusion flames. Annu Rev Fluid Mech 30:665–691 15. Santoro VS, Linan A, Gomez A (2000) Propagation of edge flame in counterflow mixing layers: experiments and theory. Proc Combust Inst 28:2039–2046 16. Qiao L, Gan Y, Nishiie T, Dahm WJA, Oran ES (2010) Extinction of premixed methane/air flames in microgravity by diluents: effects of radiation and Lewis number. Combust Flame 157:1446–1455 17. Hamins A, Yang JC, Kashiwagi T (1992) An experimental investigation of the pulsation frequency of flames. Symp Combust 24 (1):1695–1702. doi:10.1016/S0082-0784(06)80198-0 18. Cetegen BM, Ahmed T (1993) Experiments on the periodic instability of buoyant plumes and pool fires. Combust Flame 93:157–184. [2] Drysdale D (1985) An introduction to fire dynamics. Wiley, Chichester, p 146

Discussion on Heat Lost Through Solid Boundaries in Modelling Atrium Fires Under Mechanical Exhaust

10

Liang Yi, Danyang Sun, Yuanzhou Li, Ran Huo, Wan-Ki Chow, and Nai-Kong Fong

Abstract

Mechanical smoke exhaust systems are usually required in large atria in the Far East. Systems are designed based on empirical expressions in the literature. Most of those design equations were derived from a two-layer zone modelling approach under some geometrical restrictions. Full-scale burning tests on mechanical smoke exhaust system were carried out. From these experiments, it was found that there are significant differences between the measured results in some of the tests and those calculated from the design equations using a two-layer approach. Mass exchange through the smoke layer and cool air layer was identified earlier to be a key point under some conditions. But heat lost through the solid boundaries is another factor as pointed out recently. This point will be further discussed in this paper. A simple model based on a two-layer approach reported earlier will be applied to justify those design equations and experimental results. # Springer 2015. Selection and peer-review under responsibility of the Asia-Oceania Association for Fire Science and Technology. Keywords

Mechanical exhaust
Smoke layer temperature
Heat lost to solid boundaries
Atrium

Nomenclature A Aw

Cp

Plane area of fire room (m2) Area of convective heat exchange between the smoke layer and the walls and ceiling of the fire compartment (m2) Specific heat of air under constant pressure (kJkg1K1)

H hw m_ c m_ e m_ i m_ p

L. Yi
D. Sun Institute of Disaster Prevention Science and Safety Technology, Central South University, Changsha, Hunan Province, China Y. Li
R. Huo State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei, Anhui, China W.-K. Chow (*)
N.-K. Fong Department of Building Services Engineering, The Hong Kong Polytechnic University, Hong Kong, China e-mail: [email protected]; [email protected]

ms Q_ : Q_ c Q_ loss t Ta, Ts

Height of the fire compartment (m) Convective heat transfer coefficient between the smoke layer and the walls and ceiling (kWm2K1) Mass entrainment rate of air at the layer interface (kgs1) Mass flow rate of smoke extraction (kgs1) Mass flow rate of intake air through floor air inlet due to exhaust (kgs1) Mass flow rate of plume at the smoke layer interface (kgs1) Mass of the smoke layer (kg) Heat release rate of fire (kW) Convective portion of heat release rate of the fire (kW) Heat lost of smoke layer (kW) Time (s) Temperature of ambient and smoke (K)

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_10

105

106

L. Yi et al.

Tst, ΔTst

Average absolute temperature and temperature rise of smoke layer at quasi-steady vented stage (K) Volumetric flow rate of smoke extraction (m3s1) Height from top of fuel to smoke layer interface, clear height (m) Clear height at quasi-steady vented stage (m) Density of ambient air and smoke (kgm3) Density of smoke at quasi-steady state (kgm3) Total heat loss factor from the smoke layer to the fire compartment boundaries

V_ e z Zst ρα, ρs ρst χ

10.1

Introduction

Mechanical smoke exhaust systems (or dynamic smoke exhaust systems in some local codes) are required in many countries in the Far East [1]. Full-scale burning tests on mechanical smoke exhaust systems in atrium fires were carried out and reported [2]. In studying the smoke-filling process in an atrium, a two-layer approach is usually employed [3]. For a fire which occurs in a room, a hot smoke layer will be formed eventually after some time. In the two zone approach, the room concerned is divided into two homogeneous layers, an upper smoke layer and a lower cool air layer. The smoke layer will descend as air that is supplied from the plume. If there is a ceiling mechanical exhaust system with floor air inlet as shown in Fig. 10.1, smoke will be extracted upon operating the exhaust fan from the hot layer. The smoke layer interface height might be kept constant when the smoke entering the hot layer is equal to the mass extraction rate. There might be mixing across the smoke layer interface. This point had been studied separately [4], and the mass flow rate across the smoke layer interface is found to be about 30 % of the exhaust rate under higher exhaust conditions (>8 air changes per hour). The two zone approach is not valid when there is mass flow across the smoke layer interface under high exhaust rates. By modifying the key . me

Smoke exhaust

Air intake . mi Fig. 10.1 A two-layer zone model

. mp

10.2

Design Equations in NFPA-92B

Equations for calculating the positions of the smoke layer interface height Zst upon operating the mechanical smoke extraction system at an exhaust rate V_ e appear in NFPA- 92B [3]. In a big hall of height H and area A, the mass of the smoke layer ms of density ρs at height Z is

Smoke layer . mc

equation in a two-layer zone model with a constant mass flow rate across the interface under high exhaust rates, the predicted smoke layer temperature and interface height agreed better with the experiments. However, heat lost through the solid boundaries is another factor [2, 5]. All these arguments suggested that experimental fullscale burning tests on mechanical smoke exhaust system in an atrium [4, 6] are necessary. The effect of exhaust rates on the smoke temperature and layer interface height should be investigated. A big burning hall of length 22.4 m, width 12.0 m and height 27.0 m was built [7] in China for studying atrium fires. A series of burning tests with pool fires of heat release rate from 1.8 to 5 MW and mechanical exhaust rates from 3 to 16 air changes per hour (ACH) were carried out over a period of 8 months. The fire burnt for about 10 min in each test. The minimum clear height and maximum average smoke layer temperature rise were deduced from measuring the vertical temperature profiles by two thermocouple trees in the atrium [7] and visual inspection. On each thermocouple tree, there were 27 type-K thermocouples placed at intervals of 1 m. The highest thermocouple was at the ceiling. The smoke layer interface height (or clear height) was deduced from the vertical temperature profile measured from the thermocouple trees located in the atrium by the N-percentage rule [8, 9]. In this rule, the smoke layer interface is the position where the temperature rise dropped to N % of the maximum temperature rise. In large spaces with a relatively small fire, the temperature rise of smoke is low. Using a small N in the N-percentage rule, the temperature rise after multiplying by N is still too low to be recorded by the thermocouples. Therefore, the smoke layer interface was determined at the position where the temperature rise started to be larger than 30 % of the highest temperature rise along the vertical direction [4].

ms ¼ Aρs ðH  ZÞ

ð10:1Þ

H Z

Conservation equation of ms gives dms ¼ m_ p þ m_ c  m_ e dt

ð10:2Þ

In the above equations, m_ p is the mass flow rate of the plume up to the smoke layer interface; m_ e is the mass flow

10

Discussion on Heat Lost Through Solid Boundaries in Modelling Atrium Fires Under. . .

rate of exhaust; and m_ c is the mass entrainment rate of air at the layer interface. Applying the ideal gas law to the hot smoke layer and neglecting its pressure variation which is small compared with the absolute ambient pressure [10], ρs T s ¼ ρa T a ¼ ρT  353 kgKm3

ð10:3Þ

The volumetric exhaust flow rate V_ e is specified in designing most of the mechanical exhaust systems. The mass flow rate m_ e can be calculated in terms of the smoke density at the exhaust outlet ρs from: m_ e ¼ V_ e ρs

ð10:4Þ

Under a ‘quasi-steady’ venting condition with a quasisteady smoke layer of temperature Tst and interface height Zst, conservation equations of mass and energy can be written as: m_ p þ m_ c  m_ e ¼ 0

ð10:5Þ

and
 CP m_ p þ m_ c ðT st  T a Þ ¼ Q_ c ð1  χ Þ

ð10:6Þ

where χ is the total heat loss factor from the smoke layer to the fire compartment boundaries. The average smoke layer temperature rise ΔTst (in this way, density of smoke can be taken as density of air) under a quasi-steady state due to Q_ c with a roughly constant clear height is given by:
 ΔT st ¼ ð1  χ ÞQ_ c = ρa CP V_e

The results of Zst and ΔTst calculated from NFPA-92B under experimental conditions and taking χ as 0.65 from the experiments are calculated. Equation (10.7) suggests that ΔTst under quasi-steady venting is proportional to Q_ c and inversely proportional to V_ e . However, experimental measurement reported earlier before did not support this. A clear smoke layer was not observed under some conditions in those burning tests. A possible explanation for such deviation is due to the mass flow across the layer interface, i.e. m_ c might not be zero.

10.3

ð10:7Þ

Z st ¼

c

Self-developed model NFPA 92B Experiments with 2 MW fire Experiments with 3 MW fire Experiments with 5 MW fire

!3=5 ð10:10Þ

80

DTst/ ⬚C

ð10:9Þ

Substituting Zst to Z in the above equation, and combining with Eqs. (10.4, 10.5, 10.6, 10.7, and 10.8) with a proper expression of m_ c , the values of Tst and Zst can be determined. For example, taking m_ e ¼ m_ p by neglecting m_ c , the quasisteady clear height Zst is V_ e ρst  0:0018Q_ c 0:071Q_ 1=3

100

ð10:8Þ

The mass entrainment m_ p relations for an axisymmetric plume at height Z above the luminous flame height in NFPA92B is 5=3 m_ p ¼ 0:071Q_ 1=3 þ 0:0018Q_ c c Z

A Simple Two-Layer Zone Model

A simple fire scenario as shown in Fig. 10.1 in a room with constant cross-sectional area is considered. A zone model was developed with theory reported elsewhere [5]. The predictions of the model are compared to data from recent experiments in an atrium. The convective portion of the total heat release rate of the fire is taken as 0.7, and the average heat transfer coefficient hw is taken as 0.025 kWm2 K1 from empirical data. The mass entrainment rate of air at the layer interface m_ c was taken as 30 % of m_ e . The steady clear height and smoke temperature were calculated at different volumetric exhaust rates and different heat release rates of fire with this adjustment. The predicted steady clear heights by the NFPA equation and zone model agreed well with experiments after including m_ c . As shown in Fig. 10.2, the NFPA smoke temperature equation is suggested to improve [2] by including heat lost to solid boundaries.

From the equation of state given by (10.4), smoke density at quasi-steady state ρst is ρst ¼ 353=ðΔT st þ T a Þ

107

60

40

20

0 0

2

4

6

8

10 12 ACH

14

16

18

20

22

Fig. 10.2 Comparison of predicted and experimental smoke layer temperature

108

L. Yi et al.

Twi

Ts

Two

1/(hciAw)

δ/(λAw)

Ta

1.0

1/(hcoAw)

2 MW 3 MW 5 MW

0.9

Fig. 10.3 Diagram of heat transfer through the solid boundaries with thermal resistances χ

0.8

0.7

10.4

Heat Lost Through Solid Boundaries

Smoke layer temperatures obtained using equations in NFPA 92B are different from the measured results. Values predicted by the zone model agreed better with experiments. A possible explanation might be due to the heat lost to solid boundaries. Heat lost of the smoke layer Q_ loss in NFPA-92B [3] is taken as a fraction χ of the convective heat release rate of the fire Q_ c , i.e. Q_ loss ¼ χ Q_ c

ð10:11Þ

The values of χ are constant, varying from 0 to 1, and taken as 0.65 in this study. In an atrium fire, smoke temperature is not too high due to the large space volume. Radiation heat flux and the fraction of radiative heat lost to solid boundaries would be smaller than those in a room fire. Heat lost from the smoke layer first transfers to the inner surface of the walls and ceiling by convection, then to the outer surface of the atrium by conduction and then to the ambient air by convection. Similar to Ohm’s law, the overall heat transfer from the smoke to the ambient air can be modelled as a ‘series circuit’ by introducing thermal resistance into the heat transfer procedure at each surface of the atrium and inside the walls and ceiling, as shown in Fig. 10.3. The heat loss rate of the smoke through the solid boundaries of the atrium can be expressed as: Q_ loss ¼

1 hci Aw

Ts  Ta þ λAδw þ hco1Aw

ð10:12Þ

where hci is the convective heat transfer coefficient between the inner surface of the atrium and the smoke, hco is the convective heat transfer coefficient between the outer surface of the atrium and the ambient air and δ and λ are the thickness and conductivity of the solid boundaries, respectively. In the zone model for atrium fire, heat lost was calculated through an overall heat transfer coefficient hw which depends on the convective heat transfer coefficient between the surfaces of the solid boundaries and the surrounding gases, thickness and thermal properties of the solid boundaries and area Aw of the ceiling and the part of the wall: Q_ loss ¼ hw Aw ðT s  T a Þ

ð10:13Þ

Selected value for NFPA-92B: 0.65

0.6

0.5 0

2

4

6 8 10 12 14 16 Air changes per hour /ACH

18

20

22

Fig. 10.4 Heat loss fractions of smoke layer in self-developed zone model

Expressing the ratio of heat loss of the smoke layer to the convective portion of the heat release rate of the fire for the self-developed zone model as: χ¼

hw Aw ð T s  T a Þ Q_ c

ð10:14Þ

the values of χ for the zone model are plotted against the exhaust rate for different Q_ : of 2 MW, 3 MW and 5 MW in Fig. 10.4. It is observed that χ decreases as the exhaust rate increases. More heat on the smoke layer is removed for higher exhaust rate; hence, the fraction of heat loss from smoke to solid walls through convection, conduction and radiation decreases. The values of χ varied slightly under different heat release rate of the fire.

10.5

Conclusions

Heat transfer in a fire is very complicated and has to be simplified in a zone fire model. Two methods of simplifying the heat lost to solid boundaries in a two-layer zone model were discussed in this paper and compared to experimental results. The method reported in this paper gives better predictions on the smoke layer temperature in an atrium fire under mechanical exhaust. Smoke layer temperature depends on the heat loss of smoke layer to solid boundaries by convection, conduction and radiation. Since the smoke temperature is not high in an atrium fire due to the large volume of the space, heat loss of the smoke through solid boundaries can be calculated through a heat transfer coefficient hw and the area of walls and ceiling. The ratio of heat loss through solid boundaries to the fire heat

10

Discussion on Heat Lost Through Solid Boundaries in Modelling Atrium Fires Under. . .

release rate, χ, depends on the geometrical characteristics, ventilation conditions and others. Taking χ as a constant might not always work, as the value changes with the exhaust rate. Further investigations on the effects of other factors such as properties of the wall materials and localised flows including those near to the solid boundaries in affecting χ are required. Finally, increasing the exhaust rate might not necessarily move the smoke layer interface height up. As observed in the experiments, a clear stable smoke layer was not formed under some conditions such as exhaust rate higher than 12 ACH and fire with heat release rate smaller than 3 MW. From this study, providing a high exhaust rate for large atria should be considered carefully. More works are required to verify any proposed equations. At the moment, full-scale burning tests site such as the Australian hot smoke test [11–13] should be considered to demonstrate that the installed smoke exhaust system can keep the smoke layer sufficiently high. Acknowledgement The work is supported by the National Nature Foundation of China under Grant No. 51406241. All authors are members of the PolyU/USTC Joint Research Laboratory on ‘Fire Safety and Technology Research Centre for Large Space’.

References 1. Chow WK (2005) On carrying out atrium hot smoke tests. Archit Sci Rev 48:105–108 2. Yi L, Li YZ, Huo R, Chow WK (2005) Full-scale burning tests on atrium mechanical smoke exhaust system under quasi-steady vented stage. In: Kasapi N, Maruyama S, Yoshida H, Inoue T (eds) Proceedings of the 6th world conference on experimental

109

heat transfer, Fluid Mechanics, and Thermodynamics, Matsushima, Miyagi, Japan, 17–21 April 2005, pp 486–487 3. National Fire Protection Association (2000) NFPA 92B-2000, guide for smoke management systems in malls, atria, and large areas. National Fire Protection Association, Quincy 4. Chow WK, Yi L, Shi CL, Li YZ, Huo R (2006) Mass flow rates across layer interface in a two-layer zone model in an atrium with mechanical exhaust system. Build Environ 41:1198–1202 5. Yi L, Chow WK, Li YZ, Huo R (2005) A simple two-layer zone model on mechanical exhaust in an atrium. Build Environ 40:869–880 6. Klote JH (2002) Principles of smoke management. American Society of Heating, Refrigerating and Air-Conditioning Engineers, Atlanta 7. Chow WK, Li YZ, Cui E, Huo R (2001) Natural smoke filling in atrium with liquid pool fires up to 1.6 MW. Build Environ 36:121–127 8. Li YZ (2001) Smoke flow and control in large space atrium buildings. PhD Thesis, University of Science and Technology of China, Hefei, Anhui, China 9. Cooper LY, Harkleroad M, Quintiere J, Reinkinen W (1982) An experimental study of upper hot layer stratification in full-scale multi-room fire scenarios. J Heat Trans 104:741–749 10. Tanaka T, Yamana T (1985) Smoke control in large scale spaces – Part 1. Fire Sci Technol 5:31–40 11. Standards Australia (1999) AS 4391–1999, Smoke Management Systems – Hot Smoke Test, Australian StandardTM, Standards Australia (Standards Association of Australia), North South Wales, Australia, p 25 12. ASHRAE (1999) Proceedings of one-day seminar on hot smoke test. Hong Kong Institution of Engineering-Building Services Division, American Society of Heating, Refrigerating and Air-Conditioning Engineers, 12 March, 1999, Regal Kowloon Hotel, Hong Kong 13. Atkinson B (1999) HST verification of smoke management systems using the Australian method. In: Underground on the ground and into the air, EUROFIRE ‘99 – 4th European symposium, Institute of Fire Engineers (Belgium Branch), Affligen (Essene), Brussels, Belgium, 24–27 November 1999, pp 1–18

Experimental Study on Fire Behavior in a Compartment Under Mechanical Ventilated Conditions: The Effects of Air Inlet Position

11

Ken Matsuyama, Seiji Okinaga, Yasuo Hattori, and Hitoshi Suto

Abstract

The purpose of this research is to understand the fire behavior expected in a mechanically ventilated compartment. To date, some experimental studies have been conducted for investigation of fire behavior under mechanical forced ventilation; however, it might be not enough to understand everything. We therefore carried out a series of experiments on fire behavior focused on the effect of air inlet position in a compartment with same size as an ISO 9705 room (width 2.4 m  length 3.6 m  height 2.4 m) under conditions of mechanical ventilation using a pool fire. In this paper, the effects of ventilation conditions such as air inlet position and flow rate were studied. We found that differences in the air inlet position and flow rate were one of the principal factors for determining the burning behavior. Keywords

Compartment fire  Mechanical ventilation  Pool fire  Air inlet position

11.1

Introduction

In experimental studies of compartment fires such as highrise or large-scale buildings constructed using fire resistance principles, the ventilation conditions are generally natural ventilation by convection through the openings, and examples of experiments that consider compartments under mechanical ventilated conditions such as in a nuclear power plant are not comprehensive enough to develop a predictive method [1–10]. Therefore, the purpose of this study is to exactly understand the fire behavior expected in a mechanically ventilated compartment. In recent years, we have

conducted some experimental studies on fire behavior under mechanical ventilated conditions [11, 12]. As a part of the studies, a series of fire experiments under mechanical ventilated conditions taking into account various parameters were carried out. For instance, the position of inlet air, exhaust flow rate, heat release rate, and other parameters were examined with the aim of collecting fundamental data. In this paper, the experimental results on burning behavior when the position of the air inlet and the exhaust flow rate are taken as parameters are reported. In particular, flame observation results, the vertical temperature distribution, and the oxygen concentration in the compartment are the focus in the paper.

K. Matsuyama (*)  S. Okinaga Tokyo University of Science, 2641 Yamazaki, Noda-shi, Chiba 278-8510, Japan e-mail: [email protected] Y. Hattori  H. Suto Central Research Institute of Electric Power Industry, 1646 Abiko, Abiko-shi, Chiba 270-1194, Japan # Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_11

111

112

K. Matsuyama et al.

Fig. 11.1 Configuration of the experimental compartment [Unit: mm]. (a) Plan, (b) Section

11.2

Experimental Apparatus and Procedure

11.2.1 Experimental Compartment As shown in Fig. 11.1, a compartment having the same size as an ISO 9705 room (width 2.4 m  length 3.6 m  height 2.4 m) was used in the experiments. The mechanical ventilation system in the compartment consisted of an air supply duct with diameter of approximately 100 mm and an exhaust duct also with a diameter of approximately 100 mm connected to a mechanical exhaust fan at 2.2 m height. The height of the air inlet could change from a duct center position of 2200 mm (upper) to 352 mm (lower) from the floor according to the experimental conditions. A heat exchanger was installed between the exhaust duct and fan, and the air supply flow rate was kept constant from ignition to extinction. The walls of the compartment and floor were made of steel plate and 50 mm-thick autoclaved lightweight concrete (ALC) and 100 mm-thick ALC, respectively. Airtightness of the compartment was ensured by packing gaps in the joints and the like with sealant. As a result, airflow into and out of the compartment was only through the air supply inlet and exhaust outlet.

Table 11.1 Experimental condition Fuel pan diameter (area) 0.45 m (0.159 m2) 0.45 m (0.159 m2) 0.45 m (0.159 m2) 0.45 m (0.159 m2)

Air inlet height 0.352 m 0.352 m 2.20 m 2.20 m

Exhaust flow rate 0.006 m3/s 0.050 m3/s 0.006 m3/s 0.050 m3/s

11.2.3 Measurement Items The following items were measured in order to understand the burning behavior in the compartment with mechanically ventilated conditions: temperature inside the compartment by K-type thermocouples, gas concentrations (O2, CO2, CO) in the compartment by gas analyzers, the flow rate by bidirectional probes and the temperature by K-type thermocouples in both the air supply duct and exhaust duct, and pressure difference inside and outside the compartment by differential pressure gauge. The oxygen concentration was measured by using three analyzers of magnetic wind method type (type: POT-8000 (Shimadzu corp.)). The time delay correction was 40–90 s based on the length of the sampling probe. The location of each measurement point is shown in Fig. 11.1.

11.2.2 Fire Source

11.2.4 Experimental Conditions

Ethanol was used as the fire source. A circular fuel pan with a diameter of 450 mm was used and placed in the center of the compartment. Ethanol was supplied to the circular pan from outside the compartment by a siphon in order to reduce the influence of the pressure to the measurement apparatus.

Table 11.1 shows the experimental conditions. Two different air inlet positions were used at the heights of 0.352 m (upper) and 2.2 m (lower) from floor. Two exhaust flow rates of 0.006 m3/s and 0.05 m3/s were set by means of the exhaust fan. Since the number of gas analyzers was limited in the

11

Experimental Study on Fire Behavior in a Compartment Under Mechanical Ventilated. . .

113

Lower Upper

Fig. 11.2 Extinction time of each experimental condition

experiments, simultaneous measurements at all points could not be performed, and the experiments were therefore repeated eight times under each set of conditions while changing the gas sampling positions.

11.3

Results and Discussion

11.3.1 Burning Behavior The extinction time of each experiment is shown in Fig. 11.2. The extinction time in case of the lower air inlet was usually shorter than in case of the upper air inlet. Figures 11.3 and 11.4 show the flames for the lower and upper air inlet positions, respectively, at the exhaust flow rate of 0.05 m3/s. Figure 11.7 shows the time profile of pressure inside the compartment in each experiment. As shown in Fig. 11.3, in case of the lower air inlet position, although the flames went up vertically immediately after ignition, they began to tilt toward the supply air from the air inlet 3 min after ignition. As more time passed, the burning area shifted toward the downwind side and the tilt angle of the flames was observed to increase. The extinction of the flame was observed approximately 6 min after ignition. This is thought to be because the tilt angle of the flames was directly in the airflow from the air supply inlet into the compartment. The reason why the flames did not tilt immediately after ignition is thought to be that the interior of the compartment entered a state of positive pressure due to rapid combustion that consumed the oxygen within the compartment and this obstructed the inflow of air from the air supply inlet as shown in Fig. 11.7a. Once the inflow of air from the air supply inlet resumed, the combustion became unstable because the flames were struck directly by the airflow. This is also clear from the magnitude of pressure variation inside the compartment as shown in Fig. 11.7a. A phenomenon was observed where the flame was luminous immediately after ignition but changed to bright blue near the fuel pan. The change in flame color to

Fig. 11.3 Time variation of flames; lower air inlet and exhaust flow rate of 0.05 m3/s (extinction at 6 min 25 s). (a) 1min after ignition, (b) 3min after ignition, (c) 6min after ignition, (d) 6min20s after ignition

blue is attributed to chemiluminescence from the reaction of oxygen with carbon monoxide during incomplete combustion of the ethanol due to the reduced oxygen concentration in the compartment. As shown in Fig. 11.4, in case of the upper air inlet positions, the flames remained in the center of the fuel pan from ignition until the end of the experiment. A phenomenon was observed where the size of the flames gradually decreased as time elapsed, and at some stage, the size of the flames stopped changing and the combustion stabilized. Furthermore, unlike in the case with the lower air supply inlet, the flames were not observed to tilt due to the airflow inside the compartment. Because the combustion continued until the end of the experiment, it was found that even when the air supply inlet was at the top of the compartment, some of the inflowing air from the air supply inlet supplied oxygen by convection to the bottom of the compartment where the fuel pan was located. The convection that occurred due to the combustion is thought to be gentle as can be seen from the small pressure variation inside the compartment, shown in Fig. 11.7b. The burning behavior after stabilization was found to depend on the oxygen flow rate supplied to the fire source. When the exhaust flow rate was the low value of 0.006 m3/s, the effect appeared significantly as shown in Fig. 11.5. From 3 min

114

K. Matsuyama et al.

Fig. 11.5 Time variation of flames; upper air inlet and exhaust flow rate of 0.006 m3/s (extinction at 7 min 40 s). (a) 1min after ignition, (b) 3min after ignition, (c) 6min after ignition, (d) 7min30s after ignition

11.3.2 Temperature Distribution Inside the Compartment

Fig. 11.4 Time variation of flames; upper air inlet and exhaust flow rate of 0.05 m3/s (combustion continuous). (a) 1min after ignition, (b) 3min after ignition, (c) 6min after ignition, (d) 9min after ignition, (e) 15min after ignition, (f ) 20min after ignition

after ignition when the oxygen concentration had decreased, the combustion was observed to shift from full combustion of the entire fuel pan to partial combustion at the center of the fuel pan. The flames were observed to gradually change from luminous to bright blue and then to dark blue. The dark blue emission is thought to be due to radicals (CH groups) [13]. Figure 11.6 shows the flames immediately after ignition and 4 min after ignition when the parameters were upper air supply and the exhaust flow rate of 0.006 m3/s. A phenomenon was observed where the combustion area was smaller than that shown in Fig. 11.4 and depended on the amount of oxygen supplied to the fire source.

Figure 11.8 shows the temperature measurement points inside the compartment. And also, Figs. 11.9 and 11.10 show the measurement results of the vertical temperature distributions inside the compartment according to the thermocouple trees A and E. Here, we focus on the vertical temperature distributions in regions A and E even though many temperatures were measured inside the compartment, because the main purpose of the paper is to analyze heat transfer due to the combustion and the airflow due to the air supply.

11.3.2.1 Measurement Results in Case of Lower Air Inlet Exhaust Flow Rate of 0.006 m3/s As shown in Fig. 11.9a, the temperatures at a given height in regions A and E were almost the same, indicating that a horizontal temperature gradient was not present inside the compartment, since the velocities of the airflow through the inlet and exhaust and of the plume generated by combustion was slow. The calculated smoke layer height [14] was close

11

Experimental Study on Fire Behavior in a Compartment Under Mechanical Ventilated. . .

115

a Pressure Pressure(5s avg.) The fuel mass loss rate

Exncon

b Pressure Pressure(5s avg.) The fuel mass loss rate

Fig. 11.7 Pressure difference inside and outside the compartment and fuel mass loss rate; exhaust flow rate of 0.05 m3/s. (a) lower air inlet, (b) upper air inlet

Fig. 11.6 Time variation of flames; upper air inlet and exhaust flow rate of 0.006 m3/s (combustion continuous). (a) 1min after ignition, (b) 3min after ignition, (c) 6min after ignition, (d) 9min after ignition, (e) 15min after ignition, (f ) 20min after ignition

Exhaust Fan Thermocouple tree E

Ceiling 2.2m 1.9

to floor level after about 2 min., thus the temperatures inside compartment were uniform.

1.6m

Thermocouple tree A

1.3m 1.0m

Exhaust Flow Rate of 0.05 m3/s As shown in Fig. 11.9b, unlike in the case of the exhaust flow rate of 0.006 m3/s, temperature differences were observed between measurement points at the same heights in regions A and E. There was a temperature gradient inside the compartment at the higher exhaust flow rate. As shown in Fig. 11.10, by the heat flux through the inlet side from the exhaust side, the temperatures in the lower part of tree A became higher than tree E.

Fuel Pan Air Supply

0.7m 0.4m

(Upper) Air Supply

(Lower)

0.1m Floor Measurement Point

Fig. 11.8 Schematic diagram of temperature measurement points inside the compartment

116

b

250

Tree A Tree E

200

Extinction （ 7:40）

2.4

Tree A Tree E

Extinction 2.1 （ 6:25） 1.8

2.0m

2.0m 1.4m

150

1.5

0.8m 0.5m

100

0.2m Smoke layer height

50

1.2

1.4m 0.8m 0.5m

0.9

0.2m

0.6

Smoke layer height

0.3

Height (m)

a

Temperature (⬚C)

Fig. 11.9 Temperature distribution inside the compartment for the lower air supply; (a) exhaust flow rate of 0.006 m3/s, (b) 0.05 m3/s

K. Matsuyama et al.

0.0

0 0

2

4

6

Time (min)

Fig. 11.10 Spatial temperature distribution inside the compartment for the lower air supply; exhaust flow rate of 0.05 m3/s (6 min after ignition)

11.3.2.2 Measurement Results in Case of Upper Air Inlet Exhaust Flow Rate of 0.006 m3/s As shown in Fig. 11.11a, a sudden drop in temperature was observed 7 min after ignition. This is because the oxygen present in the compartment prior to ignition was consumed. The combustion thus transitioned to being dependent on oxygen provided from the air supply inlet, resulting in a decrease in the heat release rate. As in the case of lower air supply with the exhaust flow rate of 0.006 m3/s, measurement points at a given height in regions A and E had almost the same temperature. Exhaust Flow Rate of 0.05 m3/s As shown in Fig. 11.11b, because of the increased exhaust flow rate, no temperature drop like the one at the exhaust flow rate of 0.006 m3/s was observed after the transition to combustion that was dependent on oxygen provided from the air supply inlet. Furthermore, measurement points at a given height in regions A and E, except for the measurement points at 1.9 m, had almost the same temperature, as in the case of the exhaust flow rate of 0.006 m3/s. This indicates that the temperature gradient was even inside the

8

10

0

2

4

6

8

10

Time (min)

compartment without disturb in spite of the higher exhaust flow rate. The reason for the temperature difference observed at the 2.0 m above the floor was that region A was near the air supply inlet, and so the temperature was lower compared to region E since it is depended on the outside temperature. In the case of the upper air supply, the interior of the compartment was found to be mixed to a lesser extent compared with the case of the lower air supply, even at the higher exhaust flow rate, and a vertical temperature distribution with the similar gradient inside the compartment was observed.

11.3.3 Oxygen Concentration Distribution Inside the Compartment As shown in Fig. 11.12, the oxygen concentration inside the compartment was measured at 24 points. Figures 11.13 and 11.14 show the time profile of oxygen concentration inside the compartment. Note that the sudden increase in the observed oxygen concentration is due to the consumption of oxygen stopping due to extinction of the fire and oxygen being supplied from the ventilation.

11.3.3.1 Measurement Results in Case of Lower Air Inlet Exhaust Flow Rate of 0.006 m3/s As shown in Fig. 11.13a, immediately before extinction, the oxygen concentration on the air supply side and exhaust side at the height where the fuel pan was installed was approximately 15 %, but at all other points it was almost the same value of approximately 12 %. Because of this, it is thought

Experimental Study on Fire Behavior in a Compartment Under Mechanical Ventilated. . .

a

b

250

200

2.4

Tree A

Tree A Tree E

Temperature (⬚C)

Fig. 11.11 Temperature distribution inside the compartment for the upper air supply; (a) exhaust flow rate of 0.006 m3/s, (b) 0.05 m3/s

117

Tree E

2.1

2.0m

2.0m 1.4m

1.8

1.4m

1.5

0.8m

150

0.8m

1.2

0.5m

100

0.9

0.5m

0.2m

0.6

0.2m

50

Smoke layer height

Smoke layer height

0.3 0.0

0 0

5

10 15 Time (min)

20

0

5

10 15 Time (min)

20

Fig. 11.12 Schematic diagram of oxygen concentration measurement points

b

a

（ ）

（ ）

Fig. 11.13 Oxygen concentration distribution inside the compartment with the lower air supply; (a) exhaust flow rate of 0.006 m3/s, (b) 0.05 m3/s

（

）

（

）

Height (m)

11

118

b

（ ）

a

（ ）

Fig. 11.14 Oxygen concentration distribution inside the compartment with the upper air supply; (a) exhaust flow rate of 0.006 m3/s, (b) 0.05 m3/s

K. Matsuyama et al.

（

that a boundary layer divided the oxygen concentration into two zones between the lowermost and second lowermost measurement points. This trend is the same as the temperature measurement results. Exhaust Flow Rate of 0.05 m3/s As shown in Fig. 11.13b, almost no differences in oxygen concentration were observed among all the measurement points. The interior of the compartment is therefore thought to have been uniform. The flame extinction at an oxygen concentration of approximately 15 % was observed, the same as for the exhaust flow rate of 0.006 m3/s. 3

Comparison of the Exhaust Flow Rates of 0.006 m /s and 0.05 m3/s As shown in Fig. 11.13a, b, the gradient of the reduction in oxygen concentration was found to differ between the exhaust flow rates of 0.006 m3/s and 0.05 m3/s. The oxygen concentration decreased almost linearly at the exhaust flow rate of 0.006 m3/s, but for the exhaust flow rate of 0.05 m3/s, the oxygen concentration exhibited the same reduction gradient as in the case of the exhaust flow rate of 0.006 m3/s during the first 2 min after ignition, with the gradient then becoming slightly less steep. This is attributed to the burning area being reduced by the airflow moving around the fuel pan, which reduced the oxygen consumption rate.

11.3.3.2 Measurement Results in Case of Upper Air Inlet Exhaust Flow Rate of 0.006 m3/s As shown in Fig. 11.14a, only the oxygen concentration at the height where the fuel pan was installed on the air supply side differed from the other measurement points and exhibited a slightly higher oxygen concentration. Although

）

（

）

the interior of the compartment was therefore thought to have been close to uniform, some of the air flowing in from the air supply inlet collided with the plume generated by combustion, and this supplied oxygen to the height where the fuel pan was installed from the air supply wall side. Exhaust Flow Rate of 0.05 m3/s As shown in Fig. 11.14b, no significant differences in oxygen concentration were observed among all the measurement points. The air flowing in from the air supply inlet was therefore uniformly dispersed within the compartment. Comparison of the Exhaust Flow Rates of 0.006 m3/s and 0.05 m3/s As shown in Fig. 11.14a, b, the gradient of the reduction in oxygen concentration exhibited the same trend as in the case of lower air supply. However, the gradient of the reduction in oxygen concentration at the exhaust flow rate of 0.05 m3/ s was gentler for the upper air supply compared with the lower air supply. This is thought to be because the amount of oxygen supplied to the fire source was smaller and the combustion area was gradually reduced.

11.3.3.3 Comparison of Different Air Supply Inlet Positions For both the lower air supply and upper air supply, the oxygen concentration until the flame extinction was approximately 15 % at the height of 0.1 m above floor. When the oxygen concentration at the height of 0.1 m above floor was lower than approximately 15 %, the flame extinction would be expected to occur. In all the experiments with the lower air supply, the flame extinction occurred during the experiment. In contrast, for the upper air supply, experimental results were also obtained where the combustion continued until the end of the experiment without the flame extinction.

11

Experimental Study on Fire Behavior in a Compartment Under Mechanical Ventilated. . .

11.4

Summary

The following results were obtained from these experiments.

11.4.1 Lower Air Supply • A phenomenon was observed where airflow from the air supply inlet directly struck the flame, which made the combustion unstable and flame extinct. • At the exhaust flow rate of 0.006 m3/s, a layer of air supplied from the air supply inlet was maintained at the bottom of the compartment and the compartment was stratified into two layers. • At the exhaust flow rate of 0.05 m3/s, the temperature gradient was not even inside the compartment with disturb; however, a stable airflow distribution is thought to have formed.

11.4.2 Upper Air Supply • In some of the experiments, the burning area decreased depending on the amount of oxygen supplied to the fire source, and the flame continued and was not extinct until the end of the experiment. • At the exhaust flow rate of both 0.006 m3/s and 0.05 m3/s, essential temperature gradients were mostly the same inside the compartment, and airflow disturbances due to the air supply were small. Acknowledgment The authors would like to thank Mrs. Y. Terada and T. Mizuno for helps in the experiments.

References 1. Tanaka T, Kabasawa Y, Soutome Y, Fujizuka M (1986) Preliminary test for full scale compartment fire test (lubricant oil fire test:

119

part 1). In: Fire safety science – proceedings of the 1st international symposium, International Association for Fire Safety Science, pp 799–808. doi:10.3801/IAFSS.FSS.1-799 2. Foote KL, Pagni PJ, Alvares NJ (1986) Temperature correlations for forced-ventilated compartment fires. In: Fire safety science – proceedings of the 1st international symposium, International Association for Fire Safety Science, pp 139–148. doi:10.3801/IAFSS. FSS.1-139 3. Peatross MJ, Beyler CL (1997) Ventilation effects on compartment fire characterization. In: Fire safety science – proceedings of the 5th international symposium, International Association for Fire Safety Science, pp 403–414. doi:10.3801/IAFSS.FSS.5-403 4. Chow WK, Tsui SC (1998) Temperature distribution induced by fires in a small chamber with forced ventilation. J Fire Sci 16 (2):125–145. doi:10.1177/073490419801600204 5. Mizuno K, Tanaka S, Hasemi Y (1992) A model experimental study of force-ventilated fires in a highly air tight enclosure: general characteristics of combustion and conditions of a heavy carbon monoxide generation. J Archit Plan Environ Eng 435:137–145, (in Japanese) 6. Sano T, Shirai K, Hattori Y, Suto H, Tsuchino S (2013) Comprehension of limited ventilated fire behavior and study on fire prediction method in an enclosed space. Trans Atmoic Energy Soc Jpn 12:32–42, (in Japanese) 7. Pre´trel H, Le Saux W, Audouin L (2012) Pressure variations induced by a pool fire in a well-confined and force-ventilated compartment. Fire Saf J 52:11–24 8. Vaux S, Pre´trel H (2013) Relative effects of inertia and buoyancy on smoke propagation in confined and forced ventilated enclosure fire scenarios. Fire Saf J 62(Part B):206 9. Pre´trel H, Le Saux W, Audouin L (n.d.) Determination of the heat release rate of large scale hydrocarbon pool fires in ventilated compartments. Fire Saf J 62(Part B): 192 10. Audouin L, Rigollet L, Pre´trel H, Le Saux W, Ro¨wekamp M (2013) OECD PRISME project: fires in confined and ventilated nucleartype multi-compartments – overview and main experimental results. Fire Saf J 62(Part B):80 11. Hattori Y, Matsuyama K, Suto H, Onuma E, Okinaga S (2014) Turbulence measurements in a ventilation-controlled pool fire. In: Proceedings of 16th international symposium on flow visualization 12. Hattori Y, Matsuyama K, Suto H, Onuma E, Okinaga S (2014) Entrainment process in the vicinity of pool fire under ventilation condition. In: Proceedings of the 15th international heat transfer conference, 09377 13. Niioka T, Kono M, Sato J (2001) Fundamentals of combustion phenomena, Ohmsha, pp 180–186, (in Japanese) 14. He Y, Fernando A, Luo M (1998) Determination of interface height from measured parameter profile in enclosure fire experiment. Fire Saf J 31:19–38

Part IV Egress Safety

Study of the Occupant Characteristics During Evacuation in Medium- and High-Rise Buildings in Indonesia

12

Fietrysia Leonita, Harfan Sakti, and Yulianto Sulistyo Nugroho

Abstract

One way to avoid the loss of life in building fires is by performing effective evacuation. As a consequence of economic developments, cities of the world continuously build high-rise buildings, including super high-rise ones having more than 40 storeys. When a building has to perform full evacuation, the long travel distance in a super high-rise building can cause delay to reach exit discharge locations on the ground floor. Although a number of studies have been carried out to identify the characteristics of evacuee during emergency, it is suggested that the evacuation behaviour may be affected by cultural background of the ¨ zkaya A 2001) [1]. This paper is intended to gather building evacuation time, persons (O vertical movement speed data and to study the behaviour of occupants during evacuation in medium- and high-rise buildings. The study was carried out mostly in Jakarta, Indonesia. The evacuation time and overall movement speed were observed by gathering the data during a typical evacuation drill in the medium- and high-rise buildings. The result suggested that vertical movement speed descending from the fourth floor is around 0.70–0.81 m/s. As the evacuee is descending from the higher floors, the lower movement speed was identified, e.g. around 0.62–0.75 m/s for those descending from the ninth floor and around 0.50–0.73 m/s for those descending from 19th floor. The trends are in good agreement with other works suggested in the literature. From the surveys, regardless of the occupancy, i.e. whether the building is an office, a college or shopping avenue, interesting behaviours of the evacuee during evacuation process were identified. While travelling downwards, evacuees tend to evacuate in groups, to walk side by side with other members of the groups, and some of them hold the stairs’ railing. These caused queuing and congestion, especially if the exit stairs’ width only fit two people walking side by side. Keywords

Evacuation time
High-rise building
Movement speed
Behaviour

12.1

F. Leonita
H. Sakti
Y.S. Nugroho (*) Department of Mechanical Engineering, Universitas Indonesia, Kampus UI, Depok 16424, Indonesia e-mail: [email protected]

Introduction

Life safety considerations for super high-rise building evacuation have led to many developments including the introduction of new strategies such as refuge floors and lifeboat concept using protected elevators. Although full evacuations rarely occur, when a building has to perform full evacuation, the condition of super high-rise buildings can affect the evacuation time. Longer travel distances on super high-rise

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_12

123

124

F. Leonita et al.

buildings can decrease the movement speed, which can be caused by fatigue, density on stair or other causes. Other studies suggest that the movement speed decreases due to the long travel distance. A study by Wahyu Sujatmiko et al. [2] showed that an evacuation of 21 persons travelling downstairs from the 20th floor to the ground floor would need an egress time of 187 s or 3 min and 7 s, with an average downstairs speed of 0.79 m/s. An experiment with a further travelling distance was conducted by J. Ma et al. [3]. The experiment showed that an evacuation of six persons travelling downstairs from the 101st floor to the ground floor would need an egress time of 2000 s or 33 min and 20 s, with a mean downstairs speed of 0.28 m/s. ¨ zkaya [1] suggested the impact of cultural backAydin O ground on evacuation behaviour. In this respect, it is important to conduct an evacuation study in various building conditions in Indonesia. This study performs evacuation time measurements during evacuation drill in a high-rise building in Jakarta and Depok. The aim of this study is to gather building evacuation time and vertical movement speed data and behavior of evacuees. This study is limited to observing an evacuation drill with full evacuation scenario. The observation took place on one stair. Observations were conducted on some evacuees that travel from the upper observation floor. The movement time was measured as the difference between the time entering the exit door and the time leaving the exit stairs. Building evacuation should be performed and completed before untenable conditions occur. Therefore, the Required Safe Egress Time (RSET) should be smaller than the Available Safe Egress Time (ASET) [4]. In the SFPE handbook [4], the RSET is formulated as RSET ¼ td þ ta þ to þ ti þ te

ð12:1Þ

where: td ¼ time from fire ignition to detection ta ¼ time from detection to notification of occupants of a fire emergency to ¼ time from notification until occupants decide to take action ti ¼ time from decision to take action until evacuation commences te ¼ time from the start of evacuation until it is completed The main components that affect evacuation movement time are evacuation movement speed, travel distance and queuing time. In a more detailed calculation, evacuation time can also be affected by:

(i)

Building condition when fires occur, such as the presence of smoke [5] and the visibility condition [6] (ii) Evacuation strategy such as evacuation strategy for occupants with movement limitations in high-rise building evacuation [7] (iii) Building elements such as quantity, location and dimension of exit stairs, door [8, 9], corridor, building layout, model and location exit sign [5, 10], presence of obstacle on evacuation route [11, 12] and alarm [13] (iv) Evacuee characteristics

Based on the Life Safety Code [14], evacuee characteristic is defined as the ability and behaviour of people before and during fire. The movement ability of evacuees can be affected by their physical condition, e.g. health condition, being pregnant, age [13], gender [15], body mass index (BMI) [15], impairments [16] and obesity [17]. The evacuee behaviours that can affect evacuation movement are, for example, competition of evacuee to reach stair buffer area [18], exit selecting behaviour [19] and finding members of group before evacuating behaviour [20].

12.2

Data Collection for Buildings

This study was conducted by direct observation of evacuees using cameras during an evacuation drill performed in medium- and high-rise buildings. Even though an evacuation drill condition is not exactly the same as a real evacuation condition, the characteristics of Indonesian evacuees can still be captured by means of observation.

12.2.1 Data Collection Procedures Evacuation drill data were collected by positioning video cameras, out of the way of the evacuees, to record the movement of occupants in an exit stair during the evacuation. From the video recording, the data was extracted manually. The data collected from the occupants were age and gender, whereas the data collected from the building itself were stairs and exit dimension and quantity. The calculation of evacuation time started from the alarm notification and ended after the last evacuee exited from the exit door or reached the end of the exit stairs. To collect the overall movement time on stairs, video cameras were used on one of the exit stairs. The cameras were placed in two observation floors, upper floor and ground floor. From the evacuation process that was recorded on the upper observation floor camera, some evacuees were specifically selected to be observed. The time needed to descend the stairs was calculated from the moment an evacuee entered the exit

12

Study of the Occupant Characteristics During Evacuation in Medium- and High-Rise. . .

125

Fig. 12.1 Evacuees enter and exit from the stairwell

stairs was captured on camera in the upper observation floor and finished after the evacuee exited from exit discharge in the ground floor, as seen in Fig. 12.1. Travel distance was measured in the middle of the stairs and along the stairs diagonally. The outflow rate data was obtained by placing a camera in front of the exit door on the ground floor. Each person that exited from the exit door was counted per minute to get the outflow rate. Collected data were then analysed according to evacuation condition identified from video recording and live observation. The study result was also analysed to study the effect of vertical travel distance on evacuation time and occupant downstairs speed.

12.2.2 Survey 1: Observation on Fourth Floor, Medium-Rise Building Survey 1 was conducted in April 2015, at a medium-rise building, located in South Jakarta. Survey 1 and survey 4 took place in a conjoined building complex, with survey 1 and survey 4 taking place at a 6-storey shopping avenue at the podium/base of the building complex and a 40-storey office building on top of the podium, respectively, see Fig. 12.2. Nonetheless, both surveys were conducted at different times. The evacuation drill was held by the building management in collaboration with the Fire and Rescue Department of Jakarta. The evacuation drill started in the morning, before operating hours of the shopping avenue, with only shop owners and staff participating. In general, there was no queuing seen during the evacuation process. The eight evacuees that were being observed were part of the group of evacuees that arrived early on the ground floor. The observed evacuees only met four other evacuees, during their evacuation process; therefore, it can be said that the evacuees were moving freely. In this survey, the video recording started a couple of seconds after the alarm started; therefore, data related to the alarm’s timing were not available.

12.2.3 Survey 2: Observation on the Fifth Floor, Medium-Rise Building The second survey, conducted in October 2014, was in a medium-rise building at the Universitas Indonesia in Depok. The building is a six-storey building with two exit stairs. The existing exit stairs are an open stair type. The exit stairs are located on the west and east sides of the building, connected by a corridor, Fig. 12.3. There was no public announcement given; therefore, the evacuation started after the sound of the alarm went off. The evacuation drill was informed to the lecturers and the students prior to the event. The evacuation drill started before lunch hour. To prevent the overloading of occupants in a single stairway, the building management applied a stair distribution strategy. The odd floors used the west stair, while the even floors used the east stair. The observations of the evacuation are placed in one stair, the west stair, on fifth and ground floors. The queuing in front of the west stair happened around 130 s after the alarm rang. After around 5 min, the evacuation drill was finished. To collect evacuation data, 50 evacuees were observed from the fifth floor evacuees. From 50 evacuees, 35 evacuees were male and 15 evacuees were female.

12.2.4 Survey 3: Observation on the 9th and 19th Floor, High-Rise Building Survey 3 took place at a 20-storey office building located in West Jakarta, with the help of the Fire and Rescue Department of Jakarta, in February 2015, just before lunch break. The building has two exit stairs, located at its core, see Fig. 12.4. The observation took place at the south exit stairs on the ground floor, 9th floor and 19th floor. In general, the evacuation condition was not crowded, but some queues were spotted at the ground floor. The observation, carried from the ninth floor, was focused towards 17 observed evacuees, consisting of 9 males and 8 females, while another observation, from the 19th floor, was focused towards 23 observed evacuees, 14 males and 9 females.

126

F. Leonita et al.

Fig. 12.2 Floor plan for survey 1 and survey 4 (Source: Jakarta Fire and Rescue Department Data)

Fig. 12.3 Floor plan survey 2

12.2.5 Survey 4: Observation on the 30th Floor, High-Rise Building This observation for high-rise building evacuation took place in a 40-storey office building, located in South Jakarta in December 2014, which is a part of a conjoined building complex, with the building from survey 1. Just like the other surveys, the drill was conducted by the building management in collaboration with the Fire and Rescue Department of Jakarta. The building has two exit stairs, located in the

west and east sides. The observation was conducted only on the west stair. The west stair consists of two types of stairs: the first type has three intermediate landings from the first floor to the sixth floor, and the second type is a u-shaped stair from the 6th floor to the 40th floor. The evacuation drill took place on a business day. Prior to the evacuation drill, building occupants were informed about the event but not the exact time. The evacuation participants were building occupants. The observations for this evacuation were conducted in the 30th floor and ground floor. At the start of

12

Study of the Occupant Characteristics During Evacuation in Medium- and High-Rise. . .

Fig. 12.4 Floor plan survey 3

the evacuation, the density on the 30th floor exit stairs was still low, and evacuees could move freely. Around 6 min after the alarm, the 30th floor exit stairs became crowded, and then around the ninth minute, the evacuation started to become more crowded and became congested in the 30th floor.

12.3

Result and Analysis

12.3.1 Evacuation Time and Overall Movement Speed The results from all four surveys (1 and 2 took place on a medium-rise building, and surveys 3 and 4 took place on high-rise building) are as follows: From Table 12.1, it can be seen that the time needed for all evacuees to reach the ground floor for the high-rise building was greater than that of the medium-rise building. The time needed for all evacuees to reach the ground floor was counted from the first ring of the alarm to the time all evacuees exited from the stairwell on the ground floor. In survey 4, with a 40-storey high building, approximately 1002 evacuees needed ~34 min to reach the ground floor. Whereas from survey 3, with a 20-storey building, 230 evacuees needed approximately 21 min to reach the ground floor. And from the second survey, with a six-storey building, 278 evacuees needed about ~5.5 min to travel downwards and reach the ground floor. Buildings with greater height have a further travel distance for evacuees to reach the ground floor, which contributed to the greater time needed to reach the ground floor. Nevertheless, the time needed to reach the ground floor is also influenced by other factors, such as the premovement time, the condition during evacuation, the characteristics of evacuees, the number of evacuees, etc.

127

The evacuation conditions that were spotted on the upper observation floor were not the same. In the second and fourth surveys, the queues were identified on the upper observation floor, but in the first and third surveys, no queues occurred. The width of exit stairs also varied. The width of stairs on the first and second surveys was wider than the third and fourth surveys. The maximum outflow of the evacuee from all the surveys was also varied. On the second and fourth surveys, where queues occurred on the observation floor, the maximum outflows were ~87 people/min and ~65 people/min, respectively. Meanwhile, from the first and third surveys, where – in general – there were no queues on the upper observation floor, the maximum outflows were ~32 people/ min and ~41 people/min, respectively. The number of outflows were not only influenced by evacuees’ density but also affected by the width of the stairs, as mentioned on the SFPE handbook [21]. The evacuation outflow is influenced by speed, evacuee’s density and width of the evacuation pathway [21]. The width of the exit stairs from survey 2 is greater than that of survey 4, which might contribute to the fact that maximum outflow number from survey 2 is greater than the maximum outflow from survey 4. The stair width of the building in survey 2 can accommodate up to three persons walking side by side. But, the width of stairs in the building of survey 4 was narrower which can only facilitate two evacuees walking side by side. The maximum outflows are also different between surveys 1 and 3. Maximum outflows on survey 3 was higher than survey 1, which probably caused by different density conditions. The overcrowdedness might have occurred due to some evacuees arriving at the same place around the same time, or due to obstacles on the evacuation path, that led to the forming of queues. Obstacles could be in the form of individual or groups of evacuees that blocked the exit and/or moving at a much slower speed. The behaviour of some evacuees that travelled in groups could block other evacuees’ path, as seen in the results of the surveys. Furthermore, evacuees that travelled in groups tend to move slower because of their tendency to follow the speed of the slowest member of the group, as mentioned in the SFPE handbook [21] and research conducted by Proulx [13]. As observed in the surveys, the smallest group observed was the group of two. In the first survey, which had a larger width of stairs, the effect from obstacles in the form of two people walking side by side did not make a great impact and did not create long queues. But, in the third survey, where the stairs were only enough for two people walking side by side, the group of evacuees that walking side by side could cause queues. The following figures are the results of outflow and downward speed from surveys 2, 3 and 4: From survey 2, a medium-rise building, the evacuation time was less than the evacuation time from surveys 3 and

128

F. Leonita et al.

Table 12.1 Result from surveys 1, 2, 3 and 4 Upper observation floor Occupancy Floors Evacuee age (years old) Exit stair Stair width before intermediate landing (m) Stair width after intermediate landing (m) Stair riser (mm) Stair tread (mm) Steps before and after intermediate landing Exit width (m) Number of observed evacuee (person) Observed male evacuee (person) Observed female evacuee (person) Observed evacuee age (years old) Number of observation stairs Evacuees that reach exits stairs (person) Time for all evacuee discharge from exit stairwell (s) (from the first alarm to the time evacuees exit from the stairwell on the ground floor) Highest outflow (person/minute)

Survey 1 4 Shopping 6 20–40

Survey 2 5 Lecture 6 17–55

Survey 3 9 and 19 Office 20 20–50

Survey 4 30 Office 40 20–50

1.4 1.4 170 300 14/14 1.8 8 1 7 20–40 1 ~198

1.5 1.65 180 300 9/9 1.65 50 35 15 17–55 1 ~278 ~335 (~5 min 35 s) ~87

1.2 1.2 180 270 10/10 0.9 40 23 17 20–50 1 ~230 ~1273 (~21 min 13 s) ~41

1.15 1.15 180 285 11/11 1.8 50 24 26 20–45 1 ~1002 ~2027 (~33 min 47 s) ~65

~32

Fig. 12.5 Outflow survey 2

4 – high-rise building. Figure 12.5 shows the outflow pattern of survey 2. From 0 to the first minute, the outflow increases slightly and then steady high from the ~2 to ~4 min. Minutes later, the outflow decreases until all evacuees exit from the exit door. From the third and fourth surveys, the evacuation time is greater than that of medium-rise building. The pattern of the outflow, as seen in Figs. 12.7 and 12.9, is wavy, with multiple high points of outflow. At one time the outflow is high, and then it drops to a lower level and then comes back up again. This wavy pattern of outflow might occur because at certain times some evacuees arrive at the same place around the same time, or due to obstacles that led to the forming of queues. Even though the evacuees’ speed was affected by each individual’s ability and other factors, the movement speed pattern from surveys 2, 3 and 4 seems to be related to the pattern of evacuees’ outflows. From the second survey, the outflow graph was low at the start and then comes up for a couple of minutes and then tilts down again at the end. Related to outflow graph, movement speed

Fig. 12.6 Movement speed on stair, descending from the fifth floor, survey 2

graph’s pattern, Fig. 12.6, is high at the start (the first evacuee could travel faster due to the low density), and then the graph comes down then tends to flat on evacuee numbers 2–50 that seem to be affected by density condition that’s quite steady high on minutes 2–4. From surveys 3 and 4, Figs. 12.7 and 12.9, where the patterns for outflow graph are wavy, the movement speed graph’s patterns, Figs. 12.8 and 12.10, are also similarly wavy. Table 12.2 provides information on the downward speed of all surveys. Table 12.2 suggests that building occupants require 2–6 min to reach the exit stair entrance/door, counted from the time the alarm went off. The time required to travel downstairs was greater in high-rise building than mediumrise building, which might be due to the additional travelling distance. Nevertheless, the time to reach the ground floor

12

Study of the Occupant Characteristics During Evacuation in Medium- and High-Rise. . .

Fig. 12.7 Outflow survey 3

Fig. 12.8 Movement speed on stairs, descending from the 19th floor, survey 3

Fig. 12.9 Outflow survey 4

129

know the impact of travelling distance on movement speed and the fact that the evacuee’s density was different in each survey, then it is appropriate to say that the result of this study only makes up a part of a comprehensive effort in the future to establish an inclusive data about evacuation characteristics of general building occupation in Indonesia. It is important to mention that in some cases being observed, there were queues that caused a decrease in movement speed. Therefore, the data of movement speed gathered might be affected by those queues, as in survey 2 and survey 4. The movement speed data that might reflect the evacuees’ ability to travel downstairs was observed in survey 1 and 3, where the conditions were relatively not crowded and the queues only happened at the ground floor. The downward speed from the first survey, in a medium-rise building, was 0.7–0.81 m/s. In the third survey, high-rise building, the evacuee’s downward speed was 0.5–0.75 m/s. The difference in the average downward speed on the four surveys, apart from each individual’s ability, was likely influenced by the level of crowdedness (evacuees’ density) and the distance to reach the exits. The average downward speeds from the second and fourth surveys were lower than the first and third surveys. The queuing condition that was spotted on the upper observation floor in the second and fourth surveys, and the congestion condition that was spotted in the fourth survey, might be the reason for the lower average downward speed result. This is in line with the SFPE handbook [22] that crowdedness can lead to the decreasing of movement speed, as shown in Table 12.3. Results of the average downward speed from all the surveys were approximately in line with the movement speed parameters for groups of people on a variety of means and conditions in the SFPE handbook (Table 12.3) [22], while the difference in distance could cause fatigue which can slow the evacuation time [23].

12.3.2 Behaviour of the Evacuees

Fig. 12.10 Movement speed on stairs, descending from the 30th floor, survey 4

was also influenced by other factors, such as extra time due to premovement, the condition during evacuation, the characteristics of evacuees, the number of evacuees, etc. Considering the fact that lots of surveys are needed to fully analyse the evacuee’s ability to travel downstairs and to

The characteristics of evacuees, aside from their ability, also involve their behaviour. As mentioned in the Life Safety Code [14], the characteristics of the occupants are defined as the ability or behaviour of the occupants before and during fire. From the surveys, both in medium- or high-rise buildings – regardless of whether the occupant was in an office, in a college or in shopping avenues – the behaviours noticed were the tendency of some evacuees to evacuate in groups, which appeared in Fig. 12.11, and the tendency of some evacuees to hold the stairs’ railing while travelling downwards. As observed in the surveys, the groups consisted of two to four members. Some evacuees that held the railing while travelling downwards seemed to tread carefully; therefore, they seem to travel slowly. Another behaviour was the tendency for evacuees to wait for their friends/associates,

130

F. Leonita et al.

Table 12.2 Speed of travelling downstairs, all surveys No. survey 1

2

3

3

4

Descending from which floor Descending from the fourth floor Descending from the fifth floor Descending from the ninth floor Descending from the 19th floor Descending from the 30th floor

Avg pre-observation delay time (s)

Avg time needed to travel downstairs ~69 s (~1 min 9 s)

Speed of travelling downstairs (m/s) ~0.70–0.81

Avg downward speed (m/s) ~0.75

~120 s (~2 min)

~99 s (~1 min 39 s)

~0.42–0.81

~0.51

Upper floor observation and GF: seen queues

~302 s (~5 min 2 s)

~152 s (~2 min 32 s)

~0.62–0.75

~0.66

GF: seen some queues

~356 s (~5 min 56 s)

~369 s (~6 min 9 s)

~0.50–0.73

~0.61

GF: seen some queues

~253 s (~4 min 13 s)

~1130 s (~18 min 50 s)

~0.27–0.46

~0.37

Upper floor: some queues and congestion; GF: seen some queues

Evac. condition Relatively not crowded

Pre-observation delay time(s) is the amount of time needed before evacuees travel downstairs. It is counted from the time the alarm went off until before evacuees enter the exit stair zone GF ground floor

Table 12.3 Crowd movement parameters for various facilities and conditions (SFPE handbook) [22] and survey result SFPE handbook Facility Stair Stair Stair Stair

Crowd condition Minimum Moderate Optimum Crush

Density (m2) HOPO2+M HOPO2+H¼>PO2+H2O PO2+H2O¼>HOPO2+H PO2+H+M¼>HOPO+M HOPO+H¼>H2+PO2 H2+PO2¼>HOPO+H HOPO+O¼>OH+PO2 HOPO+OH¼>PO2+H2O PO2+H2O¼>HOPO+OH PO+OH¼>H+PO2 H+PO2¼>PO+OH PO+O2¼>PO2+O HOPO+H¼>H2O+PO H2O+PO¼>HOPO+H

Eb 7140.0 0.0 38,950.0 36,000.0 4860.0 1000.0 21,000.0 285.0 2040.0 10,726.0 18,567.8 645.0 11,000.0 21,409.1 0.0 1500.0 24,538.7 0.0 19,691.7 0.0 8300.0 14,647.0

n                      

9

10 1014 1012 1013 108 106 1013 1024 1027 1019 1010 1024 1013 108 1013 106 104 1013 1017 1012 1012 105

1.5 0.0 0.0 0.0 1.5 2.0 0.0 2.3 3.0 1.8 0.4 2.0 0.0 1.2 0.0 2.0 2.8 0.0 0.9 0.0 0.0 1.7

The rate constants are expressed as k ¼ AT b exp½E=ðRT Þ In mole, cm3, s b In cal/mol a

60

Validation of the Mechanism Against Laminar Flames’ Speed

Validation of the developed mechanism was performed by comparing the measured and calculated speed of atmospheric-pressure H2/O2/N2, CH4/air, and syngas/air flames doped with TMP. The flame simulation was performed using the starting [12] and skeletal mechanism (Table 63.2). The mechanism [12], except the phosphorusinvolving reactions, contains a submechanism for oxidation of propane, methane, and hydrogen. For the flame simulation using the skeletal mechanism, it was completed with a submechanism for fuel oxidation from [12]. Figure 63.2 shows the speed of H2/O2/N2 flames doped with 400 ppm TMP at a pressure 1 bar at T0 ¼ 308 K versus the equivalence ratio. The inhibition effectiveness of H2/air flames by TMP is too low; therefore, the flames with the dilution ratio D¼[O2]/([O2]+[N2])¼0.1, 0.09, and 0.077 were chosen for the mechanism validation. The choice of diluted flames as an exploration target is connected with experimental aspects and higher inhibition effectiveness of more diluted flames [9]. A higher sensitivity of the flame speed to the additive concentration provides better mechanism refinement. Both mechanisms provide for good

D=0.1 50 flame speed, cm/s

63.3

40

D=0.09

30 20

D=0.077

10 0 0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

φ

Fig. 63.2 The speed of atmospheric-pressure H2/O2/N2/TMP (400 ppm) flames (T0 ¼ 308 K) with various dilution ratios (D) versus the equivalence ratio (ϕ); symbols – experiment, dashed line – modeling using mechanism [12], solid line – modeling using the skeletal mechanism

agreement with the experimental data [9] for flames with D ¼ 0.1 and 0.09. The speed of the most diluted flame (D ¼ 0.077) is predicted less accurately (discrepancy up to 35 %). It may be explained by the low flame speed (10 cm/s and less) and the modest accuracy of measurements. Some disagreements

63

Development and Validation of Skeletal Mechanism for Flame Inhibition by Trimethylphosphate

structure of a flame, especially of concentration profiles of labile species, is a stringent test for a reaction mechanism. The developed skeletal mechanism was also validated by simulating the flame speed for syngas/air mixtures. Earlier [11] we measured the speed for CO+H2 (95:5) mixture with air and 300 ppm TMP. Figure 63.4 shows the measured and simulated atmospheric-pressure speed of syngas/air flames doped with 300 ppm TMP versus the equivalence ratio. Comparison of the experimental results and the predictions of the skeletal and detailed [12] mechanisms demonstrates good agreement of all the data. Modeling overpredicts the speed only for the flame with ϕ ¼ 0.7: 20 % (the mechanism, [12]) and 25 % (the skeletal mechanism).

35

28 flame speed, cm/s

623

21

14

7

0 0.6

0.8

1.0

1.2

1.4

φ

Fig. 63.3 The speed of atmospheric-pressure CH4/air/TMP (600 ppm) flame (T0 ¼ 308 K) versus the equivalence ratio (ϕ); symbols – experiment, dashed line – modeling using mechanism [12], solid line – modeling using the skeletal mechanism

70

56

42

28

14

0 1

2

3

4

5 φ

Fig. 63.4 Speed of atmospheric-pressure syngas (CO:H2¼95:5)/air/ TMP (300 ppm) flame (T0 ¼ 308 K) versus the equivalence ratio (ϕ); symbols – experiment, dashed line – modeling using mechanism [12], solid line – modeling using the skeletal mechanism

between the modeling results are observed for rich flames with D ¼ 0.1 and 0.09. Figure 63.3 shows the speed of atmospheric-pressure CH4/air flames doped with 600 ppm TMP at initial temperature 308 K versus the equivalence ratio. Both mechanisms predict close values of the flames’ speed, which differ only for the richest flames. The experimentally measured speeds of CH4/air/TMP flames [7] are well predicted by both mechanisms (discrepancy does not exceed 15 %). So, the skeletal mechanism satisfactorily predicts the speed of TMP-doped hydrogen and methane flames at atmospheric pressure. Prediction of the chemical

63.4

Application to CFD Simulation of CupBurner CH4/Air Flame

One possible application of the developed skeletal mechanism is numerical modeling of flame extinguishment in fire safety problems. Below, results of laminar flame simulations in co-flow configuration are presented. The aim of these simulations was to determine the minimum concentration of TMP sufficient for extinguishment of a laminar methane flame and validate the result against the data of laboratoryscale experiments. The experimental setup is shown in Fig. 63.5; it is similar to the cup-burner experiments performed earlier for n-heptane fuel, the difference being that gaseous methane was supplied at a fixed flowrate instead of liquid fuel evaporation from a cup. Experiments were performed at the following parameters: total pressure was 1 atm and temperatures of air and methane streams were 75  C. The inner diameter of the outer pipe was 55 mm; inner diameter of the “cup” (methane supply orifice) was 12 mm. The flowrates of air and methane (at 20  C) were 200 cm3/s and 5 cm3/s, respectively. For these conditions, the linear inlet velocities of air and methane (at 75оС) were 10.0 cm/s and 5.2 cm/s; the height of visible methane/air diffusion flame was equal to 50 mm. CFD simulations were performed by an in-house code which solves in axisymmetric geometry the continuity, momentum, and energy equations for multicomponent reactive gas. A chemistry module was developed with the capability of solving the kinetics equations by VODE stiff solver; the module reads in the kinetic scheme, thermodynamic, and transport properties of species from standard CHEMKIN files, which facilitates use of kinetic schemes developed by traditional chemical kinetics tools. The fluid dynamics part of the code relies on implicit coupled pressure-velocity solver with inner iterations to eliminate the splitting errors. In CFD simulations, steady-state solutions were sought by integration of transient equations in pseudo-time until

624

T. Bolshova et al.

r, [m]

temperature:

400

600

800 1000 1200 1400 1600 1800 2000

0 .0 2 0 .0 1 0 -0 .0 5

0

0 .0 5 z, [m]

0 .1

0 .1 5

CO2

r, [m]

0.005 0.015 0.025 0.035 0.045 0.055 0.065 0.075 0.085 0.095 0 .0 2 0 .0 1 0 -0 .0 5

0

0 .0 5 z, [m]

0 .1

0 .1 5

OH

r, [m]

0.0002 0.001 0.0018 0.0026 0.0034 0.0042 0.005 0 .0 2 0 .0 1 0 -0 .0 5

r, [m]

Fig. 63.5 Experimental setup for the measurement of minimum extinguishing concentration of TMP for methane-air laminar diffusion flame

0.02 0.01 0 -0.05

0

0.05

z, [m]

0.1

0.15

r, [m]

Fig. 63.6 Computational domain for CFD simulations

0.02 0.01 0 -0.05

0

0.05 z, [m]

0.1

0.15

Fig. 63.7 Numerical grid used in CFD simulations

converged flow fields and species concentrations are obtained. The experimental geometry shown in Fig. 63.5 was represented by two coaxial pipes presented in Fig. 63.6 (the axis of symmetry is shown horizontally); methane was supplied through the inner pipe, while air-TMP mixture was supplied through the outer pipe. In Fig. 63.7, the numerical grid used in simulations is shown. All solid walls were considered adiabatic with no-slip and zero-flux boundary conditions imposed. Simulations were first carried out with pure air supplied through the outer pipe (0 % TMP). It was obtained that steady-state flame for the given co-flow conditions could not be obtained with some skeletal mechanisms of methane

0

0 .0 5 z, [m]

0 .1

0 .1 5

Fig. 63.8 Structure of methane-air laminar diffusion flame without inhibitor (0 % TMP): temperature, combustion product CO2, intermediate radical OH (top to bottom)

combustion, because the flame was “blown off” by the stream. Imposing constant high temperature of 1000 K on the end surface of the inner pipe did not help either. Several mechanisms of CH4 combustion were tried, and the most successful was the mechanism quoted in [17]. In this study, this skeletal mechanism of CH4 combustion was supplemented by the skeletal TMP mechanism described above. Further studies are necessary to establish the exact reasons for undoped flame blowoff and clarify the sensitivity of results to the choice of skeletal mechanism for fuel oxidation. In Fig. 63.8, the structure of undoped methane-air laminar diffusion flame obtained in the simulations is shown (presented are the distributions of temperature, mole fractions of product CO2, and intermediate radical OH). Further simulations were carried out with TMP added to the air supplied through the outer pipe. The volume fraction of TMP was taken 0.5, 1, 1.25, 1.3, 1.4, 1.5, 2, and 3 %. In each case, the initial conditions were taken those of the steady undoped flame; for control, some simulations were also carried out from the previous solution for lower concentration of TMP (the results turned out to be identical). It was obtained that for volume fraction of TMP in the air equal to 1.4, 1.5, 2, and 3 %, no steady-state flame was possible, while for lower TMP contents combustion persisted. Therefore, the calculated minimum extinguishing concentration can be evaluated as 1.4 % TMP. In the experiments carried out on the facility sketched in Fig. 63.5, the extinguishing concentration was measured by

63

Development and Validation of Skeletal Mechanism for Flame Inhibition by Trimethylphosphate

Table 63.3 Maximum temperature and mole fractions of radicals in methane-air-TMP flame r, [m]

temperature:

300

500

700

625

900 1100 1300 1500 1700 1900 2100

0 .0 2 0 .0 1 0 -0 .0 5

0

0 .0 5 z, [m]

0 .1

0 .1 5

CO2

r, [m]

0.005 0.015 0.025 0.035 0.045 0.055 0.065 0.075 0.085 0.095 0 .0 2 0 .0 1 0 -0 .0 5

0

0 .0 5 z, [m]

0 .1

0 .1 5

OH

r, [m]

0.0001 0.0004 0.0007 0.001 0.0013 0.0016 0 .0 2 0 .0 1 0 -0 .0 5

0

0 .0 5 z, [m]

0 .1

0 .1 5

HOPO

r, [m]

0.0002 0.0008 0.0014 0.002 0.0026 0.0032 0.0038 0.0044 0 .0 2 0 .0 1 0 -0 .0 5

0

0 .0 5 z, [m]

0 .1

0 .1 5

PO 1E-06 r, [m]

gradually increasing the TMP content in the air and registering the instant at which the flame was extinct. It was obtained that flame extinction occurs when the volume fraction of TMP is equal to 1.6  0.2 %. Therefore, our CFD predictions based on the skeletal mechanism presented in the previous sections are in fairly good agreement with the experiment. This can be regarded as a validation of the skeletal mechanism, although more experiments and simulations for other global equivalence ratios (i.e., fuel and air flowrates) are necessary in order to judge on the overall performance of the model and mechanism. In Table 63.3, the calculated maximum values of temperature and mole fractions of radicals (H, O, OH, HOPO, and PO) are presented for each simulated case. The shaded cells correspond to flame extinction. One can see that addition of TMP increases somewhat the maximum flame temperature (by 60–80 K). However, the concentration of free radicals O, H, and OH decreases, which reflects the effect of TMP as an inhibitor, causing flame extinction at high enough concentration of TMP. It is interesting to see the changes in flame structure and characteristics which addition of TMP causes. In Fig. 63.9, distributions of temperature, product (CO2), and radicals (OH, HOPO, PO) obtained for the maximum (of the calculated) concentration at which the flame still exists (1.3 % TMP) are shown. Axial temperature distributions obtained in the cases where flame existed are presented in Fig. 63.10. Importantly, for the pure methane-air case (0 % TFM), the flame length (evaluated by the position of temperature maximum) agrees very well with the experimentally measured value of 5 cm. This confirms the validity of CFD simulation of the flame itself with the adopted methane combustion mechanism. As the contents of TMP in air increases, the flame becomes longer (about 7.5 cm near the extinction limit) because inhibitor reduces the reaction rate.

4E-06

7E-06

1E-05

1.3E-05 1.6E-05 1.9E-05

0 .0 2 0 .0 1 0 -0 .0 5

0

0 .0 5 z, [m]

0 .1

0 .1 5

Fig. 63.9 Structure of methane-air laminar diffusion flame with inhibitor (1.3 % TMP): temperature, combustion product CO2, intermediate radicals OH, HOPO, PO (top to bottom)

An important experimental observation, also confirmed in CFD simulations, concerns the way in which the flame is extinguished at high enough concentrations of TMP. One can see in Table 63.3 that diffusion flame inhibition in co-flow configuration does not occur by gradually decreasing the flame temperature, as might be expected. On the contrary, the maximum temperature in the flame even increases when TMP is added. What happens is that reaction is suppressed near the mixing layer near the fuel source orifice, and the flame detaches itself from the orifice and is blown away by the gas flow. In the experiment, this looks like a mushroom-shaped burning zone detaches from the source and is carried upwards, gradually burning out. CFD simulations are carried out by steady-state solver; however, the solution behavior in pseudo-time is

626

T. Bolshova et al.

configurations. The predicted minimal extinguishing concentration of TMP (1.4 %) agrees fairly well with the measured one (1.6  0.2 %). Further validation studies must include a wider range of global equivalence ratios for co-flow flames, comparison of temperature and species profiles, as well as different fuel compositions. Also, observations of flame extinction mode (flame blowoff from the fuel source, rather than gradual global reaction suppression) emphasize the importance of proper modeling of reactions occurring near the source orifice at temperatures lower than the maximum temperature reached at the flame tip.

2500

Axial temperature, [K]

2000

1500 TMP 0% 0.5% 1% 1.25% 1.3%

1000

500

0 -0.05

0.00

0.05

0.10

Acknowledgment This research was partially supported by Grant RFBR 13-01-12058 ofi_m. 0.15

z, [m]

Fig. 63.10 Axial temperature distributions at different molar fractions of TMP

qualitatively similar: the hot zone detaches from the source and propagates upwards, trailed by cold non-reacting flow. Transient simulations are necessary in order to reproduce this effect quantitatively. Another observation concerns possible effect of radiation losses which were not taken into account in the present study. Generally, radiation losses decrease the flame temperature and affect the flammability limits. However, in this particular problem, it was found that flame inhibition occurs due to blowoff of the flame, i.e., the most critical for inhibition is the zone where the flame attaches to the source, rather than the flame tip where highest temperatures and product concentrations are observed. Therefore, it is expected that in the near-source zone, where temperatures and product concentrations are lower, radiation effects will be limited. The comprehensive study of radiation effects has yet to be performed in further validation of the skeletal mechanism by CFD simulations.

63.5

Conclusion

Thus, in this paper a skeletal mechanism for flame inhibition by organophosphorus compounds is presented. Its adequacy is confirmed by validation against the detailed mechanism predictions and experimental measurements for flame propagation speed as a function of the equivalence ratio. CFD simulations also confirmed validity of the mechanism for prediction of laminar flames in co-flow

References 1. Law CK (2007) Combustion at a crossroads: status and prospects. Proc Combust Inst 31:1–29 2. Mauss F, Peters N, Rogg B, Williams FA (1993) In: Peters N, Rogg B (eds) Reduced reaction mechanisms for premixed hydrogen flames, in reduced kinetics mechanisms for applications in combustion systems, in lecture notes in physics. Springer, New York, pp 29–43 3. Mauss F, Peters N (1993) In: Peters N, Rogg B (eds) Reduced reaction mechanisms for premixed methane-Air flames, in reduced kinetics mechanisms for applications in combustion systems, in lecture notes in physics. Springer, New York, pp 58–75 4. Boivin P, Jimenez C, Sanchez AL, Williams FA (2011) An explicit reduced mechanism for H2-air combustion. Proc Combust Inst 33:517–523 5. Boivin P, Jimenez C, Sanchez AL, Williams FA (2011) Electrical characteristic of diamond film synthesized by combustion flame. Combust Flame 158:1059–1063 6. Korobeinichev OP, Shvartsberg VM, Shmakov AG, Bolshova TA, Jayaweera TM, Melius CF, Pitz WJ, Westbrook CK, Curran HJ (2005) Flame inhibition by phosphorus-containing compounds in lean and rich propane flames. Proc Combust Inst 30:2353–2360 7. Korobeinichev OP, Shvartsberg VM, Shmakov AG, Knyazkov DA, Rybitskaya IV (2007) Inhibition of atmospheric lean and rich CH4/O2/Ar flames by phosphorus-containing compound. Proc Combust Inst 31:2741–2748 8. Jayaweera TM, Melius CF, Pitz WJ, Westbrook CK, Korobeinichev OP, Shvartsberg VM, Shmakov AG, Curran HJ (2005) Flame inhibition by phosphorus-containing compounds over a range of equivalence ratios. Combust Flame 140:103–115 9. Korobeinichev OP, Rybitskaya IV, Shmakov AG, Chernov AA, Bolshova TA, Shvartsberg VM (2009) Inhibition of atmosphericpressure H2/O2/N2 flames by trimethylphosphate over range of equivalence ratio. Proc Combust Inst 32:2591–2597 10. Korobeinichev OP, Rybitskaya IV, Shmakov AG, Chernov AA, Bolshova TA, Shvartsberg VM (2010) Mechanism of inhibition of hydrogen/oxygen flames of various compositions by trimethylphosphate. Kinet Cat 51(2):154–161

63

Development and Validation of Skeletal Mechanism for Flame Inhibition by Trimethylphosphate

11. Shvartsberg VM, Shmakov AG, Bolshova TA, Korobeinichev OP (2012) Mechanism for inhibition of atmospheric-pressure syngas/ air flames by trimethylphosphate. Energy Fuels 26(9):5528–5536 12. Jayaweera TM, Melius CF, Pitz WJ, Westbrook CK, Korobeinichev OP, Shvartsberg VM, Shmakov AG, Curran HJ (2004) Organophosphorus compounds effect on flame speeds over a range of equivalence ratios. Available at https://www-pls.llnl.gov/? url¼science_and_technology-chemistry-combustion-organophos phorus_over_range 13. Yong J, Rong Q (2010) A reduced mechanism for flame inhibition by phosphorus-containing compounds based on level of importance analysis. Chin J Chem Eng 18(5):711–720 14. Shanling L, Yong J, Rong Q (2013) The generation of a reduced mechanism for flame inhibition by phosphorus containing

627

compounds based on path flux analysis method. Chin J Chem Eng 21(4):357–365 15. Rybitskaya IV, Shmakov AG, Shvartsberg VM, Korobeinichev OP (2008) Effect of the equivalence ratio on the effectiveness of inhibition of laminar premixed hydrogen–air and hydrocarbon–air flames by trimethylphosphate. Combust Expl Shock Waves 44(2):133–140 16. Turanyi T. Mechmod v. 1.4: program for the transformation of kinetic mechanisms. Available at http://garfield.chem.elte.hu/Com bustion/mechmod.htm 17. Rogg B (1991) Sensitivity analysis of laminar premixed CH4-air flames using full and reduced kinetic mechanisms. In: Smooke MD (ed) Reduced kinetic mechanisms and asymptotic approximations for methane-air flames, lecture notes in physics, vol 384. Springer, Berlin/New York, p 159

Part XVI Flashover

64

A Validation Study of Existing Formulas for Determining the Critical Heat Release Rate for Flashover SungChan Lee and Kazunori Harada

Abstract

Many studies have been carried out to establish the boundary of flashover occurrence in compartment fires. This study gathered experimental data from the existing literatures and examined the effect of factors such as size of compartment, wall and ceiling materials, and so on. It was shown that the difference between full-scale experiments and model-scale experiments is large if the thermal inertia of the wall and ceiling materials were not taken into account but correlated with the ratio of opening factor to total internal surface area. However, by rearranging existing data considering the thermal inertia of the wall, it was found that the critical heat release rates for flashover in model-scale and full-scale tests follow almost the same correlations. The existing formulas were compared to the collected experimental data. Most of the formulas tend to fit to experimental data which were used when each specific formula was derived. Babrauskas and Thomas formulas result in overestimate of critical heat release rate for rooms with low thermal inertia walls and underestimate for rooms with large thermal inertia. Ha¨gglund’s formula results in overestimate for rooms with low thermal inertia walls and good estimate for rooms with large thermal inertia. The best estimate was obtained by MQH (McCaffrey, Quintiere, and Herkeload) formula. If the value of actual time to flashover is known, the formula gives fairly lower bound of critical heat release rate for flashover. For practical calculations, use of 1000s for characteristic time is recommended for conservative design calculations. Keywords

Flashover
Heat release rate Experiment

Nomenclature A A√H AT AT/A√H H

Opening area (m2) Opening factor (m5/2) Internal surface area of compartment (m2) Temperature factor (m1/2) Height of opening (m)

S. Lee (*)
K. Harada Graduate School of Engineering, Kyoto University, C1-487 Kyoto University Katsura Campus, Nishikyo, Kyoto, Japan e-mail: [email protected]; [email protected]




Opening factor

Q QFO tc tFO




Thermal inertia of the wall




Formula




Heat release rate of fire (kW) Critical heat release rate for flashover (kW) Characteristic time (s) Time to flashover (s)

Greek Symbols √kρc c k ρ

Thermal inertia of the wall (kWs1/2/m2 K) Specific heat (kJ/kg K) Thermal conductivity (kW/m2K) Density (kg/m3)

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_64

631

632

64.1

S. Lee and K. Harada

Introduction

As the flashover is a key event in compartment fire development, many existing studies are focused on establishing the criteria for onset of flashover. Many experiments have been carried and various formulas have been proposed [1]. However, the experimental approach to determine the flashover boundary is quite difficult and limited by practical problems. The experiments for flashover are accompanied with dangerous procedure and high costs to carry out many times. The size of compartment, selection of fire source, type and amount of fuels, and so on are highly constrained. Thus, the experimental formulas derived according to those constrained experimental conditions could be limited to a certain range of conditions. Comparison of formulas with experimental datasets was carried out by Peacock et al. [2]. However, the effects of wall lining material and compartment size were not well known. To compare between existing formulas and to discuss the range of application, this study collects experimental data from the existing literatures and correlates the critical heat release rates for flashover with experimental conditions. The effects of factors such as the compartment size and wall lining materials on the flashover are examined. In addition, the validation and range of application of existing formulas are examined by comparing with collected whole set of experimental data.

64.2

Review of Existing Formulas

In order to derive critical heat release rate for flashover, most of the existing studies considered the heat balance of the fire room. The heat release rate in a fire room is balanced with the sum of heat absorption rate by wall enclosures and heat exhaust rate from opening. Existing studies suggested simple formulas to determine the flashover occurrence, having opening factor and the thermal inertia of wall as parameters after analyzing available experimental data. In general, flashover is correlated on the assumption that the upperlayer temperature reaches around 500–600  C or the radiative heat flux to the floor becomes 20 kW/m2 or above. As will be described, Babrauskas, Thomas, and Ha¨gglund’s formulas did not include the thermal inertia of wall. In case of MQH and Chen’s formulas, the thermal inertia of wall was included explicitly to the formula. Babrauskas suggested the formula by comparing with full-scale and model-scale experimental data [7, 11]. By

considering the heat balance of upper layer of fire room at 600  C, the following formula was proposed: pffiffiffiffi  QFO =A H ¼ 1520 exp 1 f 2  1:96 h ½0:799=f p ffiffiffiffi2=3 i  f 1 ¼ 1  0:94 exp 33 AT =A H h pffiffiffiffi0:6 i  f 2 ¼ 1  0:92 exp 11:9 AT =A H

ð64:1Þ

Thomas also developed an equation based on heat balance of room [10]. The coefficients were determined by comparison with Ha¨gglund’s experimental data [8]. h  pffiffiffiffi pffiffiffiffii QFO =A H ¼ 378 1 þ 0:021 AT =A H

ð64:2Þ

Ha¨gglund et al. proposed the following formula from calculation results by a zone model [4] and compared with their own experiments [8]. The criterion for flashover was 600  C in the upper layer of the room. pffiffiffiffi QFO =A H ¼ 1050



AT pffiffiffiffi A H

"

1:2 pffiffiffiffi þ 0:247  AT =A H

#3

ð64:3Þ To determine the upper-layer temperature in a compartment fire, McCaffrey et al. conducted a regression analysis. Assuming that the flashover occurs when the upper-layer temperature reaches 500  C, the following formula was proposed [9]: qp ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffi pffiffiffiffiffi ð64:4Þ kρc=tc AT =A H QFO =A H ¼ 610 Similar to McCaffrey’s work, Chen derived an empirical equation for upper-layer temperature from experimental data. They assumed that flashover takes place when upperlayer temperature becomes 600  C. The critical heat release rate for flashover is estimated by [16]: pffiffiffiffiffiffiffi0:07  0:696 pffiffiffiffi0:52  pffiffiffiffi QFO =A H ¼ 118:3 kρc AT =A H ð64:5Þ Tables 64.1 and 64.2 show the range of experimental conditions which are referred to when each formula was derived. The thermal inertia, size of room, and opening are summarized. The range of conditions is different by formulas. Therefore, it is necessary to review whether each formulas can be applied to various thermal inertia of wall materials and sizes of compartment.

64

A Validation Study of Existing Formulas for Determining the Critical Heat Release. . .

633

Table 64.1 Thermal inertia of walls in experiments compared with existing formula Formulas Babrauskas Thomas Ha¨gglund McCaffrey Chen

Scale of room Full scale Model scale Full scale

The group for thermal inertia of wall [kWs1/2/m2K] (Small:0.10–0.40, medium:0.40–1.00, large:1.00–1.60) Large, medium (main material: normal-weight concrete, etc.) Small (main material: asbestos board, etc.) Large (main material: normal-weight concrete, lightweight concrete, etc.)

Full scale Model scale Full scale Model scale

Large, medium (main material: normal-weight concrete, etc.) Medium, small (main material: ceramic fiber, etc.) Large, medium (main material: normal-weight concrete, etc.) Medium, small (main material: unknown)

Table 64.2 The range of room and opening dimensions in experiments compared with existing formula Full scale Formulas Babrauskas Thomas Ha¨gglund McCaffrey Chen

2

AT[m ] (46.0–127.7) 56.1–96.9 (56.1–96.1) 29.1–57.5 (44.2–57.5) 30.52–57.71

pffiffiffiffi

A H m5=2 (0.78–7.51) 0.22–5.34 (0.89–5.34) 0.22–2.83 (0.22–1.96) 0.22–2.83

pffiffiffiffi

AT =A H m1=2 (16.0–65.0) 16.7–258.5 (16.7–64.0) 14.9–258.5 (20.3–258.5) -

Model scale AT[m2] (0.84–7.32) 

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A H ½m5=2  (0.03–0.68) 

pffiffiffiffi

AT =A H m1=2 (9.0–50.0) 

0.78–0.88 (0.78–0.88) 0.852–1.12 (0.852–1.12)

0.02–0.04 (0.02–0.04) 0.002–0.03 (0.002–0.03)

21.9–51.2 (21.9–51.2) 32.9–199.0 (32.9–199.0)

Values, the range of experimental condition; (values), the range of experimental condition (FO occurrence); , no data; -, unknown

64.3

Analysis of Data from Existing Experiments

64.3.1 Method of Analysis 64.3.1.1 Data on Whether Flashover Occurs or Not (Method 1) In existing studies on flashover, authors of existing formulas collected their own and other researcher’s data. The collected experimental data may correspond with various conditions. Flashover occurred in some of the cases. In such a case, boundary of flashover occurrence can be directly estimated from experimental datasets. The critical heat release rate was determined by the mean of maximum heat release rate in experiments without flashover and minimum heat release rate in experiments with flashover. 64.3.1.2 Data When Flashover Occurs (Method 2) Some literature does not include enough number of experimental conditions so as to apply method 1. However, in the case of literatures with data on upper-layer temperature and heat release rate histories available by graphs, a heat release rate when smoke-layer temperature reaches around 600  C or when the radiative heat flux to the floor of compartment

becomes 20 kW/m2 is considered as critical heat release rate for flashover.

64.3.2 Critical Heat Release Rate for Flashover in an Experiment It was possible to collect a total of 75 data on critical heat release rate for flashover in ten existing studies [3, 5, 6, 8, 12–15, 17]. Full-scale experimental data were collected from seven studies and model-scale experimental data from three studies. Table 64.3 shows the range of the ratio of opening factor to total internal surface area (so-called temperature factor), opening factor, and internal surface area of compartment in all collected experimental data. The range of thermal inertia of wall was divided into three groups (large, medium, small). Their ranges are shown in Table 64.4. In case of a large thermal inertia group, compartments were constructed by high-density materials such as normalweight concrete and lightweight concrete mostly. In case of medium thermal inertia group, compartments were constructed by plaster board and ALC (autoclaved lightweight concrete) panel mostly. In case of small thermal inertia group, compartments were lined with lightweight

634

S. Lee and K. Harada

Table 64.3 The range of surface area, opening factor, and temperature factor of compartment in all experiments pffiffiffiffi Scale AT A H Full scale 43.5–119.9 0.89–14.5 Model scale 0.79–1.05 0.002–0.04 All 0.79–119.9 0.002–14.5

Table 64.4 The range of thermal inertia of walls in all experiments pffiffiffiffiffiffiffi kρc Scale Small Full scale 0.24 Model scale 0.14–0.34 All 0.10–0.40

pffiffiffiffi AT =A H 7.90– 64.0 23.7–530.1 7.90–530.1

Medium 0.53–0.87 0.41–0.64 0.40–1.00

insulating materials. From the collected data, it was possible to investigate the effect of compartment size, opening size, and the wall and ceiling lining materials on critical heat release rate for flashover.

Large 1.15–1.52 (No data) 1.00–1.60

1500

64.3.2.2 Effect of Thermal Inertia of Wall pffiffiffiffi pffiffiffiffiffiffiffi pffiffiffiffi The correlation between AT kρc=A H and Q=1500A H in all experimental data were shown in Fig. 64.2. Comparing the minimum values (solid line) of heat release rate with flashover, the normalized heat release rate of model-scale experiments is higher than full-scale experiments. However, the normalized heat release rate of full-scale experiments without flashover (dashed line) is higher than model-scale experiments. Therefore, the range of critical heat release rate of fullscale experiments is included in the gap between solid line and dashed lines for model-scale experiments.

Q/AH1/2 [kW/m5/2]

1200

64.3.2.1 Correlations with Temperature Factor pffiffiffiffi The correlation between temperature factor AT =A H and pffiffiffiffi normalized heat release rate Q=A H in all experimental data is shown in Fig. 64.1. Comparing the minimum values (solid line) of heat release rate in data when flashover occurs in full-scale and model-scale tests, the normalized critical heat release rate of full-scale experiments is higher than that of model-scale tests. In addition, the normalized heat release rate of full-scale experiments without flashover (dashed line) is higher than model-scale experiments. Comparing gaps between minimum value with flashover (solid line) and maximum value without flashover (dashed line), the gap of model-scale experiments is larger than that of full-scale experiments. This is because existing experimental data is lacking to be comparable in the same condition in case of model-scale experiments. By definition, critical heat release rate for flashover lies in between solid and dashed lines in Fig. 64.1, possibly around the middle of them.

900

600

300

0 0

30

60 AT/AH

1/2

90 [m

-1/2

120

]

Full-scale (with F.O) Full-scale (no F.O) Model-scale (with F.O) Model-scale (no F.O) Full-scale (boundary of F.O) Full-scale (boundary of non-F.O) Model-scale (boundary of F.O) Model-scale (boundary of non-F.O) pffiffiffiffi pffiffiffiffi Fig. 64.1 The comparison of the correlation of AT =A H with Q=A H between full-scale and model-scale experiments (lines, boundary of flashed and non-flashed fires; open symbols, non-flashed experiments; filled symbols, flashed experiments)

Assuming that the normalized critical heat release rate is around the middle of solid and dashed lines, normalized critical heat release rates for flashover in full-scale

64

A Validation Study of Existing Formulas for Determining the Critical Heat Release. . .

635

1500

1.6

1.4

Babrauskas’s formula

1200

QFO/AH1/2 [kW/m5/2]

Q/1500AH1/2 [-]

1.2

1

0.8

0.6

900

600

300

0.4

small thermal inertia (Full-scale data)

0.2 0

100

0 0 0

30

60

90

120

150

180

AT(kρc)1/2/AH1/2 [-] Full-scale (with F.O) Full-scale (no F.O) Model-scale (with F.O) Model-scale (no F.O) Full-scale (boundary of F.O) Full-scale (boundary of non-F.O) Model-scale (boundary of F.O) Model-scale (boundary of non-F.O)

pffiffiffiffiffiffiffi pffiffiffiffi Fig. 64.2 The comparison of the correlation of AT kρc=A H with pffiffiffiffi Q=1500A H between full-scale and model-scale experiments (lines, boundary of flashed and non-flashed fires; open symbols, non-flashed experiments; filled symbols, flashed experiments)

experiments and model-scale experiments are almost equal to one another.

64.4

Comparison of Existing Formulas with Experimental Data

64.4.1 Formulas Without Consideration of Thermal Inertia of Wall Babrauskas, Thomas, and Ha¨gglund’s formulas include the surface area of room and opening factor as parameters not included the thermal inertia of the wall. The correlation of pffiffiffiffi temperature factor AT =A H and normalized heat release rate pffiffiffiffi QFO =A H in each formula and all experimental data were shown in the following. Figure 64.3 shows the results of comparing Babrauskas’s pffiffiffiffi formula with all experimental data. As AT =A H increases,

/AH1/2

AT

Full (√kρc-Small) Full (√kρc-Large) Model (√kρc-Medium) Babrauskas formula

200

[m-1/2] Full (√kρc-Medium) Model (√kρc-Small) Full (√kρc-1.47, used)

pffiffiffiffi pffiffiffiffi Fig. 64.3 The comparison of the correlation of AT =A H with Q=A H between Babrauskas’s formula and all experimental data

the critical heat release rate for flashover becomes larger. The experimental data of the model scale are more scattered than the full scale. The slope of model-scale data is different from that of full-scale data. The critical heat release rate for flashover is larger in full-scale experiments than in model-scale experiments. A value calculated by Babrauskas’s formula is smaller than the lower bound of full-scale experimental data and is a bit higher than median value of model-scale experimental data. However, it is almost close to the lower bound of model-scale experimental data which the group of thermal pffiffiffiffi inertia is medium. In case AT =A H is less than 20, the Babrauskas’s formula is almost close to the median value pffiffiffiffi of experimental data. However, as AT =A H increases, a gap between experimental data and formula becomes larger. There is a relatively large error among values calculated from Babrauskas’s formula and experimental data. This is because, among collected experimental data, a few data (black dot) is used when Babrauskas’s formula is derived. It may be that the experimental condition of collected data is remarkably different from the data used in deriving the formula. Figure 64.4 shows the results of comparing Thomas’s formula with all experimental data. Thomas’s formula shows almost similar to that of Babrauskas’s formula.

636

S. Lee and K. Harada

Figure 64.5 shows the results of comparing Ha¨gglund’s formula with all experimental data. Ha¨gglund’s formula is well match with full-scale experimental data of the group of large thermal inertia. This is because a lot of large thermal inertial full-scale experimental data (black dot) were referred to when the formula was derived. However, it is highly different from small thermal inertia experimental data. Therefore, Ha¨gglund’s formula can be used only in the case of large thermal inertia full-scale experiments. All of three formulas were different from small thermal inertia experimental data. Therefore, there is a possibility that the value of formulas will be the overestimated case of rooms with low thermal inertia walls.

1500 Thomas’s formula

QF0/AH1/2 [kW/m5/2]

1200

900

600

300

64.4.2 Formulas in Consideration of Thermal Inertia of Wall

small thermal inertia (Full-scale data)

0

MQH and Chen’s formula include the thermal inertia of wall qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffi as a parameter. The correlation of AT kρc=A H and QFO = pffiffiffiffi Full (√kρc-Small) Full (√kρc-Medium) 1500A H in formula and experimental data were shown in Full (√kρc-Large) Model (√kρc-Small) the following figures. Model (√kρc-Medium) Full (√kρc-Large, used) qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Thomas formula pffiffiffiffiffi First of all, the correlation of AT kρc=tc =A H and QFO pffiffiffiffi pffiffiffiffi pffiffiffiffi Fig. 64.4 The comparison of the correlation of AT =A H with Q=A H =1500A H in MQH formula and experimental data were between Thomas’s formula and all experimental data examined by using experimental data where time to flashover is known. The result is shown in Fig. 64.6. Modelscale experimental data are excessively larger than the 1500 value of formula. The value of formula is close to the minimum value of full-scale experimental data. By seeing Hägglund’s the data close to MQH formula, no systematic differences formula 1200 are found between full-scale and model-scale experimental data. In case of using MQH formula for practical design situa900 tion, characteristic time (so-called tc ) shall be assumed properly. To discuss the appropriate selection of characteristic time, MQH formula was compared with all data in Fig. 64.7. In this plot, horizontal axis does not include 600 characteristic time, but it is shown by parameter. Data close to MQH formula in Fig. 64.6 were extracted and compared with calculations with 340 and 1000 s of 300 characteristic time. The value of 1000 s was taken as suggested by original McCaffery’s paper [9]. Using this small thermal inertia value, calculation formula corresponds with lower bound (Full-scale data) of experimental data. Therefore, MQH formula with 1000 s 0 100 200 of characteristic time could be used for conservative fire 0 safety design. AT/AH1/2 [m-1/2] The value of 340 is the average time to flashover in the Full (√kρc-Small) Full (√kρc-Medium) experiments extracted. Using this value for characteristic Model (√kρc-Small) Full (√kρc-Large) Model (√kρc-Medium) Full (√kρc-Large, used) time, the formula is almost close to the mean of experimental Hägglund formula data. Also, it is possible to verify that the formula is exactly pffiffiffiffi pffiffiffiffi close to the experimental data, shown by large black circles, Fig. 64.5 The comparison of the correlation of AT =A H with Q=A H used when MQH formula was derived. In addition, the between Ha¨gglund’s formula and all experimental data 0

100

QF0/AH1/2 [kW/m5/2]

AT/AH1/2 [m-1/2]

200

64

A Validation Study of Existing Formulas for Determining the Critical Heat Release. . .

637

1.2

2.0

Chen’s formula 1.0

QF0/1500AH1/2 [-]

QF0/1500AH1/2 [-]

1.5

1.0

0.8

0.6

0.4

0.5

0.2

small thermal inertia (Full-scale data)

MQH formula eq.(4)

0.0

0

Full (√k ρc-Small)

Full (√k ρc-Medium)

Full (√k ρc-Large)

Model (√k ρc-Small)

Model (√k ρc-Medium)

McCaffrey formula

pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffi Fig. 64.6 The comparison of the correlation of AT kρc=tc =A H and pffiffiffiffi QFO =1500A H between MQH formula and all experimental data where time to flashover is known

1.2

0.8 QF0/1500AH1/2 [-]

Full (√k ρc-Small) Full (√kρc-Large) Modle (√kρc-Medium) Modle (√kρc-Small, used) Chen formula

[-]

Full (√kρc-Medium) Modle (√kρc-Small) Full (√kρc-Large, used) Modle (√kρc-Medium, used)

pffiffiffiffiffiffiffi0:07 pffiffiffiffi0:52 Fig. 64.8 The comparison of the correlation ofAT 0:696 kρc =A H pffiffiffiffi and QFO =1500A H between Chen’s formula and all experimental data

agreement is good for full-scale experiments with large thermal inertia, as shown by large gray circles. However, experimental value is larger in case of full-scale experiments with medium thermal inertia. In contrast, experimental values are far smaller than formula in case of full-scale experiments with small thermal inertia. Therefore, the forqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffi mula will be usable if the AT kρc=A H is more than

MQH formula (tc - 340s)

1.0

(AT)0.696(√kρc)0.07/(A√H)0.52

6

3 AT(kρc/tc)1/2/AH1/2 [-]

20

10

0

0.0

certain value. Figure 64.8 shows the comparison of Chen’s formula with experimental data. The Chen’s formula is almost close to the upper bound of experimental data. The formula tends to be close to model-scale experiments than to fullscale experiments. This is because model-scale experimental data were mostly used when the Chen’s formula was derived.

0.6 MQH formula (tc - 1000s) 0.4

0.2 small thermal inertia (Full-scale data) 0.0 0

100

200

AT(kρc)1/2/AH1/2 [-] Full (√kρc-Small) Full (√kρc-Large) Model (√kρc-Medium) Model (√kρc-0.15, used) McCaffrey (tc-1000s)

Full (√kρc-Medium) Model (√kρc-Small) Full (√kρc-1.47, used) McCaffrey (tc-340s)

pffiffiffiffiffiffiffi pffiffiffiffi Fig. 64.7 The comparison of the correlation of AT kρc=A H and pffiffiffiffi QFO =1500A H between MQH formula and all experimental data

64.5

Conclusions

In this study, the critical heat release rates for flashover have been gathered from literature and compared with existing formulas for critical heat release rate. 1. According to the results of analysis of the experimental data, it was observed that the critical heat release rates for

638

flashover in the model scale and full scale were almost equal to one another if the effect of thermal inertia of wall was taken into account. 2. All the formulas tend to be fairly close to experimental data which were referred to when the formulas were derived. However, in other conditions, some of the formulas do not correspond with experimental data other than referred data. In cases of Babrauskas and Thomas formulas, formulas result in overestimate of critical heat release rate for rooms with low thermal inertia walls and underestimate for rooms with large thermal inertia. Ha¨gglund’s formula results in overestimate for rooms with low thermal inertia walls and good estimate for rooms with large thermal inertia. 3. MQH formula results in good estimate of lower bound of critical heat releaser rate if actual time to flashover is known. For practical situation with no knowledge of time to flashover, characteristic time of 1000 s is appropriate to calculate lower bound of critical heat release rate.

Acknowledgment Part of this research has been supported by the grant-in-aid for scientific research by Japan Association for Promotion of Science (Principal Investigator: Kazunori Harada, Grant No. 2628904).

References 1. Walton D, Thoms P (2008) Estimating temperatures in compartment fires, SFPE handbook on fire protection engineering. In: DiNenno et al (ed) 4th edn. Section 3, Chapter 6

S. Lee and K. Harada 2. Peacock R, Reneke P, Bukowski RW, Babrauskas V (1999) Defining flashover for fire hazard calculations. Fire Saf J 32:331–345 3. Heselden AJM, Smith PG, Theobald CR (1966) Fires in a large compartment containing structural steelwork, Detailed Measurements of Fire Behaviour, F.R.Note646, Fire Research Station Borehamwood 4. Ha¨gglund B, Jansson R, Onnermark B (1974) Fire development in residential rooms after ignition from nuclear explosions, FOA Report C20016-D6-A3. Forsevarets Forskingsantalt, Stockholm 5. Quintiere JG, McCaffrey BJ, DenBraven K (1978) Experimental and theoretical analysis of quasi-steady small-scale enclosure fires, NBSIR 78-1511, National Bureau of Standards 6. Fang JB, Breese JN (1980) Fire development in residential basement rooms, NBSIR 80-2120. [US]Natl. Bur. Stand, Gaithersburg 7. Babrauskas V (1980) Estimating room flashover potential. Fire Technol 16:94–103 8. Ha¨gglund B (1980) Estimating flashover potential in residential rooms, FOA Rapport C20369-A3. Forsvarets Forkningsanstalt, Stockholm 9. McCaffrey BJ, Quintiere JG, Harkleroad MF (1981) Estimating room temperatures and the likelihood of flashover using fire data correlations. Fire Technol 17:98–119 10. Thomas P (1981) Testing products and materials for their contribution to flashover in room. Fire Mater 5:103–111 11. Babrauskas V (1984) Upholstered furniture room fires – measurements, comparison with furniture calorimeter data, and flashover predictions. J Fire Sci 2:5–19 12. Lee BT (1985) Quarter-scale room-fire tests of interior finishes. Fire Mater 9(4):185–191 13. Holborn B, Bishop S, Drysdale D, Beard A (1993) Experimental and theoretical models of flashover. Fire Saf J 21(3):257–266 14. Alex C, Bwalya AC (2005) Design fires for commercial, premises – results of phase I, Internal Report 868, Project Number B4231, September, 2005 15. Bwalya AC, Zalok E, Hadjisophocleous G (2006) Design fires for commercial premises – results of phase 2, Institute for Research in Construction, National Research Council Canada 16. Chen A, Yang S, Dong X (2011) Studies of the combined effects of some important factors on the likelihood of flashover. Fire Mater 35 (2):105–114 17. Chen A, Francis J, Dong X, Chen W (2011) An experimental study of the rate of gas temperature rise in enclosure fires. Fire Saf J 46:397–405

Experimental Study of Time to Onset of Flashover in Classroom Size Compartment

65

Tomohiro Naruse, Koji Kagiya, Jun-ichi Suzuki, Noboru Yasui, and Yuji Hasemi

Abstract

A 3-year experimental research project, involving full-scale fire tests, was undertaken to investigate whether quasi-fire-resistive wooden building with supplemental fire safety measures, if necessary, could reach the equivalent level of the fire-resistive building, especially for safe egress of occupants. Flashover is one of the most important phenomena, relating to safe egress of the occupants. In this paper, the results of five types of fire tests are reported to investigate the effects of wood interior finish and loaded combustibles within the compartment with a floor area of 63 m2 and 150 m2, on the time to onset of flashover. As the results, in the cases that fire source grew fast in the compartment with noncombustible interior finish and that fire source was 240 kW in the compartment with wood ceiling finish, the flashover occurred in 330–420 s. Equation estimating the time to onset of flashover in the compartments, which have the fire source of 240 kW, was proposed, but more data are needed to improve the correlation. Keywords

Flashover
Full-scale fire test
Wood finish
Loaded combustibles

Nomenclature A t

Surface area of (m2) Time of (m)

Subscripts FO T

Flashover Internal of compartment

T. Naruse (*)
K. Kagiya Building Research Institute, Tachihara 1, Tsukuba, Ibaraki 305-0802, Japan e-mail: [email protected] J. Suzuki National Institute for Land and Infrastructure Management, Tachihara 1, Tsukuba, Ibaraki 305-0802, Japan

65.1

Introduction

In the building standard law of Japan [1], a special building, such as a three-storey school building, is required to be fire resistive from the viewpoint of safe egress of the occupants, as well as search and rescue by the fire brigades. The promotion of the use of wood in public buildings (enacted in October 2010) requires the promotion of research to review building regulations which may have been limiting the use of wood in buildings, because of the need to reduce carbonation and to encourage sustainable forest management. With this background, a 3-year research project was undertaken from the fiscal year 2011–2013. Full-scale fire tests were conducted to assess whether a quasi-fire-resistive wooden building with supplemental fire safety measures, if neces-

N. Yasui
Y. Hasemi Waseda University, Okubo 3-4-1, Shinjuku-ku, Tokyo 169-8555, Japan

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_65

639

640

T. Naruse et al.

sary, is able to reach the equivalent level of a fire-resistive building, to facilitate safe occupant egress, and fire brigade search and rescue. Fire-resistive buildings are defined as buildings of which the principal building parts come under fire-resistive construction. Quasi-fire-resistive buildings are defined as buildings of which the principal building parts come under quasi-fire-resistive construction. Fire-resistive construction shall have fire-resistive performance (loadbearing capacity, insulation, and integrity) during and after a certain period of fire exposure. Quasi-fire-resistive construction shall have fire-resistive performance during a certain period of fire exposure [1]. Flashover is one of the most important phenomena, affecting a safe egress of occupants, not only from the compartment of fire origin but also from other compartments. In this paper, the test results are reported to investigate the effects of wood interior finish and loaded combustibles in the compartment with a floor area of 63 m2 and 150 m2, on the time to onset of flashover.

65.2

Previous Study of Time to Flashover

Because the onset of flashover in the compartment is a very important event, the time to flashover (tFO) and the heat release rate (HRR) at flashover are studied experimentally and are used as physical indices. The flashover is defined by ISO 13943 [2] as “the transition to a state of total surface involvement in a fire of combustible materials within an enclosure.” Babrauskas et al. [3] define flashover as (1) the occurrence of criticality in a thermal balance sense and (2) a fluid-mechanical filling process. They also explain the flashover event cause and effect. They mention that flashover could be quantitatively described as corresponding to a gas temperature of 600  C or a floor irradiance of 20 kW/m2 or possibly a number of other related, though not necessarily identical, occurrences [4]. As for the upper hot gas layer temperature at the time of flashover, different values, such as 500  C [5], 450  C [6], and temperature exceeding 350  C [7], are reported. As for the HRR at the time of flashover in an ISO 9705 room corner test [8], Babrauskas et al. [3] mention that 1700 kW was more suitable to describe a typical HRR value, rather than 1000 kW defined by ISO 9705. As for the HRR at flashover, the convective and the conductive heat loss through the compartment boundaries has been studied as a parameter [3–5, 9]. Lee and Harada [9] correlate the time to flashover in Eq. 65.1 from the previous test results and Eq. 65.2 from the results of calculation of the t-squared fires. tFO ¼ 914:75AT 0:28

ð65:1Þ

tFO ¼ 0:0011AT 2:43

ð65:2Þ

The effect of the loaded combustibles on the initial fire intensity was studied experimentally, and the arrangement of them played an important role on fire spread [10]. The

factors influencing the fire development in a compartment are: 1. The size and location of fire source: The greater the energy releases, the quicker fire grows. The fire source on the floor grows rapidly. 2. The combustibles: The combustible wall and ceiling finish can cause rapid fire growth. 3. The position of burning object: The surface of burning materials placed perpendicularly may cause rapid fire spread. The burning object in the corner of the room makes fire grow rapidly [11].

65.3

Plan of Fire Tests

In this research, safe egress in the case of a school fire and the use of wood for the structural material and interior finish is considered. We have already learned from the full-scale fire test in a wooden school building [12] that the fire grows very fast to flashover in the room, where wood is used for the ceiling finish. From the viewpoint of the wood use promotion law in Japan, we planned to use wood for wall and floor finishing, as much as possible. A sprinkler system is considered to be effective to suppress the fire; therefore, it was excluded from the fire tests. The size of the fire source in actual school fires was surveyed to make a suitable test plan for school fire tests, because it influences the fire development in a compartment.

65.3.1 Fire Source in School Fires Fire reports from the Fire and Disaster Management Agency during 1995–2008 in Japan [13] were surveyed. “School” of business type and “building fire” of fire type were included in search condition, and “arson,” “suspicious fire,” “playing with fire,” and “an unconscious ignition” were excluded. A total 2152 examples of school fires were searched. As for the source of ignition, top three are 547 “unknown,” 403 “cigarette,” and 49 “electric cable.” Two major sources of ignition accounting for about half the ignition process and the ignited material are shown in Table 65.1. It shows that the ignited materials are the small common nearby things. Mean time of

Table 65.1 Heat of ignition, ignition process, and ignited material Ignition source Uncertain (547 examples)

Ignition process Uncertain (537 examples)

Cigarette (403 examples)

Remaining in the inappropriate place (301 examples)

Ignited material Uncertain (332 examples) Wastepaper (35 examples) Garbage waste (27 examples) Garbage waste (93 examples) Wastepaper (92 examples) Bag, paper products (34 examples)

65

Experimental Study of Time to Onset of Flashover in Classroom Size Compartment

641

Table 65.2 Heat of ignition in the room where fire broke out Room Subtotal Uncertain Cigarette Mobile electric heater Electric cable Cable with appliance Fan Aquarium heater Other electric appliances Spotlight Voltage stabilizer Welder Gas range Others

Science room 383 59 13 19 13 14 6 0 10 0 0 0 2 247

Class room 306 80 32 12 13 4 6 19 6 2 0 0 7 125

fire brigade notification from the start of a fire is 3165 s and 3000 s corresponding to the ignition source “unknown” and “cigarette,” respectively. Table 65.2 shows the number of fires for various sources of ignition. It shows that the cigarette is the main heat of ignition in all rooms.

Gymnasium 174 44 41 1 2 3 2 0 1 12 11 9 0 48

Though the cigarette is the main heat of ignition in actual school fire and it takes long time before a fire becomes large, we assumed that the starting point of real-scale fire tests was a time when the ignited materials, such as wastepaper or garbage waste, burn well. Because the greater the energy release, the quicker the fire grows [11], 4 l of methanol in a square steel pan of 0.5 m (about 120 kW in the open space and about 240 kW in the corner of noncombustible walls) was used as the fire source in fire tests to include those in the actual fires. Also the fire source was placed on the floor, in the corner of the room. Wood was used for the wall and floor finish.

65.3.3 Loaded Combustibles Fire load comprises loaded combustibles and fixed combustibles. From the survey of the loaded combustibles in schools [14], the main combustibles are the paper and books in the bookshelves and on the desks, in the room with many combustibles, such as teachers’ room. Therefore, wood cribs were used as the combustibles while adjusting it so that total surface area and total calorific value became equivalent. A piece of Japanese cedar, 0.03  0.115  0.75 m, with a weight of about 1 kg was used to form two kinds of wood cribs. Two pieces of it were provided in each layer, ten layers in total. One is the bundle type, simulating books in a line, of

Teachers’ room 81 13 22 3 6 0 1 0 0 0 0 0 0 36

Subtotal 1154 279 172 37 34 28 21 19 18 14 11 9 9 503

the size 0.23  0.75  0.30 m, and the other is the lattice type. The desks and the chairs, which had been used in school, were also used as the loaded combustibles.

65.4 65.3.2 Fire Source in Fire Test

Lounge 210 83 64 2 0 7 6 0 1 0 0 0 0 47

Fire Tests

Five types of fire tests from Test A to Test E were conducted to measure the time to onset of flashover.

65.4.1 Test A: HRR Around Fire Source Fire tests were conducted to measure HRR and the time for fire spread from the fire source to wood wall (plywood) and the loaded combustibles, called “Test A.” HRR was measured by oxygen consumption method under the hood, as shown in Fig. 65.1. For example, the measured HRR and total heat release (THR) in the case of the arrangement of the combustibles, as shown in Fig. 65.1, are shown in Fig. 65.2. After ignition of the fire source methanol, the fire spreads to the wall soon. The flame reached a ceiling about 2 min later. HRR reached a peak of 300–400 kW, decreased and increased again as shown in Fig. 65.2. This setup was used as the fire source in Tests C–E.

65.4.2 Test B: Room Corner Test Room corner test facility, conforming to ISO 9705 [8], was used to measure HRR in the early stage of fire, called “Test B.” Wood was used for the wall finish, wood or calcium silicate board was used for the ceiling finish, and a porous propane burner of 300 kW was used with no loaded combustibles. The test results are shown in Table 65.3.

642

T. Naruse et al. 2000 HRR

1600

THR

1400 1200

1400 HRR (kW)

Bundle -type crib Steel Pan 0.5mx0.5m With Methanol 4Li ers

1600

1800

1200

1000

1000

800

800

600

600

400

400

200

200

calcium silicate board 0.05m spacing

THR (MJ)

Half size of Lace -type crib Ceiling calcium silicate board 2.7m

0

0 Time (second)

0.1mx0.1m plywood to judge the ignion to floor

0.2m spacing

Fig. 65.3 HRR and THR in Test C1

Fig. 65.1 The test setup for Test A and the example of the arrangement of the combustibles

800 700 600 500

500 450

HRR THR

400 350 300 250

400 300 200 100

200 150 100 50

0

THR (MJ)

HRR (kW)

1000 900

0 Time (second)

Fig. 65.2 HRR and THR in Test A

Table 65.3 The results of Test B and C Test No. B1 B2 C1

Ceiling lining Wood Gypsum board Gypsum board

Time to 1000 kW 56 s 94 s 3570 s

Time to 1700 kW 64 s 112 s 3612 s

HRR exceeded 1000 kW after 56 s and 94 s from ignition in Test B1 and B2 respectively.

65.4.3 Test C: HRR in the Compartment A test facility of the size 7.3  8.0  2.8 m (height), with two openings, each 0.9  2.1 m (height) [12], was used to measure HRR in the early stage of fire, called “Test C.” Wood was used for the wall finish, and gypsum board of thickness 0.027 m (noncombustible [1]) was used for the

ceiling finish. Fire source and the loaded combustibles were placed in a back corner of the compartment, as shown in Fig. 65.1. Also four lattice-type cribs, piled up, were placed 1.1 m from walls to measure HRR at the time of fire spreading from the fire source to the cribs, due to a limited of the capacity of the hood. The test result is shown in Fig. 65.3 and Table 65.3. At 3420 s after ignition, fire spreads to the cribs. At 3570 s and 3612 s after ignition, HRR exceeded 1000 kW and 1700 kW, respectively.

65.4.4 Test D: Measuring Time to Flashover Fire tests, using a 7.9  8.0  2.7–3.8 m (height) room with an opening 6.3  2.0 m (height) located 0.85 m above the floor, a floor area of 63 m2 [12], were conducted to measure time to onset of flashover, as shown in Figs. 65.4, 65.5, and 65.6, called “Test D.” There are two series in Test D. The main purpose of the first series of Test D, called “Test Da,” was to investigate the effectiveness of countermeasures, such as eaves and balcony, to prevent the fire spreading upstairs by the external flame from the compartment opening. Therefore, the fire was designed to grow fast, using wood cribs placed densely around the fire source, and flashover occurred rapidly. A noncombustible curtain 0.7 m high was positioned on the opening to reduce heat loss and was removed after flashover, in the first series. The other test condition is indicated in Table 65.4. The main purpose of the second series, called “Test Db,” was to investigate the external flame similar to the first series and the spread of fire in the early stage. Therefore, a fire source as described in Sect. 3.2 and shown in Fig. 65.1 was used. While holding a sliding door open, window panes were installed in Test Db. The other test condition is indicated in Table 65.4.

65

Experimental Study of Time to Onset of Flashover in Classroom Size Compartment

壁

横貼り

TLin

∼

着火源 床

TCs Tree

8.0m

Ignition source ∼

TCin

Wood Cribs

3 piled up 2 piled up ∼

TRin

開口幅

1 piled up

6.3m 7.9m

Fig. 65.4 The plan of test facility (Test Da) and arrangement (Test D2)

∼

着火源

TLin 床

TCs Tree Ignition source

8.0m

TCin ∼

Wood Cribs 3 piled up 2 piled up ∼

TRin

開口幅

1 piled up

6.3m 7.9m

Fig. 65.5 The plan of test facility (Test Db) and arrangement (Test D 6)

∼

着火源

TLin TCs Tree 8.0m

TCin

Ignition source

∼

Wood Cribs 3 piled up

床 

2 piled up

TRin ∼ 開口幅

1 piled up

6.3m 7.9m

Fig. 65.6 The plan of test facility (Test Db) and arrangement (Test D10)

643

Simulating a teachers’ room and a classroom, the loaded combustibles were placed as shown in Figs. 65.5 and 65.6. Thermocouple trees (TCs) were installed as shown in Figs. 65.4, 65.5, and 65.6, at the heights of 0.8 m (T-4), 1.8 m (T-3), 2.8 m (T-2), and 3.0 m (or 3.8 m) (T-1), from the floor. Temperatures were recorded every 2 s. Two kinds of thermocouples were used. Type-K glass braid insulated thermocouples of 0.65 mm in diameter were used for the center TC. Type-K stainless steel sheathed thermocouples of 3.2 mm in diameter were used for the other TCs. No correction was done for the measured temperatures. The plywood of thickness 0.027 m and the gypsum boards of thickness 0.027 m or 0.0095 m (quasinoncombustible [1]) were used for interior finish. In all tests, the plywood was used for flooring. Two pieces of Japanese cedar were placed under each bundle-type cribs as the spacers. The summary of test results is as follows. In Tests D1–D4 and D11, after ignition, fire spreads rapidly and flashover occurred about 330–420 s from ignition. From these results, in the cases where the fire source grew fast in the compartment with a noncombustible interior finish and where the fire source was 240 kW in the compartment with wood ceiling finish, the flashover occurred around 330–420 s. In Test D5, the crib was placed adjacent to the fuel pan, which was placed in the corner. Methanol burned for more than 600 s. The floor, close to the fuel pan, was ignited by radiation from fire source, and the flame spread on the floor to the cribs, which were placed 1.1 m from the walls and then flashover occurred. We judged that this fire spread was not suitable for this fire test. In Tests D6–D12, temperature histories were similar to Fig. 65.3, except for D11. Even though the burning lattice-type cribs dropped off from the top of the bundle-type ones and ignited the floor and the discretely placed cribs, it did not become a direct cause of the onset of flashover. The measured temperatures of Tests D2, D6, and D10 are shown in Figs. 65.7, 65.8, and 65.9, for examples. They show the sudden steep temperature rise in the lower cool layer clearly, following the temperature rise in the upper hot layer, at the onset of flashover. Depending on the definition of flashover by ISO 13943 [2], the time to onset of flashover is listed in Table 65.5. In most cases, there is no large difference between the time to flashover defined by video record and the time when the flame spreads to the whole combustible materials from video record in the compartment. In the case of long time to flashover, the actively burning area, moved from the back corner to the center of the wall in the depth direction. The TCin shows as high temperature as those on the fire source. Judging from temperature rise, it took an average of 4 s that

644

T. Naruse et al.

Table 65.4 The test condition of Test D Ignition source Material Crib

The combustibles Loaded 700 MJ/m2

Crib

700 MJ/m2

Crib

700 MJ/m2

(Da) 4

Crib

700 MJ/m2

(Da) D5 (Db) D6 (Db) D7 (Db) D8 (Db) D9 (Db) D10 (Db) D11 (Db) D12 (Db)

Methanol 8l Methanol 4l Methanol 4l Methanol 4l Methanol 4l Methanol 4l Methanol 4l Methanol 4l

700 MJ/m2

No D1 (Da) D2 (Da) D3

1400

700 MJ/m2 700 MJ/m2 700 MJ/m2 700 MJ/m2 700 MJ/m2 700 MJ/m2

1000 800 600 400

1400

1000 800 600 400 200 0

Test D2

0 0

100

200

300 400 Time (second)

500

600

3.8 m 3.8 m

3.8 m

3.8 m 3.8 m 3.8 m 3.8 m 3.0 m 3.0 m 2.7 m 2.7 m

Test D6

0

200

Ceiling height 3.8 m

6-TLin-1 6-TLin-2 6-TLin-3 6-TLin-4 6-TCin-1 6-TCin-2 6-TCin-3 6-TCin-4 6-TRin-1 6-TRin-2 6-TRin-3 6-TRin-4

1200 Temperature (⬚C)

Temperature (⬚C)

400 MJ/m2

2-TLin-1 2-TLin-2 2-TLin-3 2-TLin-4 2-TCin-1 2-TCin-2 2-TCin-3 2-TCin-4 2-TRin-1 2-TRin-2 2-TRin-3 2-TRin-4

1200

Finish Ceiling: gypsum board Wall: gypsum board Column, beam: wood Ceiling: gypsum board Wall: gypsum board Ceiling: gypsum board Wall: plywood (1.6 m) Column, beam: wood Ceiling: gypsum board Wall: gypsum board Column, beam: wood Ceiling: gypsum board Wall: plywood Ceiling: gypsum board Wall: plywood Ceiling: plywood Wall: plywood Ceiling: gypsum board Wall: plywood Ceiling: plywood Wall: plywood Ceiling: gypsum board Wall: plywood Ceiling: plywood Wall: plywood Ceiling: gypsum board Wall: plywood

600 1200 1800 2400 3000 3600 4200 4800 5400 Time (second)

Fig. 65.8 The measured temperatures in Test D6

Fig. 65.7 The measured temperatures in Test D2

65.4.5 Test E: Full-Scale Fire Test

hot gas propagated from the location of TCin to that of TRin at the opening.

A full-scale fire test using school building [12] was conducted with the same fire source as shown in Fig. 65.1, called “Test E.” The difference is the location of fire source

Experimental Study of Time to Onset of Flashover in Classroom Size Compartment 1400

10-TLin-1 10-TLin-2 10-TLin-3 10-TLin-4 10-TCin-1 10-TCin-2 10-TCin-3 10-TCin-4 10-TRin-1 10-TRin-2 10-TRin-3 10-TRin-4

Temperature (oC)

1200 1000 800 600 400

1BY

Video

4.95m

Test D10

0 600

1200

1800

1EY

2400

3000

3600

4200

Time (second)

TCs Tree Ignition Source Wood Cribs

200

0

645

9.46m

65

16 m

Fig. 65.10 The plan of ignited room and arrangement in Test E

Fig. 65.9 The measured temperatures in Test D10

Table 65.5 Time to onset of flashover in Test D No D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D11 D12

Flame chip to ceiling height

121 135 117 138 109 118 117 124

Wall charring at ceiling height No video recording No video recording No video recording No video recording 175 165 202 193 162 169 165 197

in the corners of the wall-wall and the column-wall of wood. The plan of the compartment is shown in Fig. 65.10 and it has the floor area of 150 m2. The gypsum board of thickness 0.0095 m was used for the ceiling finish, and the plywood of thickness 0.012 m was used for the wall and the floor finish. TCs were placed as shown in Fig. 65.10, at the heights of 1.8 m (T-1), 2.4 m (T-2), 3.0 m (T-3), and 3.53 m (T-4) (under the ceiling), from the floor. Temperatures were recorded every 2 s. Type-K glass braid insulated thermocouples of 0.65 mm in diameter were used for TCs. No correction was done for the measured temperatures. While holding a sliding door open in the exterior wall, windowpanes were installed in the half of the opening of 4.95  1.87 m (height), as shown in Fig. 65.10. As the test result, at 2759 s after ignition, fire spread to three lattice-type cribs, which were piled up and placed 1.1 m away from walls. Till 2885 s after ignition, flame had not spread to the whole room yet, judging from the

Ignition of center crib

608 2153 620 1586 1997 2318 291 –

Onset of flashover Video Temperature 414 402 328 424 620 620 2806 2806 4230 4242 5520 5518 2012 2012 2664 2656 329 330 3327 3324

video record, located at 0.5 m high from the floor of the ignited room. After that, no image had been recorded in the video. The measured temperatures at the positions of 1BY and 1EY are shown in Figs. 65.11 and 65.12. At the position of 1EY close to the video camera, the temperature increased rapidly at 2774 s after ignition. There is a large difference in the time when temperatures in two locations rose, it seems that the flame had not spread to the right-hand side of the compartment yet, because the temperature at the position of 1EY stayed at about 400  C for a while, when the temperature at the position of 1BY rose rapidly.

65.5

Discussion

Figure 65.13 shows the results of our test data in the figure which Lee and Harada [9] arranged the previous data of the time to flashover in real-scale fire tests. It shows large

646

T. Naruse et al.

1400 1200

1BY-2

1000

1BY-3

Previous real-scale test data arranged by Lee and Harada [9] Test B Test D1 Test D2 Test E

1F Teachers' room

1BY-4

tFO (second)

Temperature (⬚C)

1BY-1

800 600 400 200 0

0

600

1200 1800 2400 Time (second)

3000

3600 AT (m2)

Fig. 65.11 The measured temperatures in Test E (1BY)

Fig. 65.13 Relationship among previous test results

1400 1EY-2

1000

1EY-3

1F Teachers' room

6000 Fire tests using the gypsum board for the ceiling

5000

Test B Test C Test D Test E

1EY-4 800 tFO (second)

Temperature (⬚C)

1EY-1 1200

600 400 200 0

0

600

1200 1800 2400 Time (second)

3000

3600

4000 3000

scattering among the data. The difference of test conditions influencing the fire development [11] is considered as the reason. Figure 65.14 shows the relationship between tFO and AT for the test data using gypsum board for the ceiling finish. Also, it shows Eq. 65.1 from the previous test results and Eq. 65.2 from the results of the calculation for the t-squared fires [9]. It seems strange that Eq. 65.1 shows that as the AT becomes larger, tFO becomes shorter, in spite of a different relationship indicated in Eq. 65.2. Even though the data are limited and they vary widely, from tFO of Tests D6, D8, D10, and D12, including B2, which the gypsum board is used for the ceiling finish, it is shown in Eq. 65.3, approximated by the least square method. tFO ¼ 0:1414AT 1:8742

ð65:3Þ

tFO = 0.1414AT1.8742 R² = 0.956

2000

Assuming the half room area Classroom arrangement

1000 0

Fig. 65.12 The measured temperatures in Test E (1EY)

Eq.(2) by Lee and Harada[9]

Eq.(1) by Lee and Harada[9]

0

100

200 300 AT (m2)

400

500

Fig. 65.14 Relationship between tFO and AT

The data are limited; more data are needed to discuss quantitatively. At the time of onset of flashover in a large room fire, such as classroom in school, which may occur at a certain HRR, there seems to be relatively large temperature distribution in the upper hot gas layer horizontally and vertically, because the flame enters the upper hot gas layer and changes the direction along the ceiling and spreads under the ceiling. If only a part of the upper hot layer has enough radiant emittance to ignite the nearby combustibles, it can cause an onset of flashover. And it takes time in a large room so that the hot gas layer propagates to involve the whole combustibles [15]. If this is valid, time to onset of flashover in the large room is not directly proportional to the room or internal surface area. It means that the time to flashover might be

65

Experimental Study of Time to Onset of Flashover in Classroom Size Compartment 6000

tFO(second)

2. Equation estimating time to onset of flashover in the compartments, which have a fire source of 240 kW, was proposed, but more data are needed to improve the correlation. 3. The effect of ceiling height on time to onset of flashover was investigated. In both compartments with a ceiling finish of plywood and the gypsum board, when the ceiling height reduced to 3.0 m from 3.8 m, time to flashover became less than half.

Test B1 (the wood ceiling) Test B2 (the gypsum board ceiling) Test D (the wood ceiling) Test D (the gypsum board ceiling) Test E (the gypsum board ceilng)

5000 4000 3000 2000

Classroom arrangement

1000 0

0

1

2 Ceiling height (m)

3

4

Fig. 65.15 Relationship between tFO and ceiling height

expressed by the equation, showing the convex shape in Fig. 65.14. From Figs. 65.11 and 65.12, temperatures of 1EY are relatively low and almost half of those of 1BY. Also, there was no opening in the right-hand side of ignited room, before flashover. Therefore, if heat loss in the righthand side of ignited room is ignored, AT of only left-hand side may influence the time to onset of flashover. In Fig. 65.14, this is indicated. Figure 65.15 shows the relationship between tFO and the ceiling height. In both compartments with the ceiling finishes of plywood, and the gypsum board, when the ceiling height reduces to 3.0 m from 3.8 m, tFO became less than half. In the compartments with the ceiling finish of plywood, the lower the ceiling becomes, the shorter time to flashover becomes. Because the data are limited and Fig. 65.15 shows the large scatter among the data, the clear correlation between tFO and the ceiling height has not been derived. More data are needed to discuss quantitatively.

65.6

647

Conclusion

Five kinds of fire tests were conducted to measure the time to onset of flashover in the school building. The effects of wood interior finish and loaded combustibles on the time to onset of flashover in the classroom size compartments with a floor area of 63 m2 and 150 m2 were studied experimentally. Based on the results, the following conclusions are derived. 1. In the cases that fire source grew fast in the compartment with noncombustible interior finish and that fire source was 240 kW in the compartment with wood ceiling finish, the flashover occurred in 330–420 s.

Acknowledgment The project was conducted by the cooperative body consisting of Waseda University, Akita Prefectural University, Mitsui Homes Inc., Sumitomo Forestry Inc., and Gendaikeikaku Architectural & Planning Office, under the cooperation with the National Institute of Land Infrastructure Management and Building Research Institute. The experiment was funded by the Ministry of Land, Infrastructure, Transportation and Tourism. We would like to acknowledge the great support from both cities of Gero and Nakatsugawa, Gifu-Pref. and their fire department, and volunteer fire brigades for the safe execution of the full-scale tests. The authors are also indebted to Nakajima Public Shop Inc. in the construction of the test buildings and the arrangements for the Gifu test site. We would like to thank also the cooperation of numerous individuals and organizations which helped us for the execution of the large-scale fire experiments.

References 1. The Building standard law of Japan (2011) Edited and published by The Building Center of Japan 2. ISO 13943 (2008) International Organization for Standardization 3. Babrauskas V, Peacock R, Reneke P (2003) Defining flashover for fire hazard calculations: Part II. Fire Saf J 38:613–622 4. Babrauskas V (1984) Upholstered furniture room firesmeasurements, comparison with furniture calorimeter data, and flashover predictions. J Fire Sci 2:5–19 5. McCaffrey BJ, Quintiere JG, Harkleroad MF (1981) Estimating room temperature and the likelihood of flashover using fire data correlations. Fire Technol 17(2):98–119 6. Heselden AJM, Melinek SJ (1975) The early stage of fire growth in a compartment, a co-operative research program of the CIB W 14, Fire Research Note, No.1029, Jan 7. Hasemi Y, Yoshida M, Nakabayashi T, Yasui N (1992) Transition from room corner fire to flashover in a compartment with combustible walls and noncombustible ceiling. In: Proceedings of the first Asian conference of fire science and technology, China, pp 269–274 8. ISO 9705 (1993) International Organization for Standardization 9. Lee S, Harada K (2014) A formula for determining the critical of heat release rate for flashover considering scale of the room. In: Proceedings of JAFSE annual symposium, (in Japanese) 10. Lai C, Tsai M, Lin T (2010) Experimental investigations of fire spread from movable to fixed fire loads in office fires. J Fire Sci 28:539–559

648 11. Plastic flammability handbook (2004) Principles, regulations, testing, and approval, 3rd edition, Edited by Troitzsch J, Hanser Publishers, pp 38–40 12. Hasemi Y, Yasui N, Kato K, Ida A, Itagaki N, Izumi J, Osaka T, Kaku T, Naruse T, Hagiwara I, Kagiya K, Suzuki J (2014) Fullscale fire tests of 3-storey Wooden School Building, World Conference on Timber Engineering, Canada 13. Fire report, 1995–2008, Fire and Disaster Management Agency in Japan

T. Naruse et al. 14. Hasemi Y, Yasui N, Kato K, Ida A, Itagaki N, Kaku T, Izumi J, Osaka T, Hebiishi T (2011) Feasibility study for the development of fire safety standard of large-scale timber-based buildings. In: AIJ annual meeting, Tokyo, (in Japanese), pp 95–108 15. Poon L (1998) Predicting time of flashover. In: Proceedings of the third Asia-Oceania conference of fire science and technology, Singapore, pp 283–294

Experimental Study and Model Analysis of Flashover in Confined Compartments

66

Ruowen Zong, Ruxue Kang, Weifeng Zhao, and Changfa Tao

Abstract

Catastrophe analysis of flashover has been studied based on a series of fire experiment in three different scales of compartment. The experimental data have been compared with the theoretical values predicted by a nonlinear fire growth model, which were in reasonable agreement with the theoretical bifurcation curves. Several critical conditions for flashover to occur have been also investigated, including the ceiling height, the fire radius, and the compartment scale size. The results showed that fire radius played the dominating role in the occurrence of flashover, and the characteristic value of fuel, χHc/Hvap, could be regarded as the dominating factor for the width of unstable range at the theoretical curves. And the scaling correlation between the compartment size and the critical fire radius appeared to be Rf / Lc3=4 from the tests. Keywords

Flashover  Confined compartment  Nonlinear dynamics  Scale  Experimental study

Nomenclature H_ 0 m_ a m_ f q_ f 1 q_ f 2 Q_ w a Af Aw b

Net enthalpy flow rate associated with mass flow out of the vent (W) Mass flow rate of air into the room (kg/s) Mass combustion rate of combustible gas (kg/s) Heat flux from the fire base (W/m2) Heat flux from hot smoke layer to fire base (W/m2) Energy transfer rate out through the walls and floor (W) Coefficients relating to the maximum heat flux output of the fire (W/m2) Area of fire base (m2) Area of room exposed to hot smoke layer (m2) Coefficients relating to the mean beam length of the fire (m-1)

R. Zong (*)  R. Kang  W. Zhao  C. Tao University of Science and Technology of China, No.443, Huangshan Road, Hefei, Anhui 230027, China e-mail: [email protected]

c cD cp cw D g G Hc Hi hj Hr HR ht Hv Hvap kw L Lc m N

Convective heat transfer coefficient (W/m2K) Vent flow coefficient Specific heat at constant pressure of hot smoke layer (J/kg K) Specific heat at constant pressure of wall (J/kg K) Dimensionless height of discontinuity plane Gravitational constant (m/s2) Energy gain rate of hot smoke layer (W) Heat of combustion (J/kg) Bottom thermocouple height (m) Height of thermocouples (m) Top thermocouple height (m) Height of room (m) Heat transfer coefficient (W/m2 K) Height of room vent (m) Heat of vaporization (J/kg) Thermal conductivity of wall (W/m K) Energy loss rate of hot smoke layer (W) Dimensionless compartment side length Mass of hot smoke layer (kg) Dimensionless neutral height

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_66

649

650

R. Zong et al.

R Rf Sr t T Ta Tb Tf Tw Uc Wv

Radius of fire (m) Dimensionless fire radius Stoichiometric ratio Time (s) Temperature of hot smoke layer (K) Ambient air temperature (K) Temperature of fire base (K) Temperature of flame (K) Temperature of compartment wall (K) Wall temperature factor Width of room vent (m)

Greek Symbols β ε ρ0 ρw σ χ

Ratio of the specific heat at constant pressure of combustible gas to that of hot smoke layer Hot smoke layer emissivity Density of ambient air (kg/m3) Density of wall (kg/m3) Stefan–Boltzmann constant (W/m2K4) Combustion efficiency

66.1

Introduction

Flashover is a rapidly occurred transitional event during fire growth period that a small localized fire suddenly undergoes a rapid increase in its size and intensity. It can be easily occurred in the confined compartments where ventilation is poor. This transition is characterized by rapid increase in both the temperature of upper layer (which forms under the ceiling) and the burning rate of fire itself [1]. Some theoretical models of fire growth have been developed to explain the phenomenon, suggesting a nonlinear process. The model of flashover involving with the basic thermal explosion theory was first presented by Thomas [2, 3]. Further development of this approach has been made by Bishop [4], Holborn [1], and Beard [5]. A simple two state-variable transient model of fire growth in a compartment has been constructed by them with the application of modern nonlinear dynamics. The critical conditions for flashover were given by Graham [6], Holborn [7], and Bishop [8] in a general form in terms of several nondimensional parameters. However, the considerable theoretical and experimental research for critical conditions of flashover was rare in previous studies, especially in the compartments with different scales. In this paper, the experimental validation and model analysis have been studied. In order to investigate the critical conditions for flashover, including the compartment height, fuel area, and compartment scale, the model’s prediction of nonlinear behavior has been compared with the results

obtained from experiments conducted in three different scales of model compartments. All these studies might help to get a better understanding of the critical conditions of flashover and a theoretical and experimental support for the safety of building structure.

66.2

Nonlinear Dynamics of Flashover

The nonlinear model of flashover is based on zone models in which the compartment has been divided into two uniform zones, hot upper smoke layer and lower cooler layer. The interface between two zones is parallel to the floor [2, 3]. According to the formulation of Bishop [4], the principle of energy conservation for the upper layer follows the equation as: dT G  L ¼ dt cp m

ð66:1Þ

On one hand, the energy gain rate of upper zone depends on whether the ratio of the air mass flow rate and the fuel volatilization rate is greater than (fuel-controlled fire) or less than (ventilation-controlled fire) the stoichiometric ratio and can be given by: 8 m_ a > >  Sr < χ m_ f H c if m_ f G¼ ð66:2Þ m_ a m_ a > > < Sr : χ H c if Sr m_ f The fuel volatilization rate is given by:   q_ f 1 þ q_ f 2 Af m_ f ¼ Hvap

ð66:3Þ

The calculation of heat flux from fire to fire base has been revised in Holborn’s paper [1], related with the one in Bishop’s model. An additional term considering the convective heat flux from flame has been added to the existing radiative term, as following:   q_ f 1 ¼ a½1  expðbRÞ þ c T f  T b ð66:4Þ The heat flux from smoke layer is depended on the radiative heat transfer and can be given by: q_:f 2 ¼ σ½εT 4 þ ð1  εÞT 4w  T 4a 

ð66:5Þ

The mass flow rate of air into the compartment [9] is given by:

66

Experimental Study and Model Analysis of Flashover in Confined Compartments

2 m_ a ¼ cD ρ0 W v H v 3=2 3

ffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi    Ta D 2g 1  ðN  D Þ N þ 2 T

651

respect to temperature fluctuations. In contrast the intersection at point B is unstable [4].

ð66:6Þ

66.3 Bishop [4] introduced a dimensionless wall temperature parameter, Uc, which controlled the fraction of hot smoke layer temperature that the wall reaches. Uc ¼

Tw  Ta T  Ta

U c 2 ½0; 1

ð66:7Þ

One the other hand, the energy loss rate from hot smoke layer can be assumed to be composed of two terms: L ¼ H_ o þ Q_ w

ð66:8Þ

where H_ 0 , the net enthalpy flow rate, is associated with mass flow out of the vent and is given by:   H_ o ¼ m_ a þ m_ f cp ðT  T a Þ ð66:9Þ and Q_ w , the energy transfer rate out through the walls and floor, is given by:     Q_ w ¼ Aw εσ T 4  T 4w þ ht ðT  T w Þ

ð66:10Þ

Two curves could be created by Bishop after the equations of the gain and loss rates of hot smoke layer have been reached, as shown in Fig. 66.1 where the three different loss curves L1, L2, and L3 represent different room sizes, ventilation conditions, or types of wall material/thickness separately. The intersection at point A1 or A2 corresponds to a large controlled fire, while point C1 or C2 represents a small localized fire. Both states are stable with

Experimental Descriptions

Three different scales of model compartment have been constructed of 6 mm thick iron shell with three sides infilled with 8.5 mm thick calcium silicate board and lined with 3 mm thick plywood inside. For the convenient observation of fire growth, the front side of these compartments was made of fireproof glass. The length of internal side of these different sizes of cubic cell (Cell A, Cell B, and Cell C) was 0.3 m, 0.6 m, and 1.2 m, respectively. There were two openings in each compartment, one as a window and the other as a door. Details of the Cell C are shown in Fig. 66.2. There were 15 thermocouples (chromel/alumel) located under the ceiling of each compartment to monitor the temperature of hot gas layer. They were arranged in three layers equably, and the location of three layers was one to eighth, quarter, and half of the height of the compartment (HR/8, HR/4, HR/2) under the ceiling, respectively. The position of thermocouples in each layer is shown in Fig. 66.3. Regular gasoline (SINOPEC oil, 93#) has been burned in the square trays made of iron which were centrally positioned on the platform attached to an electronic balance located below the compartment, as shown in Fig. 66.2. The output of the

Thermocouples inlet

0.4m 0.4m 0.2m

Glass Wall 0.4m

0.2m

0.4m×0.4m

0.2m L1

L2

Gain and Loss Rates

A1 B

A2

L3 G

1.2m

Fire tray 0.6m

1.2m 0.4m

0.4m 1.2m

C1 C2

D

Ta

Temperature of upper layer

Electronic balance

0 T

Fig. 66.1 Qualitative form of gain and loss rates curve for fixed time

Fig. 66.2 Sketch map of Cell C

652

R. Zong et al.

thermocouples has been calibrated in a data transmission module, and the analogue signal was transmitted to the computer through serial interface, along with the balance signal. The signal acquisition system was programmed with Visual Basic 6.0 independently, and the data was saved in Excel files. In order to identify the critical condition for flashover in different sizes of fuel tray and different scales of compartment, the sizes of tray ranged from 60  60 mm to 250  250 mm in different scales of compartments. The gasoline was sufficient for the flashover to occur. The height of ceiling in Cell B could be shifted from 40 to 80 cm above the floor to study the influence of ceiling height on flashover. The fire burned out, lasting for 10–20 min. Burning rates and gas temperatures have been measured per second right

0.2m Thermocouple

after the ignition. Then, the onset of flashover was clearly distinguished by vigorous external flaming.

66.4

Results

All the experimental parameters and steady-state results have been shown in Table 66.1. Three factors that may influence the fire growth have been considered in these experiments, including the compartment size, the tray area, and the ceiling height (which could only be changed in Cell B). The results on temperature, as the average temperature of gas layer, were calculated by the following method. The average value of results measured by the five thermocouples in each layer (T1, T2, and T3) was calculated, and data reduction methods for averaging temperatures based on equation of state [10] have been used to figure out the average gas layer temperature: Hr  Hi T avu ¼ Z Hr 1 dy T Hi

0.4m

0.4m

0.2m

ð66:11Þ

Then, the denominator of Eq. 66.11 was presented by: Z

Hr Hi

Fig. 66.3 Top view map of thermocouples position in each layer of Cell C

  L X hjþ1  hj T jþ1 1 dy ¼ ln T T jþ1  T j Tj j¼1

ð66:12Þ

As for the calculation of fuel mass loss rates, the measured values of electronic balance were first differentiated, and then Savitzky–Golay method has been

Table 66.1 Steady state results from scale compartment fire experiments Experiment number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Compartment size Cell A Cell A Cell A Cell B Cell B Cell B Cell B Cell B Cell B Cell B Cell B Cell B Cell C Cell C Cell C

Ceiling height (cm) 30 30 30 60 60 60 60 60 80 70 50 40 120 120 120

Tray area (cm2) 66 7.5  7.5 99 10  10 11  11 12  12 13  13 15  15 15  15 15  15 15  15 15  15 17  17 20  20 25  25

Maximum gas temperature (K) 761.15 813.15 844.15 594.15 579.15 886.15 896.15 965.15 847.15 873.15 967.15 1031.15 532.15 896.15 926.15

Maximum fuel mass loss rate (g/s) 0.25 0.44 0.62 0.55 0.52 1.15 1.68 2.97 1.95 1.61 2.77 3.02 3.55 5.94 11.06

Flashover occurrence Negative Positive Positive Negative Negative Positive Positive Positive Positive Positive Positive Positive Negative Positive Positive

Experimental Study and Model Analysis of Flashover in Confined Compartments

Upper layer temperature (ć)

600

EXP 2

Upper layer temperature Fuel mass loss rate

500

653

1.0

0.8

400 0.6 300 200

0.4

100

0.2

Fuel mass loss rate (g/s)

66

0 0.0 0

200

400

600

800

Time (s) 700

4

400

3

300

2

200 1 100

)XHOPDVVORVVUDWH

4

8SSHUOD\HUWHPSHUDWXUH

500

3

400 300

2

200

Fuel mass loss rate (g/s)

)XHOPDVVORVVUDWH 8SSHUOD\HUWHPSHUDWXUH

500

EXP 4 600 Upper layer temperature (ć)

600

5

5

(;3

)XHOPDVVORVVUDWHJV

8SSHUOD\HUWHPSHUDWXUHć

700

1 100

0

0 0

200

400

600

800

1000

0

0 0

7LPHV

200

400

600

800

1000

7LPHV EXP 14

Upper layer temperature (ć)

14

Upper layer temperature Fuel mass loss rate

600

12

500 10 400 8 300 6 200 4

Fuel mass loss rate (g/s)

700

100 2 0 0

50

100

150

200

250

300

0 350

Time (s)

Fig. 66.4 Layer temperature and fuel mass loss rate time histories (the scale for the mass loss rate is not common)

used to smooth the differentiated results with ORIGIN 8.0 where polynomial order was two and points of window was ten. The result illustrated the direct-viewing change of the fuel mass loss rates, and most information of data trend has been saved. Four typical histories of average gas layer temperature and fuel mass loss rate are illustrated in Fig. 66.4, including experiment #2 (7.5 cm square tray in Cell A with the positive flashover occurrence), #6 (12 cm square in Cell B with the positive flashover occurrence), #14 (20 cm square trays in Cell C with the positive flashover occurrence), and #4 (10 cm square tray Cell B with the negative flashover occurrence). The results showed that there was an initial rise in temperature and mass loss rate due to the rapid burning of gasoline. In experiment #4, the fire burned at a relatively low

and steady speed, while the results of other three experiments were far more dramatic. After ignition, the development of upper layer temperature showed the trend of slow growth at first and then fluctuation, which indicated that the temperature fluctuated near the intersection of G and L curves until a large and rapid change in both temperature and fuel mass flow loss rate occurred. Whether flashover took place or not, fluctuation could be found in all experiments to some extent. It was also interesting to note that although the bigger compartment contained larger tray with more fuel than the smaller one, the fuel supply burned out more rapidly. In experiment #14, the mass loss rate was higher. At the growth phase, the fuel burned more quickly due to its larger fuel area for reaction. After this phase, the higher mass loss rate might be resulted from the stronger

654

R. Zong et al.

thermal radiation. The maximum burning rates were 78.2 g/ m2s in Cell A, 79.86 g/m2s in Cell B, and 148.5 g/m2s in Cell C, respectively, where the difference might be resulted from the high-level radiant heat developed inside the compartment during the period of post-flashover.

66.5

those used in previous simulations [1, 4]. All the parameters adopted for the simulations are shown in Table 66.2. The wall temperature factor, Uc, cannot be 1, but it did have the maximum value depending on the thermophysical property of compartment wall. The maximum value was given by [15]:

Discussions

66.5.1 The Effect of Ceiling Height Table 66.2 Specified parameters used in model simulations

In experiments #9, #10, #8, #11, and #12 where the ceiling height changed from 80 to 40 cm, the influence of ceiling height on flashover has been detected. Figure 66.5 shows the effect of ceiling height on the maximum fuel mass loss rate and the maximum gas layer temperature in Cell B. It was obvious that there was linear relationship between the ceiling height and the maximum upper layer temperature. The correlation coefficient was up to 0.9396. Meanwhile, the linear relationship between the ceiling height and the maximum fuel mass loss rate was relatively poor, and the correlation coefficient was 0.6591.

66.5.2 Model Analysis As the fire growth simulation model described by Bishop [1, 4], fire radius (R) and wall temperature factor (Uc) controlled the upper layer temperature. With the application of model introduced above, the comparison of model analysis and experiment data has been presented here. In this model analysis, the fire area (radius) has been considered as a parameter, and other parameter values have been appointed to match the experiment situation as much as possible, as

Parameter Description Compartment parameters HR ¼ 0.3, 0.6, 1.2 m Height of scale compartment WR ¼ 0.3, 0.6, 1.2 m Width of scale compartment LR ¼ 0.3, 0.6, 1.2 m Length of scale compartment Hv ¼ 0.15, 0.3, Height of vent 0.6 m Wv ¼ 0.16, 0.33, Width of vent 0.66 m Fuel parameters Sr ¼ 14.7 Stoichiometric ratio Hvap ¼ 338,900 J/ Heat of vaporization kg Hc ¼ 43,700,000 J/ Heat of combustion kg Tf ¼ 1200 K Flame temperature Tb ¼ 600 K Af¼πR2 Fluid parameters cD ¼ 0.7

Fire base temperature Fire area

ρ0 ¼ 1.25 kg/m3

Ambient air density

Ta ¼ 280 K

Max Layer Temperature (ć)

10

8

700 650 TMax= 940.8-4.62 × HR

600

6

r2=0.9396

4

550 500 450

2

MMax= 4.444-0.033 × HR r2=0.6591

Max Fuel Mass Loss Rate (g/s)

Max Layer Temperature Max Fuel Mass Loss Rate

750

0

400 40

50

60

70

80

Ceiling Height (cm)

Fig. 66.5 Linear fitting curve of maximum layer temperature and maximum fuel mass loss rate vs. ceiling height

Vent flow coefficient

Ambient room temperature cp ¼ 1003.2 J/kg K Specific heat at constant pressure of air Heat transfer parameters ht ¼ 7 W/m2 K Heat transfer coefficient σ ¼ 5.66  1108 Stefan–Boltzmann W/m2K4 constant Uc ¼ 0.496 Wall temperature fraction χ ¼ 0.65 Combustion efficiency ε ¼ 0.41 Layer emissivity a ¼ 102,000 W/m2 Heat flux from flames to fire base parameters b ¼ 1.12 m1 c ¼ 25 W/m2K Convective heat transfer coefficient

Comments Cells A, B, and C Cells A, B, and C Cells A, B, and C Cells A, B, and C Cells A, B, and C

From SFPE [11] From SFPE [11] From SFPE [11] Estimate of typical value Estimate [1]

From Quintiere et al. [12] From Quintiere et al. [12] From measurement From Takeda [13]

From Takeda [13]

From Eq. 66.13 Estimate of typical value Estimate [1] From Emmons [14] From Emmons [14] From Holborn [1]

66

Experimental Study and Model Analysis of Flashover in Confined Compartments

Fig. 66.6 Theoretical simulation and experimental comparison of fire radius vs. max layer temperature (a) Cell A, (b) Cell B, (c) Cell C

Max Layer Temperature (K)

a

1100 1050

Theoretical Value Experimental Value

1000 950 900 850 800 750 700 0.030

b Max Layer Temperature (K)

655

1100

0.035

0.040 0.045 Fire Radius (m)

0.050

0.055

Theoretical Value Experimental Value

1000 900 800 700 600 500 0.05

0.06

0.07 Fire radius (m)

0.08

0.09

c Max Layer Temperature (K)

1200

Theoretical Value Experimental Value

1100 1000 900 800 700 600 500 0.10

B1

U c ðMaxÞ ¼ eβðkw ρw cw Þ

ð66:13Þ

where β ¼ 0.593 and B1 ¼ 0.338, as the constant. The primary material of the wall in these experiments was plywood. The thermophysical property was listed as follow: kw ¼ 0.0011 kW/mK, cw ¼ 0.96 kJ/kgK, and ρw ¼ 2050 kg/

0.12

0.14 Fire radius (m)

0.16

0.18

m3; thereby, the maximum wall temperature factor was Uc(Max) ¼ 0.496. When Uc increased to Uc(Max) gradually with the development of fire, the upper layer temperature reached the climax theoretically. With the substitution of Uc(Max) into energy conservation Eq. 66.14, the peak upper layer temperature with varying fuel area was obtained.

656

R. Zong et al.

GL¼0

ð66:14Þ

Figure 66.6 shows that how the change of fuel area affected the theoretical simulation and the experimental comparison of the peak upper layer temperature, where the broken line as an unstable solution branch. And in Fig. 66.6b, the theoretical path follows the curve of an “S” shape [1] with a lower, fuel-controlled branch and an upper, ventilation-controlled branch linked with an unstable middle branch. It was obvious that the critical radius could be predicted to be around 0.07 m. If the fire radius surpassed the point, the flashover jump would be occurred. The test results were in reasonable agreement with the theoretical curve. Figure 66.6c shows that the match is just acceptable, where the critical fire radius can be predicted about 0.15 m while it was around 0.11 m in real tests. However, there was no good agreement in Cell A, as shown in Fig. 66.6a, where the theoretical curve is much higher than the test data. To some extent, the overestimate of upper layer temperature can be explained by a simple assumption made in the model that the hot gas layer filled the entire compartment. In fact, the layer was observed to fill more than half of the compartment in most of the tests. And this assumption might be available when the compartment scale became smaller. According to the discussion above, it is found that in these experiments, when the side length of the compartment increased twice, the relevant critical fire radius for flashover just rose to 1.68 times (approximately 23/4). So the scale could be deduced as the following [16]: Rf / L3=4 c

ð66:15Þ

More specific model study will be needed to validate this assumption in the future. Another issue should be considered that the unstable range of these “S” sharp curves is too narrow (as illustrated in Fig. 66.6), compared with the results of Holborn’s experiments [1]. The reason can be explained by the different fuels used in the tests. Polyethylene pellets were used in Holborn’s experiments, whereas gasoline was used here. In order to solve this problem, the nonlinear dynamic model should be studied. When the diameter of the tray increases to a certain range, there will be three intersections between G and L, C2, B, and A1 (as illustrated in Fig. 66.1), and then B will jump rapidly to A1. In this way, two solutions (C2 and A1) could be made in Eq. 66.14. When the diameter of the tray increases continuously, there will be only one intersection, A2, which represents that the Eq. 66.14 only has one solution. So there is a range within the two solutions. When the size of tray is included in this range, the flashover in the

compartment might be occurred. Fuel-controlled heat gain can be transformed as below: G ¼ χ m_ f H c ¼ χH c

ðq_ f 1 þ q_ f 2 ÞAf χH c ¼ gðT, RÞ ð66:16Þ Hvap H vap

The value χHc/Hvap is the dominating factor for the shape of fuel-controlled curve G. Because of the high combustion heat and the low vaporization heat of gasoline, this value was much higher than that of polyethylene. So curve G was sharper which made the unique solution in Eq. 66.14 during the fuel-controlled phase, but for polyethylene, this value was lower, which produced a flatter curve G. So when curve L moved rightward, two intersections between fuelcontrolled G and L emerged, where one stands for non-flashover and the other indicated an unstable solution which might jump to flashover. For the confined compartment fire with different fuels, the expression has been verified with the experimental results [17].

66.6

Conclusion

Fire growth data in the compartment of three different scales has been obtained and compared with the theoretical prediction from the nonlinear dynamics model. The theoretical prediction of the steady maximum layer temperature with the change of fire radius appeared to be consistent with the test results, except for the smallest cell in the study. With the study of flashover occurrence with the varying ceiling height, it was found that there was linear relationship between the ceiling height and the steady maximum layer temperature as well as the maximum mass loss rate of fuel. The study of the critical fire radius for flashover occurrence in the compartment of different scales indicated that the scaling correlation between the compartment dimension and the critical fire radius could be speculated to be Rf / and the value of χHc/Hvap has been proved to be the L3=4 c dominating factor for the shape of fuel-controlled curve G and then controlled the size of unstable (or critical) range.

References 1. Holborn PG, Bishop SR (1993) Experimental and theoretical models of flashover. Fire Saf J 21:257–266 2. Thomas PH, Bullen ML, Quintiere JG et al (1980) Flashover and instabilities in fire behavior. Combust Flame 38:159–171 3. Thomas PH (1980) Fires and flashover in rooms – a simplified theory. Fire Saf J 3:67–76 4. Bishop SR, Holborn PG (1993) Nonlinear dynamics of flashover in compartment fires. Fire Saf J 21:11–45

66

Experimental Study and Model Analysis of Flashover in Confined Compartments

5. Beard AN, Drysdale DD, Holborn PG, Bishop SRA (1992) Non-linear model of flashover. Fire Sci Technol 12:11–27 6. Graham TL, Makhviladze GM, Roberts JP (1995) On the theory of flashover development. Fire Saf J 25:229–259 7. Holborn PG, Bishop SR, Drysdale DD, Beard AN (1992) The effect of variation of ventilation and other control parameters in compartment fire model. Int Commun Heat Mass Transfer 19:711–719 8. Bishop SR, Holborn PG, Drysdale DD (1995) Experimental comparison with a compartment fire model. Int Commun Heat Mass Transfer 22:235–240 9. Rockett JA (1976) Fire induced gas flow in an enclosure. Combust Sci Technol 12:165–175 10. He Y (1997) On experimental data reduction for zone model validation. Fire Sci 15:144–161 11. DiNenno PJ, Drysdale D, Beyler CL, Walton WD, Custer RLP, Hall JR, Jr, Watts JM, Jr (eds) (2002) SFPE handbook of fire protection engineering [M]. 3rd edn. National Fire Protection Association, Quincy, Massachusetts, USA. pp 3-134, A-36, 3-26 12. Quintiere JG, McCaffrey BJ, Den Braven K (1978) Experimental and theoretical analysis of quasi-steady small-scale enclosure fires.

657

In Proceedings of the seventeenth symposium (international) on combustion, pp. 1125–1137 13. Takeda H (1987) Transient models of early stages of fire growth. In: Mehaffey JR (ed) Mathematical modelling of fires, ASTM STP 983. American Society for Testing and Materials, Philadelphia, pp 21–34 14. Emmons HW (1978) Prediction of fires in buildings. In: Proceedings of the seventeenth symposium (international) on combustion, pp 1101–1112 15. Beard AN, Drysdale DD (1995) A non-linear model of major fire spread in a tunnel. Fire Saf J 24:333–357 16. Ruowen zong (2008) Fire reconstruction research of flashover in the special confined compartment [D]. University of Science and Technology of China, Hefei, Anhui, China. pp 73–75 17. Weifeng Zhao, Ruowen Zong, Bin Yao, Jiaxin Gao, Guangxuan Liao (2013) Analysis of influencing factors on flashover in the longnarrow confined space [J]. Proceeding of the 9th Asia-Oceania symposium on fire science and technology, 62(2013)72, 250–257

Part XVII Material Flammability

Experimental Study on Radiation Blockage of Small-Scale Vertical PMMA Fires

67

Zhen Li, Naian Liu, Shaojie Zhang, Xiaodong Xie, and Wei Gao

Abstract

In this work, experiments were conducted to evaluate the effect of exterior radiation on vertical PMMA fires. The mass loss, flame height, and radiant heat flux were measured. A critical heat flux of 10 kW/m2 was determined as a boundary for radiation blockage effect. For lower exterior radiation heat fluxes (10 kW/m2). Recall in Eq. 67.2, mass loss rate was a function of exterior radiant flux and 00 00 flame heat feedback ( q_ f r , q_ f conv ). Therefore, the results shown in Fig. 67.3 indicate that flame heat feedback was dominant (relative to the exterior radiation) in inducing mass loss when the exterior radiation was below (>10 kW/m2). Under higher external radiant heat fluxes, Fig. 67.3 shows that the mass loss rate had a nearly linear relation with the exterior radiation. The critical heat flux value of 10 kW/m2 is also reflected by the data of flame height. As shown in Fig. 67.4, the flame height was nearly 0.20 m for free burning case and then increased rapidly with increasing exterior radiation heat flux up to 10 kW/m2. For higher exterior radiation heat fluxes (>10 kW/m2), the flame height remained nearly constant.

00

The total radiation heat feedback q_ total can be measured by radiation heat flux meter.

67.3.2 Mass Loss Rate and Flame Height In tests, it was difficult to determine the steady-state burning stage only by the data of mass loss rate, since in combustion, there was plastic deformation on the PMMA surface and also flowing fire would form when the exterior radiation heat flux was over 30 kW/m2. In this work, the steady-state burning stage was determined by combined analysis of the video images and the mass loss data. As shown in Fig. 67.3, the variation of sample mass flux with increasing exterior heat radiation undergoes two distinct stages. For lower radiation (1.2. Equation 84.6 also shows that the gas phase properties will influence the flame spread rate. For the thermal conpffiffiffiffiffiffiffiffiffiffiffiffi ductivity of an ideal gas, we have kg / cv, g =σ 2g  T=Mg , where cv,g, σ g T, and Mg are constant volume heat capacity, molecular radius, temperature, and molar mass, respectively, with the subscript g denoting gas phase [21]. The constant volume heat capacity could be calculated using cv, g ¼ cpg – nR, where cp,g, n, and R are constant pressure heat capacity, amount of substance, and gas constant, respectively. Meanwhile, the constant pressure heat capacity is a function of temperature with the form of cp ¼ f ðT Þ ¼ a0 þ a1 T þ a2 T 2 þ   . Thus, the thermal conductivity and constant pressure heat capacity of an ideal gas are independent of pressure normal range. However, the density of the gas is proportional to pressure:
 ρ ¼ P= Rg T / P. In Eq. 84.5, it is implied that the induced flow is influenced by pressure. However, the mass loss rate increases with pressure. This will offset the pressure’s influence on induced flow speed to some degrees. Moreover, the induced flow is a buoyancy flow which is passively induced and strengthened by the flame. This indicates that when the induced flow is in a counter effect with that of pressure on flame spread, the former would not overwhelm the latter. Thus, the flame spread rate, flame height, and mass loss rate increase with pressure.

84.4

Conclusions

In this work, a series of experiments was conducted to evaluate the sidewall and pressure effects on downward flame spread over insulation material. It is found that: 1. Flame spread rate and mass loss rate increased with the increasing of sidewall geometrical factor α. Trends were observed that both parameters decelerated as α increased.

829

2. The maximum increase for flame spread rate and mass loss rate between experiments without and with sidewalls was approximately 20 % and 40 %, respectively. 3. Flame height increased with α. However, no deceleration was observed. 4. Flame spread rate, flame height, and mass loss rate were in positive relationships with pressure. A theoretical analysis was performed and an air entrainment analysis is provided. The theoretical analysis showed that: 5. Flame spread rate is proportional to induced flow speed which increases with sidewall geometrical factor α and mass loss rate and decreases with back wall width. 6. Induced flow speed increases while decelerates with the increasing of α. When α >1, it could be regarded that the induced flow speed enters a steady stage, indicating that the flame spread will approach a limit after α >1, which agrees well with experimental results. However, due to the limitation of the low-pressure facility, the induced flow speed could not be obtained. Further work is needed for the better understanding of sidewall and pressure effect. Acknowledgment This work is supported by National Basic Research Program of China (973 Program, Grant. No.2012CB719702), the Research Fund for the Doctoral Program of Higher Education (No. 20113402110023), and Key Technologies R&D Program of China during the 12th Five-Year Plan Period (No. 2013BAJ01B05).

References 1. Atreya A, Baum HR (2002) A model of opposed-flow flame spread over charring materials. Proc Combust Inst 29:227–236 2. De Ris JN (1969) Spread of a laminar diffusion flame. Symp (Int) Combust 12:241–252 3. Kudo Y, Ito A (2002) Propagation and extinction mechanisms of opposed-flow flame spread over PMMA. In: Twenty-ninth international symposium on combustion Hokkaido University Sapporo Japan. July 21, 2002–July 25, 2002. Sapporo, Japan, pp 237–243 4. Ya-Ting T, T’Ien JS (2011) A comparison of flame spread characteristics over solids in concurrent flow using two different pyrolysis models. J Combust, pp 250391 (9 pp.)-250391 (9 pp.) 5. Zhang Y, Ji J, Huang X, Sun J (2011) Effects of sample width on flame spread over horizontal charring solid surfaces on a plateau. Chin Sci Bull 56:919–924 6. Gong J, Yang L, Zhou X, Deng Z, Lei G, Wang W (2012) Effects of low atmospheric pressure on combustion characteristics of polyethylene and polymethyl methacrylate. J Fire Sci 30:224–239 7. Huang XJ, Wang QS, Zhang Y, Yin Y, Sun JH (2012) Thickness effect on flame spread characteristics of expanded polystyrene in different environments. J Thermoplast Compos Mater 25:427–438

830 8. An W, Huang X, Wang Q, Zhang Y, Sun J, Liew KM et al (2013) Effects of sample width and inclined angle on flame spread across expanded polystyrene surface in plateau and plain environments. J Thermoplast Compos Mater 2:2013 9. Gollner MJ, Huang X, Cobian J, Rangwala AS, Williams FA (2013) Experimental study of upward flame spread of an inclined fuel surface. Proc Combust Inst 34:2531–2538 10. Jiang L, Xiao H, Zhou Y, An W, Yan W, He J, Sun J (2014) Theoretical and experimental study of width effects on horizontal flame spread over extruded and expanded polystyrene foam surfaces. Journal of Fire Sciences 32(3):193–209 11. Zhou Y, Xiao HH, Sun JH, Zhang XN, Yan WG, Huang XJ (2015) Experimental study of horizontal flame spread over rigid polyurethane foam on a plateau: effects of sample width and ambient pressure. Fire and Mater 39(2):127–138 12. Tsai K-C (2007) Upward flame spread on a flat surface, in a corner and between two parallel surfaces. J Chin Soc Mech Eng 28:341–348 13. Tsai K-C (2009) Width effect on upward flame spread. Fire Saf J 44:962–967

W. Yan et al. 14. Tsai K-C (2011) Influence of sidewalls on width effects of upward flame spread. Fire Saf J 46:294–304 15. An W, Wang Z, Xiao H, Sun J, Liew KM (2014) Thermal and fire risk analysis of typical insulation material in a high elevation area: influence of sidewalls, dimension and pressure. Energy Convers Manag 88:516–524, 12 16. Quintiere JG (2006) Fundamentals of fire phenomena. Wiley, Chichester 17. Heskestad G (1983) Luminous heights of turbulent diffusion flames. Fire Saf J 5:103–108 18. Williams FA (1985) Combustion theory: the fundamental theory of chemically reacting flow systems. Perseus Books Group, New York 19. Bhattacharjee S, West J, Altenkirch RA (1996) Determination of the spread rate in opposed-flow flame spread over thick solid fuels in the thermal regime. Symp Combust 26:1477–1485 20. Cussler EL (2009) Diffusion: mass transfer in fluid systems. Cambridge University Press, Cambridge 21. Gregory Nellis SK (2008) Thermal conductivity of a gas. In: (ed) Heat transfer. Cambridge University Press

Heat Loss and Kinetic Effects on Extinction and Critical Self-Sustained Propagation of Forced Forward Smoldering

85

Jiuling Yang, Haixiang Chen, and Naian Liu

Abstract

This work studied the forced forward smoldering mechanism of foam by a transient one-dimensional numerical model. A three-step reaction scheme, including foam oxidation and pyrolysis, char oxidation was adopted. Based on the first principles of mass, momentum, and energy, the mass fractions of solid and gas species as well as temperature evolutions in self-propagation regime were examined. Foam oxidation and pyrolysis fronts were found to be ahead of char oxidation front where oxygen is not completely consumed. The critical kinetic parameters for self-sustained smoldering were determined. It was found that pyrolysis endothermic reaction, heat losses, and limited kinetics of char oxidation are favorable for smoldering extinction. Especially, smoldering extinction is more sensitive to kinetic parameters of char oxidation reaction than that of other two reactions. The halfway quenching of smoldering is mainly due to the weakening of char oxidation reaction, which cannot offset the pyrolysis endothermicity and heat loss to environment. Keywords

Smoldering
Extinction
Self-propagation
Kinetic effects
Heat losses

85.1

Introduction

Smoldering is a kind of heterogeneous combustion reaction taking place at the phase interface between a porous fuel and oxygen [1]. Smoldering can self-sustain in oxygen-starved environment for a very long time, and once it transits to flaming combustion, human life and property security would be threatened. So it is imperative to study the factors that influencing smoldering extinction and self-sustained propagation. Besides the experimental investigations, the numerical model is essential for a better understanding of smoldering mechanisms. Based on the smoldering experiments

J. Yang
H. Chen (*)
N. Liu State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei, Anhui 230026, People’s Republic of China e-mail: [email protected]

conducted in the space shuttle flights [2–5], a transient one-dimensional model in microgravity with a three-step reaction was put forward [6], in which the smoldering velocity was predicted to be linear with the airflow once selfsustaining propagation was supported. Afterward, Rein et al. expanded this model to five-step reaction [7] and successfully captured every reaction structure in both forward and opposed smoldering combustion with the same kinetic parameters extracted from thermogravimetric data by genetic algorithms [8]. Dodd et al. also examined the two-dimensional cone structure of the smoldering front with a seven-step reaction mechanism [9]. These studies promote the development of numerical smoldering propagation models. Most of numerical studies on smoldering focus on the propagation, rather than extinction of smoldering. Two extinction mechanisms including blowoff and quenching extinction are reported in literature. The former extinction is caused by the strong convective cooling effect due to

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_85

831

832

J. Yang et al.

overblowing. The latter extinction means the halfway quenching of the smoldering front after turning off the igniter, so the combustion cannot propagate through all the fuel due to insufficient heat release. Few literatures involve the extinction mechanism, especially for the effects of heat loss and char oxidation reaction on smoldering quenching extinction. For horizontal reverse smoldering that the inlet air flows in the opposed direction as the reaction front propagation, Schult et al. [10] employed a large activation energy asymptotic method (AEA) with one-step reaction to study smoldering blowoff extinction, which occurred as the incoming mass flux was further increased. Leach et al. [11, 12] adopted a numerical model with two-step reaction to predict the extinction regime, which indicated that overblowing gas flow would lead to blowoff extinction. These numerical results [10–12] agreed with the earlier experimental observations [13]. For forward smoldering that the external air flows in the same direction as the smoldering front propagation, Torero et al. [14] observed that an unexpected extinction stage (0.3–1.2 mm/s) of the downward forward smoldering was caused by buoyancy effects in normal gravity, but no extinction was found in upward smoldering due to the same direction between buoyancy and smoldering propagation. With the absence of buoyancy in microgravity, experiments [2–5] indicated that a lower airflow velocity would lead to the extinction of forward smoldering due to insignificant oxygen mass flux. The above literature shows that the mechanism of smoldering extinction is complex and varies with smoldering patterns and environmental conditions. The objective of present work is to study the factors closely related to smoldering extinction, such as inlet gas speed, heat loss, and kinetic parameters, by a numerical model.

85.2

One-Dimensional Numerical Model

the pressure-driven flow. The heat loss to the environment is considered as a heat sink item in the solid energy conservation equation [6, 7]. Three gas species, including oxygen, nitrogen, and gaseous products (all the species have equal diffusion coefficients), and three solid species, containing virgin foam, char, and ash, are involved in this model.

85.2.2 Kinetic Mechanism A three-step reaction scheme given earliest by Roger and Ohlemiller [18] is as follows: ðaÞ½Foam þ μ1o ½O2  ! μ1c ½Char þ μ1gp ½Gas ðbÞ½Foam ! μ2c ½Char þ μ2gp ½Gas ðcÞ½Char þ μ3o ½O2  ! μash ½Ash þ μ3gp ½Gas The reaction rates are described by Arrhenius-type equations with one order exponent: ωo ¼ Ao y1 yo2 eE1 =RT ωp ¼ Ap y1 eE2 =RT ωa ¼ Aa y2 yo2 eE3 =RT

85.2.3 Governing Equations Essential conservation equations are presented as follows: ∂y1 ¼ ωo  ωp ∂t

ð85:1Þ

∂y2 ¼ μ1c ωo þ μ2c ωp  ωa ∂t

ð85:2Þ

∂y3 ¼ μash ωa ∂t

ð85:3Þ

85.2.1 Model Assumptions Polyurethane foam is chosen because of easy availability of its kinetic parameters from literatures. And after smoldering propagating through all the foam, the foam can still maintain its structural integrity, [4] and thus no volume shrinkage due to reactions is assumed. Additionally, one-dimensional smoldering structure is assumed from experimental observations [3, 14–16]. Due to large surface-to-volume ratio of solid porous matrix [7] and small temperature difference between solid and gas media [17], thermal equilibrium between solid and gas is assumed. Because the gas flowing velocity through the porous fuel is so low, Darcy’s law is available to solve


 ∂ ερg ∂t

þ


 ∂ ρg u ∂x

¼ mg

!1 3 X pMmix , Mmix ¼ ρg ¼ ci =ρg , RT i   mg ¼ ρ0 ð1  μ1c Þωo þ ð1  μ2c Þωp þ ð1  μash Þωa ð85:4Þ u¼

K ∂p μ ∂x

ð85:5Þ

Heat Loss and Kinetic Effects on Extinction and Critical Self-Sustained. . .



ð85:6Þ

85.3

   ∂T ∂T ∂T ∂ keq þQ þ ρg Cg u ¼ ρCp eq ∂x ∂t ∂x ∂x

 HðT  T 0 Þ, H  U e AL ρCp eq ¼ ð1  εÞρs Cs þ ερg Cg keq ¼ V ¼ εk g þ ð1  εÞks Q  ¼ ρ0 Δhf , o ωo þ Δhf , p ωp þ ρc Δhc, o ωa

ð85:7Þ

Non-extinction refers to that the smoldering front can maintain a self-sustained propagation after switching off the heater. To understand the smoldering characteristics of non-extinction regime, seven probes are inserted through the whole domain at 1.5, 3.5, 5.5, 7.5, 9.5, 11.5, and 13.5 cm from the heater. The simulated spatial profiles of heat release rates and mass fractions of solid species at these probes are explored in Fig. 85.1. Figure 85.1a shows that the foam oxidation and pyrolysis front are ahead of char oxidation front, which is consistent with the experimental observations [14]. The heat magnitudes of the foam pyrolysis and oxidation are comparable, but that of char oxidation is four times higher than that of foam oxidation, which corroborates that char oxidation reaction is the major heat source to support self-sustaining smoldering.

a ð85:8Þ

  ∂T ¼ q t tign k ∂x Here, convective cool boundary condition is imposed at the left side of the foam sample that differs from a zero temperature gradient [12] or a constant external temperature [19]. At the right boundary, a constant heat flux q for a fixed period of time is imposed, and air flows through the porous fuel from the same side. The fuel sample is 14 cm long. Initial conditions: T 0 ¼ 300K, yO2 ¼ yin , cO2 ¼ ρg yin =wO2 , p ¼ 0 cN2 ¼ ρg ð1  yin Þ=wN2 , cproducts ¼ 0, u ¼ 0

Results and Discussions

85.3.1 Non-extinction Regime

Here, y1, y2, y3 in Eqs. 85.1, 85.2, and 85.3 are mass fractions of virgin foam, char, and ash, respectively. Equations 85.4, 85.5, 85.6, and 85.7 indicate mass and momentum conservation of total gas, individual gas molconcentration conservation, and energy conservation, respectively. Boundary conditions: x ¼ 0 cm ∂T ∂ci k ¼ hðT 0  T Þ, p ¼ 0, Di ¼0 ∂x ∂x x ¼ 14 cm : u ¼ ui , coxygen ¼ ρg yo2 =wo2  cN2 ¼ ρg 1  yo2 =wN2 , cproducts ¼ 0

833

is selected, and the whole domain is divided into 1400 space elements. The initial time step is 0.001 s, and the maximum step is 0.1 s for a better convergence of solutions.

ð85:9Þ

y1 ¼ 1, y2 ¼ y3 ¼ 0

85.2.4 Solution Method

4x104 Heat release rate [W/kg]

  ∂ci ∂ ∂ci ∂ci Di þ ¼ Ri þu ∂t ∂x ∂x ∂x RO2 ¼ ρ0 ðωo μ1o þ ωa μ3o Þ=wO2
 Rproducts ¼ ρ0 μ1gp ωo þ μ2gp ωp þ μ3gp ωa =wgas ρg yi RN2 ¼ 0, DO2 ¼ Dproducts ¼ DN2 , ci ¼ wi

foam oxidation foam pyrolysis char oxidation

3x104 2x104 1x104 0

-1x104

b

-2x104 Mass fraction of solid & gas species

85

0

4

1.0

y

6

y

10

12

14 700

T(K) III

IV

0.6

8

1

0.8

II

600

I

g

500

0.4 y 0.2 y 0.0 0

The equations are solved by a package of software named COMSOL Multiphysics, which is a finite element solver with the capability of solving coupled partial differential equations. Backward differentiation formula (BDF) method

2

2

4

O2

y 2

6 8 10 Axial Distance (cm)

400

3

300 12

14

Fig. 85.1 (a) Heat release rates and (b) mass fractions of species and temperature vs. distance at t ¼ 600 s for an oxygen concentration of 0.23 and an inlet gas velocity of 5 mm/s

J. Yang et al.

Mass Fraction of ash

Fig. 85.2 Mass fractions of (a) foam, (b) char, (c) ash, and (d) total mass loss and loss rate vs. time for the base case of forward smoldering (Two positions at 0.5 cm and 1.5 cm from the heater, followed by six locations with 2 cm interval are monitored). The inlet gas velocity is 5 mm/s and oxygen concentration is 0.23

a 1.0 0.6 0.4 0.2 0.0

b

0

200

400

600

800

1000

1200

1400

0

200

400

600

800

1000

1200

1400

0

200

400

600

800

1000

1200

1400

0.18 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00

c

0.06 0.05 0.04 0.03 0.02 0.01 0.00

d Total mass loss

0.5cm 1.5cm 3.5cm 5.5cm 7.5cm 9.5cm 11.5cm 13.5cm

0.8

0.06

1.0

turning off the heater

0.05

0.8

MLR

0.04

0.6

0.03

0.4

0.02

0.2

0.01 0.00

0.0 0

200

400

600

800

1000

1200

Mass loss rate [g/s]

Mass fraction of Char Mass fraction of Foam

834

1400

Time [s]

In order to reveal the reaction dominating foam consumption, a scaling factor representing the foam depletion percentage by oxidation or pyrolysis is defined as Eq. 85.10. The data in Fig. 85.1a produces So  36 % and Sp  64 %, which indicates that the pyrolysis reaction dominates the foam consumption: Z

L

Z

L

Qo dx Δh f,o So ¼ Z L 0 ¼ Z L 0   Qp Qo ρ0 ωo þ ωp dx þ dx Δhf , o Δhf , p 0 0 ρ0 ωo dx

Sp ¼ 1  So ð85:10Þ Figure 85.1b presents four regions in the sample during smoldering. All the reactions are completed in Region I (ash zone). Region II is char oxidation zone, where char begins to be consumed and no fuel degradation reaction happens. Meanwhile, a peak temperature (700 K) is found at x ¼ 8.5 cm. Oxygen mass fraction decreases to 0.172 from 0.230 in this region. In Region

III (fuel degradation zone), foam (y1) is consumed by pyrolysis and oxidation. The mass fraction of char (y2) increases to 0.165 and that of gas products (yg) increases to 0.66. The slower increase of temperature at 593 K implies the onset of an endothermic pyrolysis reaction [6, 14]. In Region IV, foam is preheated mainly by the convective heat transfer due to gas flow, so the temperature increases monotonically. In Fig. 85.2a–c, mass fraction histories of solid species are presented at an interval of 2 cm. The average propagation speed of foam degradation front is 0.160 mm/s, and that of char oxidation front is 0.155 mm/s. Due to the convective cooling boundary, char oxidation is weaker in the last 1–2 cm to the outlet (12–13 cm far away from the heater), i.e., a higher char mass fraction but a lower ash mass fraction at x ¼ 13.5 cm as shown in Fig. 85.2b, c. The sample mass in Fig. 85.2d decreases with time after turning off the heater. The mass loss rate firstly increases sharply due to the consumption of foam near the heater, but then it reaches a steady period with an average value 0.045 g/ s due to a stable smoldering propagation.

Heat Loss and Kinetic Effects on Extinction and Critical Self-Sustained. . .

85

85.3.2 Possible Extinction Regimes 85.3.2.1 Inlet Gas Velocity and Oxygen Concentration Effects Leach et al. [11] analyzed the possible extinction regime of forced reverse smoldering and concluded that higher inlet gas velocity could lead to a sudden extinction. However, the forced forward smoldering underwent extinction at a lower gas velocity rather than a higher value [6]. A criterion given by Titus et al. [20] said that a section of sample was assumed to have smoldered if the temperature in that section exceeded 320  C. The present work uses this criterion and temperature profiles to discriminate where and when the smoldering wave is extinguished. Figure 85.3a shows that the smoldering velocity (average ratio of adjacent locations (2 cm) divided by the time interval corresponding to the peak temperature before extinction) is very low at an inlet gas velocity of 3.5 mm/s. The reaction rates of three-step reactions indicate that the char oxidation reaction away from the heater is almost not initiated. a Smoldering velocity [mm/s]

0.28 0.24

Numerical calculated J.L. Torero et al's Moldel predicted

835

Consequently, without the heat source from char oxidation, both foam pyrolysis and oxidation reactions are weaker and weaker until smoldering is extinguished at a distance of 7.5 cm from the heater. The final normalized foam mass for 3.5 mm/s inlet gas velocity is 39.8 % at t ¼ 1800s, calculated by Eq. 85.11. Z mf ¼ m0

L

ρ0 Aðy1 þ y2 þ y3 Þdx

0

ρ0 AL

ð85:11Þ

With the increase of inlet gas velocity, the extinction regime experiences a transition to another regime; especially when inlet velocity is higher than a critical threshold, the final normalized foam mass decreases insignificantly. So we can roughly regard 3.51 mm/s as the critical extinction threshold of the gas inlet velocity at the turning point, shown in Fig. 85.3b, in which self-sustained propagation is no longer possible. Taking heat loss into consideration, J.L. Torero et al. [13] proposed an analytical smoldering model based on the energy balance as Eq. 85.12. The speed of smoldering propagation was expressed as Eq. 85.13, where Qloss was the heat loss flux (100 W/m2 [5]): Usml

0.16

yO2 ρg ug Qsml  Qloss AL =AC i ð1  ϕÞρs cs þ ϕρg cg ðT s  T i Þ  ð1  ϕÞρs Qp þ yO2 ρg Qsml

0.12

ð85:12Þ

0.20

¼h

 ρg yO2 ug  U S Qsml ¼

0.08 0.04

extinction non-ignition

0.00 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 Inlet gas flow velocity [mm/s]

b

100

Normalized total mass (%)

90 80

non-initiation

70 60 50 40 30 20 10

extinction self-propagation

0 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 Inlet gas velocity (mm/s)

Fig. 85.3 (a) Smoldering velocity vs. inlet gas velocity and (b) normalized total foam mass for an oxygen concentration of 0.23

Qloss AL =AC h i þUs ð1  ϕÞρs cs þ ϕρg cg ðT s  T i Þ Us ð1  ϕÞρs Qp

ð85:13Þ Figure 85.3a compares the results of numerical calculation and J.L. Torero’s model. The numerical results agree well with Torero’s predictions at low inlet gas velocity except for the extinction and non-ignition cases. Nevertheless, Torero’s model overestimates the smoldering velocity at high inlet gas velocities, which is because that Torero’s model assumes complete consumption of oxygen in smoldering front, but in fact a small fraction of oxygen is still left. Similarly, the critical oxygen concentrations of extinction for different inlet velocities are determined and listed in Table 85.1. Lower inlet velocity requires higher critical oxygen concentration to ensure enough oxygen mass flux to reach the reaction front. Otherwise, the smoldering reaction will be extinguished or not initiated by the heater below these critical values.

836

J. Yang et al.

85.3.2.2 Heat Loss Effect Four different heat loss scenarios (different values of parameter H (W/m3) in Eq. 85.7) have been investigated in Fig. 85.4. At H ¼ 50 W/m3 in Fig. 85.4a, heat loss is too high, and the smoldering combustion is extinguished after switching off the heater, which is a non-initiation case. At H ¼ 42 W/m3 in Fig. 85.4b, there is a sudden peak temperature drop at 9.5 cm away from the heater where an extinction of char oxidation happens. Without the heat generation from char oxidation, heat release of foam oxidation is not enough to support a self-propagation, i.e., extinction happens. About 5.562 g foam and 2.420 g char calculated by Eq. 85.14 at t ¼ 1500s are unburned.

Table 85.1 Critical oxygen concentrations and final total mass for different inlet velocities us (mm/s) 0.1672 0.1458 0.1379 0.1244 0.1328

Fig. 85.4 Temperature distributions of every spatial location as a function of heat loss for (a) H ¼ 50, (b) H ¼ 42, (c) H ¼ 26, and (d) H ¼ 0 W/m3 at an inlet gas velocity of 5 mm/ s and oxygen concentration of 0.23 (interval of 2 cm from the heater to the outlet (1.5–13.5 cm))

TS (K) 705 703 699 691 698

w1(%) 8.29 8.17 8.27 8.88 8.87

L

mf ¼

Z

L

ρ0 Ay1 dx, mc ¼

0

ρ0 Ay2 dx

ð85:14Þ

0

When H is 26 W/m3 in Fig. 85.4c, the smoldering front almost has self-propagated to the outlet due to a less heat loss. The final mass of virgin foam is 1.0 g and only 0.24 g char is left. One special case is captured in Fig. 85.4d when H is 0, i.e., an adiabatic case, where the temperature does not drop after the arrival of smoldering front, so both the smoldering velocity (0.163 mm/s) and peak temperature (731.4 K) are larger than the nonadiabatic cases (a)(b) (c). The smoldering velocities and peak temperatures at different heat loss coefficient are shown in Fig. 85.5. In stable self-propagation regime, a quasi-linearly decreasing peak temperature and a parabolic diminishing smoldering velocity with increasing the heat loss to environment are observed.

85.3.2.3 Effect of Enthalpy of Reaction In Case 1a–e in Table 85.2 of Appendix A, heat release of foam oxidation ΔhO changes in a range from 500 to 900 J/ g. Comparison of Case 1a–c and the base case shows that lower ΔhO results in a smaller peak temperature and

a

b 650 600

Temperature [K]

yO2,c 0.35 0.225 0.200 0.170 0.155

550 500 450

700

1.5cm 3.5cm 5.5cm 650 7.5cm 9.5cm 600 11.5cm 550 13.5cm 500 450

400

400

350 300 250

350 H = 50 0 200 400 600 800 100012001400

c

Temperature [K]

ug0(mm/s) 3 4.5 5 6 7

Z

300 250

H = 42 0 200 400 600 800 100012001400

d 700

700

600

600

500

500

400

400

300

H = 26 0 200 400 600 800 100012001400 Time (s)

300

H=0 0 200 400 600 800 100012001400 Time (s)

85

Heat Loss and Kinetic Effects on Extinction and Critical Self-Sustained. . .

(Case 3e, 3f) has little effect on the smoldering velocity and peak temperature.

0.20

650

0.16

Non-initiation half-way quenched

0.14 0.12

600 550 500

stable smoldering propagation

450

Temperature [K]

Smoldering Velocity [mm/s]

700 0.18

400 0.10

VS(mm/s)

350

TS(K)

0.08 0

10 20 30 40 50 Heat loss coefficient H (W/m^3)

837

300 60

Fig. 85.5 Smoldering velocity and peak temperature as a function of volumetric heat loss coefficient H for an inlet gas velocity of 5 mm/ s and an oxygen concentration of 0.23

smoldering velocity but slightly a higher total mass, which is attributed to less accumulation of energy. As ΔhO is further decreased to 637.5 J/g, only the foam near the heater is consumed, and more than half of the foam is unreacted. Likewise, larger total mass (85.33 %) is left under a much smaller heat release (Case 1e). An approximate critical value (Case 1c*) is obtained by a method of bisection [21]. In Case 2a–e, heat release of foam pyrolysis Δhp is varied from 1000 J/g to 775 J/g. Smoldering velocity and peak temperature decrease with the increase of Δhp due to greater energy absorbed by the endothermic reaction. Especially for Case 2d where the smoldering wave is extinguished at 5.5 cm away from the heater, the remaining total foam mass fraction is 34.87 %. Greater increase (Case 2e) of Δhp will result in non-ignition because of inadequate net heat release. Case 3ac indicate that larger Δha (heat release of char oxidation) leads to higher smoldering velocity and peak temperature. In Case 3d, extinction happens at 9.5 cm away from the heater due to weaker char oxidation. In contrast to Cases 1 and 2, a weak reaction regime happens at smaller Δha in which the smoldering velocity and peak temperature are significantly lower than that of the selfpropagation regime. In weak reaction regime, the heat source sustaining a weaker and weaker foam degradation results from foam oxidation, and almost no ash is formed due to absence of char oxidation, so a great deal of char is formed by the foam oxidation and pyrolysis, which prevents heat loss to environment. Interestingly, further decrease of Δha

85.3.2.4 Effects of the Frequency Factor and Activation Energy Case 1a–d in Table 85.3 of Appendix A indicates that a decrease in frequency factor of foam oxidation reaction (AO) leads to a reduction of smoldering velocity and peak temperature but an increase in final foam weight, which is due to the weakened foam oxidation reaction. Contrary, in Case 2(a–d), the smoldering velocity and peak temperature undergo a reduction with the increasing frequency factor of foam pyrolysis reaction (Ap), which attributes to the enhanced endothermic pyrolysis reaction. In Case 3a–f, the effect of the frequency factor of char oxidation reaction (Aa) is interesting. An increase of Aa will result in an insignificant increase of smoldering velocity and final foam weight, which is inconsistent with results of Leach et al. [12]. But a diminishing peak temperature with increasing Aa is unexpected, which is consistent with the model results of the reference [22]. It can be explained from the curves in Fig. 85.6. Three frequency factors of 2  1013[1/s], 4  10 13 [1/s], and 6  1013[1/s] are considered. A faster char oxidation reaction rate but a lesser char mass fraction is obtained under larger Aa in Fig. 85.6a, b, that is, the insulated hot char layer behind the smoldering front is thinner, so a lower peak temperature is appropriate due to less heat accumulation in the localized char layer as indicated in Fig. 85.6c. When Aa is too small to compensate for the heat absorption and heat loss to environment (Case 3e), the smoldering wave is suddenly interrupted at a distance of 7.5 cm away from the heater. It should be emphasized that a weak reaction regime (Case 3f) occurs because of an extremely small Aa and thus hot char with a significant amount of fuel is left behind the smoldering front, which differs from the effects of AO and Ap. Similarly, the activation energy of each reaction step plays a vital role in smoldering regimes as indicated in the Table 85.4 of Appendix A. An increase in activation energy of foam oxidation (Case 1a–e) or a decrease in activation energy of foam pyrolysis (2a–d) will lead to a reduction of smoldering velocity and peak temperature but an increase in final foam weight, which is ascribed to the weakened foam oxidation and enhanced foam pyrolysis reaction. It is noted that smoldering regime is extremely sensitive to pyrolysis activation energy. As compared with base case and Case 3a–b, the increase of Ea has insignificant effect on smoldering velocity and final foam weight. Larger Ea results in smaller char oxidation rate but higher

J. Yang et al.

Fig. 85.6 (a) Char oxidation reaction rate, (b) char mass fraction, (c) temperature vs. distance at t ¼ 200 s, 400 s, 600 s, and 800 s under different frequency factors, 2e13, 4e13, and 6e13 1/s, of char oxidation reaction for an inlet gas velocity of 5 mm/s and an oxygen concentration of 0.23

char oxidation reaction rate [kg/s]

838

a

0.12

2e13 4e13 6e13

0.10

800s

600s

400s 200s

0.08 0.06 0.04 0.02 0.00 0

2

b

4 800s

Char mass fraction

0.12

8

600s

6e13 4e13 2e13

0.16

6

10

12

400s

14 200s

0.08 0.04 0.00

c

0

2

4

Temperature [K]

800 2e13 4e13 6e13

700 600

6

8 600s

800s

10

12

400s

14 200s

500 400 300 0

2

char mass fraction, so peak temperature increases reasonably. Below the critical threshold in Case 3a–c, the smoldering wave can sustain self-propagation; otherwise, it will encounter extinction (Case 3c) or a weak reaction (Case 3d, e). It is noted that the smoldering propagation hardly depends on the char oxidation reaction when Ea exceeds 198[kJ/mol] due to non-initiation of char oxidation.

85.4

Conclusions

A transient one-dimensional model of forced forward smoldering is analyzed to study the mechanism of extinction. The mass fractions of solid and gas species for non-extinction regime indicate the sample can be divided into ash zone, char layer zone, foam degradation zone, and virgin foam zone from heating end to the outlet.

4

6 8 Axial Distance [cm]

10

12

14

Effects of gas flow velocity, heat loss, and kinetic factors on smoldering process, especially for extinction regime, are investigated in this work. A lower inlet gas velocity will lead to extinction or non-ignition. Heat loss has the most prominent effect on smoldering regimes which is primarily related to the sample size, such as lateral area to volume ratio of the smoldering front and the cross-sectional area of the fuel. However, the kinetic effects primarily lie in enthalpy, frequency factor and activation energy. Large activation energy of char oxidation can speed up extinction due to more heat required to trigger smoldering reactions. In general, extinction is caused by (1) larger heat loss to environment and (2) limited kinetics. Acknowledgment This work was sponsored by National Basic Research Program of China (973 Program, NO. 2012CB719702) and International Science and Technology Cooperation Program of China (2014DFG72300). HX Chen was supported by the Fundamental Research Funds for the Central University (WK2320000020).

85

Heat Loss and Kinetic Effects on Extinction and Critical Self-Sustained. . .

839

Appendix A

Table 85.2 Effect of chemical reaction heat on smoldering regimes

Cases Base case

Δho [J/g] 900

Δhp [J/g] -775

Δha [J/g] 4600

Ts [K]

VS [mm/s]

w [%]

Smoldering regimes

708

0.1561

7.69

Self-propagation

1a 700 -775 4600 700 0.1367 8.65 Self-propagation 1b 680 -775 4600 698 0.1349 8.85 Self-propagation 648 -775 4600 679 0.1331 9.29 Critical condition 1c* 1d 638 -775 4600 — — 57.08 Non-ignition 1e 500 -775 4600 — — 85.33 Non-ignition 2a 900 -875 4600 702 0.1443 8.33 Self-propagation 2b 900 -925 4600 700 0.1388 8.88 Self-propagation 2c* 900 -928 4600 700 0.1390 8.86 Critical condition 34.87 Extinction at x=5.5cm 2d 900 -930 4600 624 0.0842a 2e 900 -1000 4600 — — 69.46 Non-ignition 3a 900 -775 4000 701 0.1558 7.80 Self-propagation 3b 900 -775 3500 694 0.1477 8.45 Self-propagation 900 -775 2986 685 0.1410 9.20 Critical condition 3c* 20.52 Extinction at x=9.5cm 3d 900 -775 2980 642 0.1007a 3e 900 -775 1800 576 0.1365 37.05 Weak reaction 3f 900 -775 600 573 0.13 43.49 Weak reaction a For extinction, the smoldering velocity Vs is calculated by the adjacent a space and time intervals in the self-sustained smoldering region

Table 85.3 Effect of the frequency factor on smoldering regimes

Cases Base case 1a 1b 1c 1d 2a 2b 2c 2d 3a 3b 3c 3d 3e 3f

Ao [1/s] 2×1012 3×1012 1.75×1012 1.65×1012 2×1011 2×1012 2×1012 2×1012 2×1012 2×1012 2×1012 2×1012 2×1012 2×1012 2×1012

Ap [1/s] 5×1015 5×105 5×1015 5×1015 5×1015 6×1015 7×1015 7.8×1015 8×1015 5×1015 5×1015 5×1015 5×1015 5×1015 5×1015

Aa [1/s] 4×1013 4×1013 4×1013 4×1013 4×1013 4×1013 4×1013 4×1013 4×1013 10×1013 6×1013 2×1013 5×1012 4.85×1012 1×1012

TS [K] 708 715 701 698 — 703 698 687 — 691 698 717 720 663 574

VS [mm/s] 0.1561 0.1607 0.1435 0.1395 — 0.1501 0.1461 0.1285 — 0.1622 0.1588 0.1544 0.1467 0.1010 0.1315

w [%] 7.69 7.41 8.12 8.80 98.41 6.97 7.49 8.64 60.17 7.76 7.74 7.65 7.50 22.97 44

Smoldering regimes Self-propagation Self-propagation Self-propagation Critical condition Non-ignition Self-propagation Self-propagation Critical condition Non-ignition Self-propagation Self-propagation Self-propagation Critical threshold Extinction at x=7.5cm Weak reaction

840

J. Yang et al.

Table 85.4 Effect of the activation energy on smoldering regimes

Cases Base case 1a 1b 1c* 1d 1e 2a 2b* 2c 2d 3a 3b 3c 3d 3e

Eo [kJ/mol] 155 154 153 156 157 160 155 155 155 155 155 155 155 155 155

Ep [kJ/mol] 200 200 200 200 200 200 199.5 199 198 180 200 200 200 200 200

Ea [kJ/mol] 185 185 185 185 185 185 185 185 185 185 180 190 197.4 250 800

References 1. Rein G, Cohen S, Simeoni A (2009) Carbon emissions from smouldering peat in shallow and strong fronts. Proc Combust Inst 32:2489–2496 2. Stocker DP, Olson SL, Urban DL, Torero JL, Walther DC, Carlos Fernande-Pello A (1996) Small-scale smoldering combustion experiments in microgravity. Proc Combust Inst 26:1361–1368 3. Walther DC, Fernandez-Pello AC, Urban DL (1999) Space shuttle based microgravity smoldering combustion experiments. Combust Flame 116:398–414 4. Bar-Ilan A, Rein G, Fernandez-Pello AC, Torero JL, Urban DL (2004) Forced forward smoldering experiments in microgravity. Exp Thermal Fluid Sci 28:743–751 5. Bar-Ilan A, Rein G, Walther DC, Fernandez-Pello A, Torero JL, Urban DL (2004) The effect of buoyancy on opposed smoldering. Combust Sci Technol 176:2027–2055 6. Rein G, Bar-Ilan A, Fernandez-Pello AC, Ellzey JL, Torero JL, Urban DL (2005) Modeling of one-dimensional smoldering of polyurethane in microgravity conditions. Proc Combust Inst 30:2327–2334 7. Rein G, Carlos Fernandez-Pello A, Urban DL (2007) Computational model of forward and opposed smoldering combustion in microgravity. Proc Combust Inst 31:2677–2684 8. Rein G, Lautenberger C, Fernandezpello A, Torero J, Urban D (2006) Application of genetic algorithms and thermogravimetry to determine the kinetics of polyurethane foam in smoldering combustion. Combust Flame 146:95–108 9. Dodd A, Lautenberger C, Fernandez-Pello A (2009) Numerical examination of two-dimensional smolder structure in polyurethane foam. Proc Combust Inst 32:2497–2504 10. Schult D, Matkowsky B, Volpert V, Fernandez-Pello A (1995) Propagation and extinction of forced opposed flow smolder waves. Combust Flame 101:471–490

TS [K] 708 719 727 698 — — 703 697 — — 702 726 635 573 573

VS [mm/s] 0.1561 0.1909 0.2189 0.1376 — — 0.1496 0.1417 — — 0.1572 0.1543 0.1062 0.1282 0.1276

w [%] 7.69 7.01 6.67 8.97 87.29 98.27 7.76 8.20 76.36 97.75 7.66 7.62 26.98 45.19 45.12

Smoldering regimes Self-propagation Self-propagation Self-propagation Critical condition Non-ignition Non-ignition Self-propagation Critical condition Non-ignition Non-ignition Self-propagation Self-propagation Extinction at x=5cm Weak reaction Weak reaction

11. Leach S, Ellzey J, Ezekoye O (1998) Convection, pyrolysis, and Damko¨hler number effects on extinction of reverse smoldering combustion. In Symposium (International) on combustion. Elsevier. pp 2873–2880 12. Leach S, Ellzey J, Ezekoye OA (1998) Convection, pyrolysis, and Damko¨hler number effects on extinction of reverse smoldering combustion. Proc Combust Inst 27:2873–2880 13. Torero J, Fernandez-Pello A, Kitano M (1993) Opposed forced flow smoldering of polyurethane foam. Combust Sci Technol 91:95–117 14. Torero J, Fernandez-Pello A (1996) Forward smolder of polyurethane foam in a forced air flow. Combust Flame 106:89–109 15. Torero J, Fernandez-Pello A (1995) Natural convection smolder of polyurethane foam, upward propagation. Fire Saf J 24:35–52 16. Walther DC, Anthenien RA, Fernandez-Pello AC (2000) Smolder ignition of polyurethane foam: effect of oxygen concentration. Fire Saf J 34:343–359 17. Dodd AB, Lautenberger C, Fernandez-Pello C (2012) Computational modeling of smolder combustion and spontaneous transition to flaming. Combust Flame 159:448–461 18. Rogers F, Ohlemiller T (1980) Smolder characteristics of flexible polyurethane foams. J Fire Flammability 11:32–44 19. Leach SV, Rein G, Ellzey J, Ezekoye OA, Torero JL (2000) Kinetic and fuel property effects on forward smoldering combustion. Combust Flame 120:346–358 20. Titus R, Bar-Ilon A, Fernandez-Pello C (2006) The effects of environmental parameters on smoldering propagation in small polyurethane foam samples. In 44th AIAA Aerospace Sciences Meeting and Exhibit, p 742 21. Weber R, Sidhu H, Sawade D, Mercer G, Nelson M (2006) Using the method of lines to determine critical conditions for thermal ignition. J Eng Math 56:185–200 22. Di Blasi C (1995) Mechanisms of two-dimensional smoldering propagation through packed fuel beds. Combust Sci Technol 106:103–124

Numerical Study of the Radiative and Turbulent Heat Flux Behavior of Upward Flame Spread Over PMMA

86

Alexander Karpov, Artem Shaklein, Mikhail Korepanov, and Artem Galat

Abstract

Upward flame spread has been analyzed by the coupled approach including turbulent flow dynamics, radiative flame heat transfer, modified combustion model, and solid fuel pyrolysis through numerical solution of two-dimensional equations. A contribution of the radiative surface heat transfer has been evaluated. Pulsating heat flux on the solid fuel surface has been found for the regime approaching the transition from laminar to turbulent. Solidphase dissipative transfer characteristics at the surface show under-relaxation behavior in response to turbulent fluctuations. Keywords

Upward flame spread Coupled analysis

Nomenclature C D E G g k L p Pr Q q R R0 Sc T t u

Specific heat capacity (J/kg K) Diffusion coefficient (m2/s) Activation energy (J/mole) Incident radiation intensity (W/m2) Gravity acceleration (m/s2) Kinetic energy of turbulence (m2/s2) Fuel slab thickness (m) Pressure (Pa) Prandtl number () Heat release (J/kg) Heat flux (W/m2) Specific gas constant (J/kg K) Universal gas constant (J/mole K) Schmidt number () Temperature (K) Time (s) Velocity (m/s)

A. Karpov (*)
A. Shaklein
M. Korepanov
A. Galat Institute of Mechanics, 34 T.Baramzinoi, Izhevsk 426067, Russia e-mail: [email protected]




Turbulent combustion

W x Y y




Radiation model




PMMA burning




Reaction rate (s1) Coordinate normal to fuel surface (m) Mass fraction () Coordinate along fuel surface (m)

Greek Symbols ε λ κ μ v ρ ω

Turbulent dissipation (m2/s3) Thermal conductivity (W/m K) Absorption coefficient (m1) Dynamic molecular viscosity (Pa s) Stoichiometric coefficient () Density (kg/m3) Turbulent frequency (s1)

Subscripts a F g O P

Ambient Fuel Gas Oxidizer Products

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_86

841

842

r s sgs t w

86.1

A. Karpov et al.

Radiative Solid Sub-grid scale Turbulent Wall

Introduction

The upward flame spread phenomenon has been historically developed into a significant branch of the fire safety science. The systematic approach for studying such a process was founded in the early experimental works (e.g., [1–5]), where sizes of pyrolysis and flame zones have been measured, a flame spread rate has been determined, and simplified algebraic correlations between these parameters have been formulated. Such, a PMMA combustion [1, 4, 5] and a flame spread over vertical surfaces moistened by liquid fuel substances [2, 3] were studied. In [3], a flame spread rate has been linked with the normalized heat flux from a flame to a solid surface in the form of an integral model. This model was further developed [6, 7] and is widely used in upto-date engineering practice. Despite the fact that a technical level of measurement technique has been considerably improved [8, 9], a general well-recognized approach to quantitative estimation of the flame spread parameters is based on the correlations introduced in 80 s. Nowadays, the progress achieved in the development of physical models and numerical methods and continually rising computing capabilities (which likely states as the most important factor) turn to be close enough to perform comprehensive theoretical analysis of considered upward flame spread by means of computational fluid dynamics. Thus, CFD basic principles relatively to fire modeling have been outlined (e.g., [10]). Present efforts are intended to develop an advanced approach, which could integrate possibly all essential physical factors affecting the specific flame spread process. In regard to upward flame spread considered here, primary effects are caused by dominant influence of natural convection providing the growing size of flame zone, which results in substantial contribution of turbulent and radiative heat transfer. Then, proper analysis of heat transfer in the heterogeneous “gassolid” system should be based on the coupled approach, which ensures the existence of self-sustained mode for gaseous flame spreading over solid fuel surface. A non-coupled model of PMMA pyrolysis suggested in [11] showed fairly good agreement with the experimental data. A coupled algorithm for the calculation of feedback interaction of heat transfer between gas phase and solid fuel was developed [12]. However, this model failed to predict correctly the position of leading edge of the flame spreading upward, which was explained by lack of resolution applied to gas-phase physics, especially in a view of employing the RANS turbulent models. Then, DNS approach is claimed to

be applied [13] to the modeling of upward turbulent flame spread. Meanwhile, such an implementation seems to be rather doubtful since two-dimensional formulation is in use there. An appropriate usage of DNS requires an involvement of full (three-dimensional) Navier-Stokes equations into analysis [14, 15]. Despite the fact that some (occasionally reasonable) vortex flow structure may be obtained through 2D simulation, a shear stress tensor could not be accurately resolved in that case, because one of the velocity components is missed. In fact, present DNS challenge to predict reacting flows in large-scale flame stands certainly far from the total success in practice. Without matching of a mesh size to the Kolmogorov length scale (the requirement staying far from being met now), just the quasi-DNS can be achieved [16]. From a viewpoint of the modeling of gas-phase processes, very advanced model has been reported in [17], where the following technique is employed for the description of physical effects: turbulent flow parameters are modeled by LES approach with for low Reynolds modification in a form of WALE (Wall-Adapting Local EddyViscosity) model for near-surface characteristics; turbulent combustion is described by EDC (Eddy Dissipation Concept), or EBU (Eddy Breakup) in other terms, modified for accounting the laminar combustion regime; and radiation heat transfer model is based on a thin optical layer approximation. Again, there is a certain limitation of model [17] if it would be applied to wider flame spread phenomena because fuel supply from the surface is prescribed. Say, an uncoupled analysis takes place. Summing up above considerations, the following requirements are drawn up for the proper modeling of upward turbulent diffusion flame spread over solid combustibles: couple analysis should be involved to determine the character of feedback interaction between gas-phase flame and solid fuel through the conjugate heat transfer at the surface; effects of turbulence and radiation are to be predicted as carefully as available technique allows both in the means of developed physical models and affordable computing capabilities. Some efforts in resolving of this problem are presented below.

86.2

Formulation

86.2.1 Gas-Phase Fluid Dynamics By using the LES concept, the set of Favre-filtered equations describing an unsteady non-isothermal turbulent multicomponent flow, including the effects of combustion reaction and radiative heat transfer, are as follows: uj ∂ρ ∂ρe þ ¼ 0; ∂t ∂xj

ð86:1Þ

86

ρ

Numerical Study of the Radiative and Turbulent Heat Flux Behavior of Upward Flame Spread Over PMMA

∂e ui ∂e ui þ ρe uj ¼ ∂t ∂xj



 ∂e ∂p ∂  ui þ μsgs þ μ þ ðρa  ρÞgi ; ∂xi ∂xj ∂xj

ð86:2Þ ρC

e e ∂T ∂T þ ρCe uj ¼ ∂t ∂xj

eQþ þ ρW

  e μsgs ∂T ∂p ∂ C þλ þ ∂xj Prsgs ∂t ∂xj

∂e q jr ; ∂xj

ð86:4Þ 

ð86:11Þ

e ¼ μsgs G e μ; G

ð86:12Þ

1 ∂k ∂ω 10 ¼ max 2σ ω2 ; 10 ; ω ∂xj ∂xj

ð86:13Þ



ð86:3Þ

  ∂Ye F ∂Ye F ∂ μsgs ∂Ye F e þ ρe ρ uj ¼ þ ρD  vF ρW; ∂t ∂xj ∂xj ∂xj Scsgs

ρ

Other parameters are expressed as [22]:   uj u i ∂e u i ∂e e μ ¼ ∂e þ ; G ∂xj ∂xj ∂xi

843

CDkω

( F1 ¼ tanh

  4 ) pffiffiffi

k 500μ 4ρσ ω2 k min max * ; 2 ; ; β ωx ρx ω CDkω x2 ð86:14Þ



eO ∂Y ∂Ye O ∂ μsgs ∂Ye O e þ ρe uj ¼ þ ρD  vO ρW; ∂t ∂xj ∂xj ∂xj Scsgs

ð86:5Þ

( F2 ¼ tanh

 pffiffiffi  2 ) 2 k 500μ ; max * ; 2 β ωx ρx ω

ð86:15Þ

  where x is a distance to a nearest wall. ∂Ye P ∂ μsgs e All model constants are computed by blending function þ ρD þ ðvO þ vF ÞρW ; ∂xj ∂xj Scsgs ϕ ¼ F1 ϕ1 þ ð1  F1 Þϕ2 and have the following values: ð86:6Þ σk1 ¼ 0.85034, σk2 ¼ 1.0, σω1 ¼ 0.5, σω2 ¼ 0.85616, α1 ¼ 0.5532, α2 ¼ 0.4403, β1 ¼ 0.075, and β2 ¼ 0.0828. e p ¼ ρRT: ð86:7Þ Other constants are CDES ¼ 0.65, β* ¼ 0.09, Prsgs ¼ 1.0, Scsgs ¼ 1.0, a1 ¼ 0.31, and c1 ¼ 10.0. As pointed out above on the nature of couple “gas-solid” problem, the structure of core vortex flow should be resolved as well as local surface characteristics. Since LES approach 86.2.2 Combustion Model [18–20] does not penetrate close to predict wall flow behavior, some modifications are to be applied. Here, the DDES Here, the analysis is carried out for the single combustion (Delayed Detached Eddy Simulation) as approved by [21– macro-reaction [24]. EDC/EBU model [25, 26] is modified to merge turbulent combustion regime with the kinetic 23] is engaged: mechanism occurring under laminar flow:    ∂k uj ∂ρk ∂ρke ∂  *   e c1 β ρkω þ ¼ σ k μsgs þ μ þ min G; e ¼ min W e kin ; W et ; ∂xj ∂t ∂xj ∂xj W ð86:16Þ *  ρωβ ek; ð86:8Þ   e O exp Eg =R0 T e ; eFY e kin ¼ kY ð86:17Þ W  ∂ω uj ∂ρω ∂ρωe ∂  e μ  ρβω2 þ ¼ σ ω μsgs þ μ þ αρG ∂t ∂xj ∂xj ∂xj ! eF Y eO   Y  ρðF1  1ÞCDkω : ð86:9Þ e t ¼ ASij min W ; ; ð86:18Þ vF νO Here, turbulent sub-grid scale viscosity is defined as   qffiffiffiffiffiffiffiffiffiffiffiffi   S ij . S ij e where e S ij  ¼ 2e a1 k     μsgs ¼ ; ð86:10Þ e max a1 ω; S F2 eP eP ∂Y ∂Y þ ρe ρ uj ¼ ∂t ∂xj

where e S ij ¼ 12



∂e ui ∂xj

 ∂e u þ ∂xij is the strain rate tensor.

86.2.3 Radiation Model Radiative transfer is modeled by the first-order spherical harmonic method (also known as P1), where radiation intensity, being the transferred variable, is integrated over all

844

A. Karpov et al.

possible directions [27]. Thus, radiative heat flux is introduced to follow gradient heat transfer approach: e e ∂G ; qejr ¼ Γ ∂xj

ð86:19Þ

The rate of thermal decomposition is expressed as W s ¼ ks expðEs =R0 T s Þ;

ð86:26Þ

which yields the linear pyrolysis rate of the injection of gasified fuel component from the solid fuel surface:

where

Z0 e¼ Γ

1 : 3ðe κ þ σeS Þ  Ce σS

ð86:20Þ

Us ¼

ð86:27Þ

W s dx: L

The radiative heat transfer is considered to be isotropic, while scattering is neglected. So, Eq. 86.20 is simplified to

86.2.5 Boundary Conditions e¼ 1 : Γ 3e κ

ð86:21Þ

A differential equation of the incident radiation transfer is [27] !   e 1 1 ∂G e ¼ 0: e4  G ð86:22Þ þe κ 4σ T ∂xj 3e κ ∂xj

Then, boundary condition set for two-dimensional statement is formulated. For the gas phase (Fig. 86.1): e ¼ Ta, Y eO ¼ Y a , Y e F ¼ 0, Y e P ¼ 0, y¼0: T O ∂e u =∂y ¼ 0, ∂e v =∂y ¼ 0, e ¼ Ta, Y eO ¼ Y a , Y e F ¼ 0, Y e P ¼ 0, x ¼ Lg : T O ∂e u =∂x ¼ 0, ∂e v =∂x ¼ 0,

Wall radiative heat flux is expressed as !   e 1 ∂G εw r e w ; ð86:23Þ e4  G e 4σ T ¼ qw ¼  w 3e κ ∂x 2 ð2  ε w Þ w

where εw is the surface emissivity. The absorption coefficient of the gas-phase mixture is expressed as X e ie e κ¼ Y κ i: The gas medium is gray, with the absorption coefficient being averaged over the whole spectrum. The absorption coefficient dependence upon temperature is taken into account by HITEMP tables [27]. The data are approximated in the form of fifth-order polynomial on 300–2500 K temperature interval: e 3α þ a4 Te 4α þ a5 Te 5α : ð86:24Þ e κ ¼ a0 þ a1 Te α þ a2 Te 2α þ a3 T

e =∂y ¼ 0, ∂Y e m =∂y ¼ 0, m ¼ fF; O; Pg, y ¼ h : ∂T ∂e u =∂y ¼ 0, ∂e v =∂y ¼ 0, x ¼ 0 : ρe u ¼ ρs U s , e v ¼ 0,  ρD

e w , m ¼ fF; O; Pg, e m ¼ ρs U s Y uY þ ρe m e w ¼ 1, Y e w ¼ 0, Y e w ¼ 0, k ¼ 0, Y F O P sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi  2  pffiffiffi 2 k 6μ ω¼ þ Cμ χx ρβ1 x2

For the solid-state domain: x ¼ L, y ¼ 0, y ¼ h : ∂T s =∂n ¼ 0: The coupled heat transfer on the solid fuel surface is described by the following boundary conditions: e T s ¼ T;

86.2.4 Solid Fuel Pyrolysis The energy conservation in the solid fuel is described by the following equation: ρs Cs

∂T s ∂ ∂T s ¼ λs þ ρs W s Q s : ∂xj ∂xj ∂t

ð86:25Þ

em ∂Y ∂x

∂T s þ ðρvCT Þs ¼ λs  ∂x  e μsgs ∂T eþe þ ρe þλ v CT q wr ;  C ∂x Prsgs

ð86:28Þ

ð86:29Þ

where e q wr radiative heat flux at a solid surface is determined through Eq. 86.23.

86

Numerical Study of the Radiative and Turbulent Heat Flux Behavior of Upward Flame Spread Over PMMA

845

Fig. 86.2 Surface heat flux distribution: 1 radiative, 2 molecular, 3 total (sum of 1 and 2)

Fig. 86.1 Computational domain

86.3

Solution Procedure

The finite volume method (e.g., [28]) is used for numerical solution of Eqs. 86.1, 86.2, 86.3, 86.4, 86.5, and 86.6, 86.8 and 86.9, 86.22, and 86.25. The fluid and solid energy conservation equations are strongly coupled. In the solution procedure, two numerical grids are presented. The fluid mesh is used to discretize the fluid equation set 86.1, 86.2, 86.3, 86.4, 86.5, and 86.6, 86.8, and 86.9, and 86.22; the compound mesh is used to solve the coupled energy Eqs. 86.3 and 86.25. The pressure and velocity fields are linked by the PISO method [28, 29]. The following grid size was arranged: 220 elements along x-direction (normal to the fuel surface) of which 60 elements relate to solid body (fuel) and 160 ones to gas phase and 2000 elements are included along y-direction (longitudinal, being upward at their interpretation). In order to provide correct numerical approximation of heat flux on the fuel surface, an irregular grid is arranged in x-direction with the minimal grid size adjacent to the surface which is set to 2 105 m for the gas phase and 3 104 m for the solid fuel. The grid size gradually increases from the surface to gas and solid. As for the y-direction, constant grid size of 2.5 103 m is employed.

86.4

Results

First, a tentative numerical study of the laminar upward flame spread over vertical PMMA surface is carried out. The sample slab is h ¼ 0.25 m high and L ¼ 0:006 m

Fig. 86.3 Surface heat flux distribution in case of flame spreading: 1 upward (experiment [5]), 2 downward (experiment [5]), 3 upward and downward (present simulation)

thick. Thermophysical and chemical kinetics parameters used here have been taken from [30–32]. The computed heat flux at the solid surface at the time 250 s from the ignition is presented in Fig. 86.2. Analysis of curves one to two shows that the radiative heat flux is at least of the same order as molecular one along the fuel surface. Moving upward to the flame plume, radiation level increases noticeably. Thus, a quite expected behavior is concluded that radiative heat transfer could not be neglected for the upward flame spread even in the case of such non-large-scale flame. Then, comparison of the computed wall heat flux with the experimental data [5] is shown in Fig. 86.3. Since present study has been carried out for the flame spreading both upward and downward from the ignition point, uniform heat flux distribution is obtained. This wall heat flux profile has good agreement with the experimental

846

A. Karpov et al.

Thus, upward flame spread over a PMMA surface is investigated. In this case, solid fuel slab height is increased up to h ¼ 5 m having thickness of L ¼ 0.006 m. Turbulence modeling is by DDES approach [21]. Here, k – ω SST model is applied to filtered equations for core flow as well as for RANS equations describing near-wall flow. Instantaneous values of heat transfer parameters affected by turbulent flow are shown in Fig. 86.4 for different time moments. Molecular (curve 2) and accordingly total (curve 1) heat fluxes have an apparent non-monotonic distributions due to turbulent fluctuations in the vicinity of a solid wall. On the other hand, radiative heat flux (curve 3) has smooth distribution along the fuel surface, because of every point along the surface receives heat by radiation from the whole hot gas flame zone integrated over all directions unlike the normal one if regarding to molecular transfer. Therefore, radiative heat transfer is rather much dissipative process, averaging the turbulent flow pulsating field (Fig. 86.4a, b). However, as flame-spreading process develops, surface temperature increases and surface irradiation level grows, resulting pulsating-like profile (Fig. 86.4c, d). Unlike the pulsating (instantaneous) heat flux distribution along the solid fuel surface at the initial stages of flame spread, wall temperature (Fig. 86.5a, b) shows smooth profile because of the inertial character of energy conservation in the solid, which has other relaxation periods. Nevertheless, as flame spread proceeds further, wall temperature is going to reflect turbulent fluctuations (Fig. 86.5c, d). The effect of turbulent transfer on the surface temperature could be estimated by comparing with the effective pyrolysis temperature (TP ¼ 615 K). Estimation of the turbulence modeling approaches has been carried out to estimate upward flame spread over solid fuels. DDES and qDNS (quasi-DNS) approaches have been employed to obtain instantaneous gas-phase flow dynamics and heat transfer parameters. This study is intended to evaluate their availabilities to predict local surface characteristics. The distribution of the pyrolysis rate along the solid surface is shown on Fig. 86.6. It could be easily noticed that RANS approach failed to predict the effect of the core flow turbulent behavior, while DDES and qDNS show apparent effect of vortical flow structure. However, qDNS results appear to be a bit less exposed upon turbulence. Fig. 86.4 Surface heat flux distribution computed by DDES model (1 total, 2 molecular, 3 radiative): (a) 0 s, (b) 120 s, (c) 160 s, (d) 200 s

86.5

values both for upward (3 vs 1) and downward (3 vs 2) flame spread. As it has been pointed out above, the primary aim of the present study is focused on the turbulent flame behavior.

Results presented above are focused on the implementation purely coupled analysis to the problem of the upward flame spread over the solid combustibles. Thus, the effects of turbulent and radiative transfer have been engaged. Basic

Concluding Remarks

86

Numerical Study of the Radiative and Turbulent Heat Flux Behavior of Upward Flame Spread Over PMMA

847

Fig. 86.6 Pyrolysis rate of the solid fuel along the surface; (a) 0 s, (b) 120 s, (c) 160 s, (d) 200 s Fig. 86.5 Temperature and heat flux distributions on the fuel surface; (a) 0 s, (b) 120 s, (c) 160 s, (d) 200 s

reasonable quantitative estimations have been reached and, which is most remarkable, coinciding to well-known recognized measurements. Estimating the moving of the

pyrolysis gasification point (TP ¼ 615 K), we got the spread rate to be about 2–5 mm/s for upward flame propagation. In fact, 2D simulation has been proceeded here, realizing that the need of full-scale flow resolution is to be finalized by the promoting 3D modeling.

848 Acknowledgments The study has been supported by the Russian Foundation for Basic Research (Project No. 16-08-00110).

References 1. Orloff L, De Ris J, Markstein GH (1975) Upward turbulent fire spread and burning of fuel surface. In: Proceedings of the 15th symposium (international) on combustion. pp 183–192 2. Ahmad T, Faeth GM (1979) Turbulent wall fires. In: Proceedings of the 17th symposium (international) on combustion. pp 1149–1160 3. Hasemi Y (1984) Experimental wall flame heat transfer correlations for the analysis of upward flame spread. Fire Sci Technol 4 (2):75–90 4. Saito K, Quintiere JG, Williams FA (1985) Upward turbulent flame spread. Fire Safety Science – Proceedings of First International Symposium. pp 75–86 5. Ito A, Kashiwagi T (1988) Characterization of flame spread over PMMA using holographic interferometry sample orientation effects. Combust Flame 71:189–204 6. Delichatsios MA, Delichatsios MM, Chen Y, Hasemi Y (1995) Similarity solutions and applications to turbulent upward flame spread on noncharring materials. Combust Flame 102:357–370 7. Tsai K.-C, Turnbull J, Will G, Drysdale D (2003) Upward flame spread: heat transfer to the unburned surface. Fire Safety Science – Proceedings of Seventh International Symposium. pp 117–127 8. Gollner MJ, Huang X, Cobian J, Rangwala AS, Williams FA (2013) Experimental study of upward flame spread of an inclined fuel surface. Proc Combust Inst 34:2531–2538 9. Leventon IT, Stoliarov SI (2013) Evolution of flame to surface heat flux during upward flame spread on poly(methyl methacrylate). Proc Combust Inst 34:2523–2530 10. Novozhilov V (2001) Computational fluid dynamics modeling of compartment fires. Prog Energy Combust Sci 27:611–666 11. Pizzo Y, Consalvi JL, Porterie B (2009) A transient pyrolysis model based on the B-number for gravity-assisted flame spread over thick PMMA slabs. Combust Flame 156:1856–1859 12. Rauwoen P, Degroote J, Wasan S, Vierendeels J, Merci B (2010) Simulations of upward flame spread by coupling a pyrolysis model with a CFD calculation. V European Conference on Computational Fluid Dynamics ECCOMAS CFD, Lisbon. 14–17 June 2010 13. Xie W, DesJardin PE (2009) An embedded upward flame spread model using 2D direct numerical simulations. Combust Flame 156:522–530 14. Moin P, Mahesh K (1998) Direct numerical simulation: a tool in turbulence research. Annu Rev Fluid Mech 30:539–578 15. Lipanov AM, Kisarov YF, Klyuchnikov IG (1999) Theoretical investigation of parameters for turbulent subsonic flows in

A. Karpov et al. compressible media: method and certain results. Dokl Phys 44 (6):380–384 16. Perzon S, Davidson L (2000) On transient modeling of the flow around vehicles using the Reynolds equations. ACFD 2000:720–727 17. Ren N, Wang Y, Trouve´ A (2013) Large eddy simulation of vertical turbulent wall fires. Procedia Engineering – 9th Asia-Oceania Symposium on Fire Science and Technology 62. pp 443–452 18. Gernamo M, Piomelli U, Moin P, Cabot W,H (1991) A dynamic sub-grid scale eddy viscosity model. Phys Fluids A(3):1760–1765 19. Sagaut P (2006) Large eddy simulation for incompressible flows. Springer, N.Y., 556 p 20. Volkov KN, Emel’yanov VN (2008) Turbulent flows in channels with injection. Results of large eddy simulation and the two-equation turbulence model. Fluid Dyn 43(4):573–582 21. Strelets M (2001) Detached-eddy simulation of massively separated flows. Proc. Meet. And Exhib. – The Thirty Ninth AIAA Aerosp. Sci., 18 p 22. Menter FR, Kuntz M, Langtry R (2003) Ten years of industrial experience with the SST turbulence model. Turb Heat Mass Transf 4:625–632 23. Spalart PR, Deck S, Shur ML, Squires KD, Strelets MK, Travin A (2006) A new version if detached eddy simulation, resistant to ambiguous grid densities. Theor Comp Fluids Dyn 20(3):181–195 24. Zeng WR, Li SF, Chow WK (2002) Review on chemical reactions of burning poly(methyl methacrylate) PMMA. J Fire Sci 20:401–433 25. Fureby C, Lofstrom C (1994) Large-eddy simulations of bluff body stabilized flames. The Combustion Institute – The Twenty-Fifth Symposium (International) on Combustion. pp 1257–1264 26. Versteeg HK, Malalasekera W (2007) An introduction to computational fluid dynamics: the finite volume method. Pearson Education, Harlow, England, 517 p 27. Modest MF (2003) Radiative heat transfer. Academic Press, San Diego, CA, 842 p 28. Jasak H (1996) Error analysis and estimation for the finite volume method with applications to fluid flows. PhD Thesis, Imperial College, University of London, 394 p 29. Issa RI (1985) Solution of the implicitly discretized fluid flow equations by operator-splitting. J Comp Phys 2:40–65 30. Bhattacharjee S, King MD, Paolini C (2004) Structure of downward spreading flames: a comparison of numerical simulation, experimental results and a simplified parabolic theory. Combust Theory Model 8:23–39 31. Wu KK, Fan WF, Chen CH, Liou TM, Pan IJ (2003) Downward flame spread over a thick PMMA slab in opposed flow environment: experiment and modeling. Combust Flame 132:697–707 32. Ayani MB, Esfahani JA, Sousa ACM (2007) The effect of surface regression on the downward flame spread over a solid fuel in a quiescent ambient. Therm Sci 11:67–86

Part XXII Suppression

Validating the Function of Absorber Plates for Auto-sprinkler System Activation

87

Kuang-Chung Tsai, Yukio Yamauchi, and Ken Matsuyama

Abstract

Building codes in many countries require an absorber plate with an auto-sprinkler system to accelerate sprinkler system activation when this system cannot be installed on a ceiling. This study evaluated experimentally the function of absorber plates. A sprinkler with or without an absorber plate was installed in a room measuring 4.3(L)  3.0(W)  3.3 (H) m. Hanging walls 1.1 m long were installed at the edges of the room ceiling to help form a hot smoke layer during a fire. Heptane and methanol into circular pans with diameters of 20–60 cm were the fire sources. The fire sources were positioned at the room center and 0.5, 1, and 1.5 m from the room center to produce fire plumes that reached both the sprinkler head and absorber plate, directly reached the absorber plate but not the sprinkler head, or reached neither the sprinkler head nor absorber plate, respectively. Temperature and velocity, two parameters important to sprinkler activation, were measured near the sprinkler head and compared for cases with and without an absorber plate. Experimental data indicate that the absorber plate did not help accelerate sprinkler head activation from temperature and velocity measurements for all cases with and without an absorber plate. Keywords

Detection  Activation  Absorber plate  Fire code

87.1

Introduction

After a fire starts in a compartment, a buoyancy-driven fire plume forms, and hot gases in the fire plume rise above the burning fuel and impinge on a ceiling. The ceiling causes the flow to turn and then move horizontally near the ceiling. When the flow reaches vertical barriers, such as walls, hot gases start accumulating below the K.-C. Tsai (*) Department of Safety, Health and Environmental Engineering, National Kaohsiung First University of Science and Technology, 2 Juoyue Rd, Kaohsiung 811, Taiwan e-mail: [email protected] Y. Yamauchi  K. Matsuyama Center for Fire Science and Technology, Tokyo University of Science, Tokyo, Japan

ceiling, generating a hot smoke layer. Many studies [1– 3] have examined the behaviors of fire plumes and ceiling jets, as both are important when dealing with issues related to fire detection and suppression [4–8]. Heskestad [1] listed the relationships between plume radius, mean excess temperature, and mean velocity at the centerlines of fire plumes; Alpert [2] presented the correlations between temperature and the velocity distributions of ceiling jets [2]. Temperature and velocity are the two key thresholds for activating an auto-sprinkler system. The response time index (RTI), which is related to the temperature, velocity, and other sprinkler sensor properties, is a measure of sprinkler sensitivity. To accelerate the activation of auto-sprinkler system during a fire, various related facilities have been developed.

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_87

851

852

K.-C. Tsai et al.

head, consequently affecting sprinkler activation. Experiments are designed to validate the function of an AP for auto-sprinkler activation during a fire. The extinguishment performance of the auto-sprinkler system was not discussed.

87.2

Fig. 87.1 A practical case of using an AP

This study evaluates absorber plates (APs). According to building codes, such as those in Japan, Taiwan, and China, auto-sprinkler systems request an AP to accelerate the activation of auto-sprinkler system when a system cannot be attached on a ceiling. Figure 87.1 shows an example in which water pipes obstructed heat from a fire from reaching sprinkler heads directly when the sprinkler heads were attached on a ceiling. Moreover, water drops may not reach fires due to obstructions after sprinkler system activation. The “Standard for Installation of Fire Safety Equipments Based on Use and Occupancy” [9] Article 47 in Taiwan’s Fire Code codifies the use of APs. Where the distance between the deflector of the nozzle and the ceiling or floor slabs is more than 30 cm, an AP shall be equipped pursuant to the following provisions: (1) The AP shall be made of metal materials, and the diameter shall be not less than 30 cm. (2) The distance between deflector and AP shall be not more than 30 cm. However, based on fire dynamics, heat flowing with a fire plume reaches an AP and a sprinkler head via three scenarios (Fig. 87.2). Figure 87.2a presents the first scenario in which a fire plume centerline reaches a sprinkler head without obstructions when a fire occurs just below the sprinkler head. Figure 87.2b shows the second scenario in which the fire plume centerline does not reach a sprinkler head and AP, but the fire plume reaches an AP. Heat flowing from the fire plume travels to the sprinkler head along the lower surface of the AP. Finally, in the third scenario, when a fire does not occur near below an AP and sprinkler head, the fire plume does not reach the AP and sprinkler head directly. The heat that reaches a sprinkler head is from the hot layer that forms below a ceiling. Additionally, in the first and second scenarios, the presence of an AP influences the flow velocity around a sprinkler

Experimental Design

Figure 87.3 shows a schematic diagram of the experiment. A fire source was placed on the floor of a room 4.3(L)  3.0(W)  3.3(H) m. Noncombustible curtains 1.1 m in length made of 10-mm-thick gypsum board were installed at the edges of the room ceiling to facilitate the formation of a smoke layer beneath the room ceiling during a fire [10]. A sprinkler head with or without an AP was located at the ceiling center in a room with a water pipe. The metal AP was 40 cm in diameter and positioned 30 cm below the ceiling. Four thermocouples (T11–T14) were installed at an interval of 10 cm moving radially from the sprinkler head at 5 cm below the ceiling to investigate the flow behavior of ceiling jets. A thermocouple tree (TR1-TR15) with a thermocouple interval of 5, 20, or 30 cm was utilized to measure room temperature. Figure 87.4 shows the experimental setup around the AP and sprinkler head. Thermocouples T1–T7 measured the temperature of, and around, the sprinkler head. Thermocouple T8 recorded the temperature rise of the water pipe. As an AP influences the fire plume around the water pipe, the pipe may receive less heat from the fire [8] than when the AP is absent. The reduction in the amount of heat reaching the water in the pipe increases the temperature difference between the sprinkler head and the water. More heat can conduct to the water from the sprinkler head, and sprinkler head activation may be consequently delayed. Thermocouples T9 and T10 were used to measure the temperature rise of the AP and were not used for scenarios without an AP. A bi-directional probe was installed near the sprinkler head to measure the horizontal velocity of flows in scenarios with an AP and near the ceiling in scenarios without an AP. Sprinkler head activation was simulated not using glass bulbs but temperature and velocity readings. These readings were used for further flow and thermal analyses. The primary fuel used was heptane; methanol was used in several cases for comparison. Table 87.1 lists the fuel type, diameter of fuel pan, and fuel position from the room center in the experiment. The fuel was poured into 2-cm-deep circular pans with diameters of 20, 40, 50, and 60 cm. Fuel depth was fixed at 1.5 cm.

87

Validating the Function of Absorber Plates for Auto-sprinkler System Activation

Fig. 87.2 Three scenarios for a fire plume to reach an AP and a sprinkler head

Fig. 87.3 Schematic diagram of the experimental design

Fig. 87.4 Detailed design of experimental setup around the AP and sprinkler head

853

854

K.-C. Tsai et al.

Table 87.1 The fuel type, diameter of fuel pan, fuel position from the room center in the experiment, condition of fire plume, and corresponding scenario. 0.2 + 0.2 m denotes that two pans with diameter of 0.2 m were used Fuel Heptane

Fuel pan diameter 0.2 + 0.2 m

Fuel position Center

0.5 m 1m 1.5 m 0.4 m

Center

0.5 m 1m 1.5 m 0.5 m

Center

1m Methanol

0.6 m

Center

1m 1.5 m

87.3

Fire plume condition Plume centerline reached AP Plume reached AP Plume did not reach AP Plume did not reach AP Plume centerline reached AP Plume reached AP Plume did not reach AP Plume did not reach AP Plume centerline reached AP Plume reached AP Plume centerline reached AP Plume reached AP Plume did not reach AP

Corresponding scenario (a)

were higher than those with an AP because the fire plumes can reach the water pipe without an AP. Additionally, the heat conducted through water with an AP was higher than that without an AP. This experimental finding was partly caused by conduction via water which influences sprinkler head activation, as Yamauchi et al. [11] demonstrated in his numerical analysis.

(b) (c) (c) (a)

(b) (c) (c) (a)

(b) (a)

(b) (c)

Results

87.3.1 Experimental Results This study used heptane and methanol as fuels, and the following experimental results from tests using the two fuels were consistent. After igniting the fuel, a fire started and a fire plume formed. A hot smoke layer then accumulated below the ceiling. After the hot smoke layer descended below the lower edge of the gypsum curtain, the hot smoke started flowing out of the experimental compartment. According to experimental observations, the T4 and T6 temperature readings were very close in all tests with an AP. This observation is consistent with what occurs in the stagnation region of ceiling jets [2]. The flow behavior below an AP resembled that of ceiling jets. The T8 readings decreased as the distance between the fire source and room center increased. The T8 readings in tests without an AP

87.3.2 Fire Plume Scenarios As described in Fig. 87.2, a fire plume interacts with an AP and sprinkler head in three scenarios in this study. However, assigning each test to one of the three scenarios by eye observation during this experiment was difficult because the fires were not very sooty. Therefore, four methods according to measurements were developed. Figure 87.5 shows a schematic of these four methods. • Method 1: When T12 readings were lower than those of T4 and T6, plume centerline reached the AP. When T12 readings were slightly higher than those of T4 and T6, plumes reached the AP, but plume centerlines did not reach the AP. When T12 readings were markedly higher than those of T4 and T6, fire plumes did not reach the AP. • Method 2: When T10 readings increased very rapidly, plume centerlines reached the AP. When T10 readings increased gradually, plumes reached the AP, but plume centerlines did not reach the AP. When T10 readings increased slowly, fire plumes did not reach the AP. • Method 3: When T4 readings increased suddenly and were higher than those of TR4 (thermocouple tree temperature at same height), plumes reached the AP. When T4 readings were lower than TR4 readings, plumes did not reach the AP. • Method 4: When T8 readings increased very rapidly, plume centerlines reached the AP. When T8 readings increased gradually, plumes reached the AP, but plume centerlines did not reach the AP. When T8 readings increased slowly, fire plumes did not reach the AP. Notably, the above methods must be considered together to group the tests. Table 87.1 lists the groups to which the conditions of the fire plumes interacting with the AP and sprinkler head belong based on the four methods.

87.4

Discussions

This study discusses the function of an AP for sprinkler head activation. Temperature and velocity are the two most important parameters. Figure 87.6 shows the three possible

87

Validating the Function of Absorber Plates for Auto-sprinkler System Activation

855

Fig. 87.5 Four methods to group the behaviors of fire plumes

from ceiling jets. The temperatures of ceiling jets were higher than those of the hot smoky layer accumulated at 30 cm below the ceiling. Therefore, the AP in this scenario (Fig. 87.2b) did not accelerate the activation of the sprinkler system. Furthermore, when a fire occurred that was not below a sprinkler head (Fig. 87.2c), T1 readings were much lower than those of T12 (data not shown). The AP in this situation (Fig. 87.2c) did not enhance the heat transfer to the sprinkler system.

Fig. 87.6 Three possible sprinkler designs in this study (a) With AP (b) Without AP (c) Sprinkler sensor below ceiling.

sprinkler designs addressed in this study. By comparing temperature and velocity in Fig. 87.6a, b, the influence of an AP on activation of a sprinkler head located 30 cm below a ceiling is evaluated. By comparing the readings in Fig. 87.6a, c, the function of an AP in enhancing sprinkler system effectiveness is assessed.

87.4.1 Temperature Figure 87.7 shows the temperatures at the three positions corresponding to the three possible sprinkler designs (Fig. 87.6) with 0.2 + 0.2-m heptane fires located at the center and 0.5 m from the room center. Clearly, when the fires were located in the room center and the plume centerlines of the fires directly reached the AP (Fig. 87.2a), the AP did not increase the heat received by the sprinkler head (Fig. 87.7a). Figure 87.7b shows the temperature histories at different locations corresponding to the three possible sprinkler designs (Fig. 87.6) with the 0.2 + 0.2-m heptane fires located 0.5 m from the room center. The plumes directly reached the AP but not the sprinkler head. The presence of the AP did not increase T1 temperature. The T12 temperatures were highest in the three possible sprinkler designs (Fig. 87.6). The heat thermocouple T1 received was from the hot smoky layer that accumulated below the ceiling while T12 received heat

87.4.2 Velocity Figure 87.8 shows the velocity measurements at two locations (v1 and v3 in Fig. 87.6) corresponding to two of the three possible sprinkler designs (Fig. 87.6a, c) with 0.2 + 0.2-m and 0.4-m heptane fires located at the room center and 0.5, 1, and 1.5 m from the room center. Experimental results show that the plume velocity decreased as the distance between the fires and the room center increased. The v1 readings were very close to v3 readings when plume centerlines directly reached the sprinkler head. However, v1 readings were smaller than those of v3 in the other two scenarios (Fig. 87.2b, c). Furthermore, the velocity v2 (Fig. 87.6b) was not measured in this study.

87.4.3 Sprinkler System Activation Table 87.2 compares temperature and velocity measurements near the sprinkler head corresponding to the three possible sprinkler designs (Fig. 87.6) and sprinkler activation assessed by the temperature and velocity comparison. When the plume centerlines directly reached the sprinkler head (Fig. 87.2a), the sprinkler systems activated almost simultaneously. When the plume centerlines did not directly reach a sprinkler head but reached an AP (Fig. 87.2b), the sprinkler system installed close to the ceiling (Fig. 87.6c) activated fastest among the three possible sprinkler designs. When the sprinkler system was equipped 30 cm below the ceiling, the AP did not cause differences in the temperature

856

K.-C. Tsai et al.

Fig. 87.7 Temperatures at different locations corresponding to the three possible sprinkler designs in Fig. 87.6 for 0.2 + 0.2 m heptane fires located at the room center (a) and 0.5 m from the room center (b)

Fig. 87.8 Velocity measurements v1 and v3 in Fig. 87.6 for 0.2 + 0.2-m and 0.4m in diameter heptane fires located at the room center and 0.5, 1, and 1.5 m from the room center. The corresponding scenarios described in Fig. 87.2 are addressed

and velocity near the sprinkler head and reduce consequent time to sprinkler activation. Furthermore, the AP did not accelerate the activation of the sprinkler system when a fire occurred far from the sprinkler system (Fig. 87.2c).

87.5

Conclusions

1. All cases can be grouped into three scenarios from the perspective of sprinkler activation and according to the interaction between a fire plume and sprinkler. The three scenarios are as follows: (a) The fire plume centerline reaches the sprinkler head and AP. (b) The fire plume reaches the AP but the plume centerline did not reach the sprinkler head or AP.

(c) The fire plume does not directly reach the sprinkler head and the absorber. 2. The temperatures of the sprinkler head and that below the AP (T4 and T6) were very close in all the cases when fire plumes reached the AP, demonstrating that ceiling jets formed below the AP. 3. The temperature of water pipe (T8) decreased when the distance between the room center and fire source increased. The T8 readings in cases without an AP were higher than those with an AP. Thus, conduction loss of heat via water in the cases with an AP was more than that without an AP. 4. From temperature and velocity measurements for all cases with and without an AP, the AP did not help accelerate sprinkler head activation. Thus, APs are not

87

Validating the Function of Absorber Plates for Auto-sprinkler System Activation

857

Table 87.2 Comparison of temperature, velocity, and time to sprinkler activation corresponding to the three possible sprinkler designs in Fig. 87.6 Sprinkler design

With AP Without AP

Fire position relative to sprinkler head Scenario Temperature Fig. 87.2a Velocity (direct hit of Sprinkler plume) activation Scenario Temperature Fig. 87.2b Velocity (partial hit of Sprinkler plume) activation Scenario Temperature Fig. 87.2c Velocity (no hit of Sprinkler plume) activation

Sprinkler sensor below ceiling

Almost equala Almost equal Almost equal

Almost equal Almost equal Almost equal

Almost equal Almost equal Almost equal

Almost equal Almost equal Almost equal

Almost equal Almost equal Almost equal

High High Fast

Almost equal Almost equal Almost equal

Almost equal Almost equal Almost equal

High High Fast

a

“Almost equal” means the values of parameters associated with the designs are very close

useful! Even when a sprinkler head is installed 30 cm below a ceiling, APs are not useful!

Acknowledgment The authors would like to appreciate the financial support from the International Young Researcher Scholarship Program, Center for Fire Science and Technology, Tokyo University of Science.

References 1. Heskestad G (2000) Fire plumes, flame height, and air entrainment. In: DiNenno PJ (ed) The SFPE handbook of fire protection engineering, 3rd edn. National Fire Protection Association, Quincy, 02269, pp 2/1 2. Alpert R (2000) Ceiling jet flows, the SFPE handbook of fire protection engineering. In: DiNenno PJ (ed) The SFPE handbook of fire protection engineering, 3rd edn. National Fire Protection Association, Quincy, 02269, pp 2/18 3. Oka Y, Imazeki O, Sugawa O (2010) Temperature profile of ceiling jet flow along an inclined unconfined ceiling. Fire Saf J 45:221–227

4. Ishii H, Ono T, Yamauchi Y, Ohtani S (1991) An algorithm for improving the reliability of detection with processing of multiple sensors’ signal. Fire Saf J 17:469–484 5. Motevalli V, Marks CH (1987) Measurement of velocity and temperature profiles in low-speed, turbulent non-isothermal flows. National Bureau of Standards Report NBS-GCR-87-535, Gaithersburg. pp 3 6. Jones W (2011) Implementing high reliability fire detection in the residential setting. Fire Technol. doi:10.1007/s10694-010-0211-8 7. Gritzo LA, Bill R, Wieczorek C, Ditch B (2011) Environmental impact of automatic fire sprinklers: Part 1. Residential sprinklers revisited in the age of sustainability. Fire Technol 47:751–763 8. Wieczorek CJ, Ditch B, Bill R (2011) Environmental impact of automatic fire sprinklers: Part 2. Experimental study. Fire Technol 47:765–779 9. The “Standard for installation of fire safety equipments based on use and occupancy” (2011) Taiwanese Regulation, Taipei, Taiwan 10. Husted BP, Holmstedt G (2008) Influence of draft curtains on sprinkler activation – comparison of three different models. J Fire Prof Eng 18:29–54 11. Yamauchi Y, Mammoto A, Dohi M, Ebata H, Morita M (2005) A calculation method for predicting heat and smoke detector’s response. Fire Sci Technol 2:179–210

An Experimental Study on the Smoke-Logging Phenomenon Using Sprinkler for Performance-Based Evacuation Safety Design

88

Dong-Goo Seo, Ung-Gi Yoon, In-Hyuk Koo, Bong-Chan Kim, Dong-Eun Kim, Ken Matsuyama, and Young-Jin Kwon

Abstract

The purpose of this study is to investigate the descending air current of a smoke layer for smoke logging during sprinkler operation. The results could be used for performance-based evacuation safety design. Smoke layers were investigated experimentally using particle image velocimetry (PIV) and were analyzed according to the size of the fire source and the type of sprinkler head. As a result, a relationship was obtained between the mean droplet diameter, droplet velocity, and spray distribution for each sprinkler head. In addition, the descending air current of smoke was studied as a function of the fire source size and the velocity of the descending air current. It was predicted by the regression equation that the smoke layer moved to the bottom layer when the descending air current speed reached 0.6 m/s and confirmed that a stable descending air current was generated only when the fire source size exceeded 100 kW. Keywords

Sprinkler
PIV system
Smoke logging
Spray droplet
Descending air current velocity

Nomenclature Ad B CD D0 g ΔHC mb md n ni Q Re

Cross-sectional area (m2) Buoyancy force of the smoke layer (N/m) Drag coefficient (-) Resistance force for the drag water droplets (N) Gravitational acceleration (m/s2) Heat of combustion (kJ/g) Burning rate (g/s) Water droplet mass (kg) Particles per unit volume number (-/m3) Particle diameter (μm) Heat release rate (kW) Reynolds number (-)

D.-G. Seo
U.-G. Yoon
I.-H. Koo
B.-C. Kim
D.-E. Kim
Y.-J. Kwon (*) Hoseo University, 165, Sechul-ri Baebang-Eup, Asan, Chungnam 336-795, South Korea e-mail: [email protected]

r Ts T0 vd W x xi

Radius of the water droplet (m) Smoke layer temperature (K) Lower layer temperature (K) Vertical particle velocity (m/s) Sprinkler watering spray amount (kg/m2∙s) Sauter mean diameter (μm) Number of particles (-)

Greek Symbols μ ρ ρs

Viscosity (Pa∙s) Smoke layer density (kg/m3) Smoke layer air density (kg/m3)

K. Matsuyama Tokyo University of Science, 2641 Yamazaki, Noda-shi, Chiba 278-8510, Japan

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_88

859

860

88.1

D.-G. Seo et al.

Introduction

A complex approach correlated by fire behavior, evacuation behavior characteristics, and firefighting facilities is needed for the performance-based fire safety design of a building. It is an important factor to predict the fire behavior and evacuation behavior in accordance with the firefighting facilities, as well as to predict the safe evacuation time based on behaviors of heat and smoke through fire behavior analysis. In particular, sprinkler equipment is a representative waterbased firefighting equipment to suppress the force of the fire in the early stage by automatically detecting the fire. In addition to the early removal of the initial fire source, sprinkler facilities are recently used actively for performancebased control, that is, the performance-based design (PBD) that uses mathematical calculations such as Fire Dynamics Simulator (FDS) model developed by National Institute of Standards and Technology (NIST) [3, 14]. Korean studies on sprinkler equipment have been performed since the 1990s; studies to improve the performance of the sprinkler equipment include the ones by Lee et al. [9] and Kim et al. [4–6] on the operation time prediction according to the response time index (RTI). In addition, Chu et al. [2], Park [15], and Kim et al. [7] have studied the methods that can be used in PBD as well as the construction of the database in consideration of the characteristics of sprinkler head spray of droplets. Korean studies tend to focus on the performance improvement of the initial fire control and utilization of PBD. Sprinklers spray water potentially into compartments where flames do not exist due to its automatic operation during fire and therefore could directly spray water to the smoke in rooms adjacent to the fire, diffusing the smoke and adversely descending the smoke by the sprayed water droplets. Such a phenomenon is called the smoke logging which negatively influences evacuation safety and firefighting activities. Smoke-logging phenomenon has been proposed by Bullen [1], who presented a formula for the relationship between smoke layer buoyancy and the resistance of the water droplets on the smoke layer. Afterward, Li et al. [10] and Zhang et al. [16] reviewed the formula proposed by Bullen [1] and proposed prediction equations for the descending air current caused by water droplets in the smoke layer. Recently, Tsuchiya [13] performed a research to derive prediction formula for the air current of the smoke layer by observing the watering pressure (MPa) watering amount (L/min) spray droplet characteristics of sprinkler head accurately with PIV (particle image velocimetry) equipment. However, Korean studies on smoke-logging phenomenon are rare, therefore are very necessary for PBD. The purpose of this study is to build the basic data for future PBD using the sprinkler equipment, by reviewing the spray droplet characteristics for Korean sprinkler head for the smoke

layer during sprinkler spray using PIV equipment and analyzing the descending air current of the smoke layer during sprinkler spray in accordance with the fire size.

88.2

Experiment of the Sprinkler Spray Droplet

It is an important factor in analyzing smoke-logging phenomenon to identify the spray droplet characteristics (water droplet particle diameter, velocity, and spray and spray distribution) of the sprinkler head. Bullen [1], Li et al. [10], and Zhang et al. [16] derived the Re and CD of the resistance to air by using Newton’s laws of resistance for D0 (N) to the B (N) of the smoke layer due to the temperature rise for the descending air current of the smoke layer, by assuming the form of the water droplets as spherical. Further, according to Tsuchiya’s [13] experimental results of the average particle diameter and the speed of sprinkler, the form of the air flow is a transition region, allowing the calculation by assuming the Reynolds number as constant. According to Newton’s law of resistance, the vertical air drag D0 of one water droplet is shown in Eq. 88.1, 0

D ¼ kd v2d

ð88:1Þ

kd ¼ CD ρs Ad =2

ð88:2Þ

where

Assuming the motion of the air flow to be in the transition zone, CD can be expressed as Eqs. 88.3 and 88.4 using Re and μ.
 =Re CD ¼ 24 1 þ 0:15R0:687 ð88:3Þ e Re ¼ 2rvd ρs =μ

ð88:4Þ

The D to the air flow per unit volume may be expressed in the Eq. 88.5 as the product of the D0 and n which can be calculated using the W, md, and vd, as shown in Eq. 88.6: 0

D¼Dn

ð88:5Þ

n ¼ W=md vd

ð88:6Þ

In addition, B from thermal current may be represented as the product of the difference between ρ0, ρs, and g and can be calculated using the Ts and T0. B ¼ ðρ0  ρs Þg ¼ fðT s  T 0 Þ=T s gρ0 g

ð88:7Þ

In this section, accordingly, experiments were performed on the spray droplet characteristics of the water particles according to the respective smoke descending.

88

An Experimental Study on the Smoke-Logging Phenomenon Using Sprinkler for. . .

88.2.1 Test Methods and Conditions Four types of sprinkler heads that used Korean (top-down, open, residential, and flush type) were selected, for the manual spraying with the thermal part removed, in order to investigate the spray droplet characteristics sprayed from the heads. In addition, spraying was performed with constant watering pressure of 0.1 MPa according to NFSC 103; the specifications of the four types of SP heads are shown in Table 88.1. The whole experiment was conducted in a facility with the pump and drainage as shown in Fig. 88.1, with the entire compartment area (W6.5  D6.5  H2.7 m) Table 88.1 Specifications of sprinkler heads Sprinkler head Symbol Pressure (MPa) Watering spray amount (L/min)

Ttype T80 0.1 80

Otype O80 0.1 80

Rtype R50 0.1 50

Ftype F80 0.1 80

861

divided by calcium silicate board (W3.2  D3 .2  2.7 m). The head of the sprinkler was installed at the upper center, positioned 200 mm away from the ceiling. In addition, the watering range was constrained by installing a water bottle in the lower bottom of the sprinkler head in order to investigate only the area of the smoke layer because this experiment was aimed to examine the smoke movement in accordance with the fire source. The compartment space was installed with PIV camera, laser, and the acrylic plate for visual observation, in the form of window, with the size of the opening 0.9 m wide and 1.8 m high. For the measurement of the sprinkler particle diameter, the flash device was installed in an area with an acrylic window and a microscope lens within the compartment space, i.e., on the opposite side. It uses the method that the image is recorded as it passes the image-sensing surface at the moment of the scintillation of light-emitting device; valid particle diameter of the droplets can be obtained if the particle diameter analysis software (VisiSize Solo) is used for the recorded image because only the droplets with exact focuses are extracted. Figure 88.2 shows the measuring position of the particle diameters. The particle diameter data was analyzed by calculating the Sauter mean diameter. The Sauter mean diameter is most widely used in the organization of spray particle size by weighted averaging with the squaring of the needed particle diameter, defining that the total sum of the surface area is the same as the total amount. Sauter mean diameter x can be expressed as Eq. 88.8 where ni is the particle diameter and xi is the number of particles. X X x¼ ni x3i = ni x2i ð88:8Þ Measurement of the particle velocity of the sprinkler was performed using a PIV system. Measuring position was at the same as the particle diameter, with the measuring range defined as 200 mm (W)  150 mm (H). Tracer that

Fig. 88.1 Overview of experiment space. (a) Plan. (b) x-x0 section

Fig. 88.2 Measure location of particle size and velocity

862

D.-G. Seo et al.

Fig. 88.3 Measure method of spray amount

Table 88.2 Results of watering spray droplet List of measurement Sauter mean diameter(μm) Particle velocity(m/s) Spray amount(mL/min∙0.1 m2)

T80 602 2.35 200

O80 502 4.92 141

R50 642 4.99 175

F80 610 6.64 208

visualizes the behavior of the gas was not sprayed to measure the velocity of particle from the sprinkler heads. Figure 88.3 shows the spray distribution, where spray was performed to the area as wide as 3/4 of the total area, after placing 25 water bottles (0.32  0.32 m), specifying numbers A–F, at the bottom of the sprinkler head. A method for spray distribution test device is shown in the attached Fig. 88.2 sprinkler head type approval and product inspection technical criteria, but this experiment used different method from the standard because the spray distribution is uniform and is meant to derive only particle diameter and particle velocity for each watering spray amount.

88.2.2 Experimental Results and Analysis Table 88.2 shows the experimental results of the spray droplet characteristics for three items of particle diameter, particle velocity, and spray distribution for each sprinkler head. Discussion on the findings follows. 1. Particle diameter measurement About 3000 sheets of valid data were obtained for each sprinkler head. Sprinkler head-specific mean particle diameters were used to calculate the Sauter mean diameter, as described above. The mean particle diameter, except for the O80 head, appeared to be about 600 μm. 2. Particle velocity measurement

This study is aimed to measure the descending air current generated by the water droplets sprayed from the sprinkler head; measurements were derived by average frequency for vector component of the lower direction because it was only for the downward movement. The results show that F80 descends the fastest at 6.64 m/s and that the flow rate is about 5 m/s except for T80. The reason why T80 is measured lower than other three types of heads is the rapid decrease in the flow rate because of the large area where water droplets hit the deflector. 3. Spray distribution measurement The measured results of the sprinkler spray distribution: T80 and R50 show uniform shape of the distribution; O80 is biased toward the wall surface; and F80 shows that spray was conducted in the lower center portion sprinkler head. This implies that spray distribution varies depending on the type of deflectors; so, the average watering spray amount was obtained. 4. Discussions According to the sprinkler head-specific measurements of the spray droplet characteristics, the correlation between particle diameter, particle velocity, and watering spray amount was analyzed. As shown in Fig. 88.4a, the three sprinkler heads, except for R50, showed linear relationship between the watering spray amount and particle diameter. The reason why R50 shows lower watering spray amount than the other three sprinkler heads despite of its relatively large particle diameter is its lower watering pressure compared to the other types of sprinkler heads. In addition, Fig. 88.4b does not show a close relationship between the particle velocity and the mean particle diameter, but Fig. 88.4c shows the linear relationship between the particle velocity and the mean particle diameter for the three sprinkler heads other than T80. The reason is believed that the measurements were conducted for the sprinkler head that sprays water for the same duration of time, so that the faster flow rate means that the more water droplets flow in the water bottle. However, T80 shows a lower flow rate distribution when compared to the droplet diameter and watering spray amount, which is due to the absence of valid data during the removal process of the error vectors. This problem requires further analysis.

88.3

Experiment of Smoke Layer Descending Air Current According to Spray Droplets

88.3.1 Experimental Methods and Conditions This experiment was carried out by installing PIV measurement equipment in the same space as the spray droplet characteristic experiments. Comparison was made for the generating status of the descending air current by a

88

An Experimental Study on the Smoke-Logging Phenomenon Using Sprinkler for. . .

863

Table 88.3 Experimental conditions Case 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Fig. 88.4 Correlation with spray droplet by sprinkler heads. (a) Sauter mean diameter and spray amount. (b) Sauter mean diameter and particle average velocity. (c) Sauter mean diameter and particle average velocity

temperature difference between the smoke layers, followed by the comparison of the generating status of

Fuel basket (m2) 0.1

Type of fuel Ethanol

n-Heptane

0.25

Ethanol

n-Heptane

Sprinkler T80 O80 R50 F80 T80 O80 R50 F80 None T80 O80 R50 F80 T80 O80 R50 F80 None

the descending air current by a sprinkler head-specific performance. The situation of the smoke layer was compared according to the heat release rate (kW) and sprinkler head performance by changing the type of fuel (ethanol and n-heptane) and the fire source area (0.1 and 0.25 m2). The conditions for this experiment are shown in Table 88.3 This experiment has been made for the SL phenomenon; the formation of the smoke layer is the most important factor. Therefore, the amount of the fuel consumption was adjusted according to the fuel type and fire source size in order to obtain a sufficient combustion time for the formation of the smoke layer. Sprinkler heads are the same as the ones used for the experiment of the spray droplet characteristics; heat release rates in the compartment were measured by mass loss rate (kg/s) using the load cell. In addition, thermocouples were installed in two places of the compartment to measure the temperatures as a function of the height of smoke layer. The vertical temperature distribution was measured after installing 13 points with a height interval of 200 mm. In addition, the top (2.6 m) and bottom (0.2 m) CO2 concentration meters were installed on the thermocouple framework near the opening, in order to judge whether the smoke layer descent is generated. Measurement of the flow rate by PIV equipment was carried out through the acrylic windows in the wall and by CCD camera and a laser device installed outside of the compartment. The measurements were carried out for the two points of 0.2 m (W)  0.15 m (H) at the height of 1.30 and 1.45 m.

Fig. 88.5 Experimental overview. (a) z-z0 section. (b) x-x0 section. (c) y-y0 section. (d) xyz coordinate

88

An Experimental Study on the Smoke-Logging Phenomenon Using Sprinkler for. . .

Fig. 88.5 shows the overview of the installation of the measuring equipment for this. Experimental procedure was set that spraying did not start until the steady state of the temperature (initial temperature) in the compartment, but started 180 s after the breakout of the fire. Detailed experimental procedures are shown in Table 88.4. In addition, the valve adjustment was ready in advance to maintain a constant watering pressure (0.1 MPa) immediately after the spray. In addition, it is possible to obtain the movement of the air flow and the movement of the water droplets at the same time from the tracer using PIV equipment. However, it is difficult to identify only the speed of the airflow from the original image due to the simultaneous illustration of water droplets. For this, Photoshop CS5 was used to remove the droplets to obtain velocity vectors for each hour using dedicated software (Koncerto) from the data for the movement of the pure gas flow.

88.3.2 Experimental Results and Analysis Figure 88.6 shows the results of the heat release rate to set the conditions in the compartment. It was confirmed that the conditions of fire source within the compartment were kept at steady state until the 180th second when the spray started and that the heat release rate remained nearly constant for 150 s from the 60th second after ignition till 210th second when PIV measurement ends. Thus, the heat release rate was defined as the mean value during the 150 s, and the heat Table 88.4 Outline of experiment No. 1 2 3 4 5 6 7 8

Time –02:00 –01:00 00:00 02:00 02:50 03:00 03:30 04:40

Contents Operate a sprinkler pump Take a camera Ignition Operate a PIV system Watering Stability of pump pressure(0.1 MPa) Finish PIV system Extinguish

Fig. 88.6 Results of heat release rate. (a) Ethanol. (b) n-Heptane. *Thin line, 0.1 m2; thick line, 0.2 m2

865

release rate Q (kW) is calculated by multiplying the burning rate mb (g/s) of the fuel by the heat of combustion ΔHc (kJ/g) as shown in the Eq. 88.9 [8]. Q ¼ mb ΔH c

ð88:9Þ

Here, the heat of combustion of ethanol (C2H6) is 44.56 (kJ/g) and that of normal heptane (C7H16) is 26.82 (kJ/g) [13]. Table 88.5 shows the heat release rate in accordance with the conditions of the fire source. In addition, Fig. 88.7 is an example (T80, n-heptane 0.25 m2) showing the temperature in the compartment, illustrating that temperature is reduced in the upper layer right after the spraying (180th second) and the temperature of the lower layer rises after approximately 60 s. In other words, smoke layer is considered to be descended to the lower portion in the compartment. In addition, as shown in Fig. 88.8, it is found that the CO2 concentration in the lower compartment has risen for about 1 min after the spray, from the example (F80, n-heptane 0.25 m2) of the CO2 concentration (%) of the upper and lower sections. It shows more clearly that the smoke layer has descended than judged by the temperature. Regarding the relationship between CO2 concentration and descending air current by sprinkler head, CO2 concentration was reduced when the descending air current reached the speed of about 0.6 m/s or higher as shown in Fig. 88.9. Tsuchiya [13] reported that, after the experiment of the rise of CO2, the smoke layer descended when the speed of the descending air current was 0.4 m/s or over and that the CO2 concentration did not increase at the nozzles when the particle velocity was 2 m/s or less or the watering spray amount was 300 mL/min∙m2 or less and presented three step intervals (interval where descending of air current period does not occur, the transition interval, and interval where descending of air current is generated). According to the results of this study, the particle velocity and watering spray amount show the distribution above those values, so it is considered that the occurrence of descending air current is a matter of course and the result of regression analysis shows

866

D.-G. Seo et al.

Table 88.5 Results of HRR by fire sources Ethanol 0.1 m2 49.7 kW

n-Heptane 0.1 m2 98.6 kW

Ethanol 0.25 m2 138.5 kW

n-Heptane 0.25 m2 239.0 kW

Fig. 88.9 The relationship between the rise of CO2 concentration and descending air current velocity

Fig. 88.7 Results of temperature

Fig. 88.10 Descending air current velocity by size of fire sources

Fig. 88.8 Results of CO2 concentration (%)

that the air current speed is 0.6 m/s somewhat higher than 0.4 m/s, still similar results. In addition, Fig. 88.10 shows the respective average descending air current according to the analysis result of the descending air current by the sprinkler head for each size of the fire source. It is difficult to believe that the distribution of the descending air current speed for each sprinkler head is related to the water droplets when the fire source size is relatively small (0.1 m2), but it was confirmed that the descending air current by the water droplets injected from the SP head descends relatively stable if the fire source

size is large (0.25 m2), namely, 100 kW or larger. This means that the fire source size should be 100 kW or more to analyze the smoke-logging phenomenon by the effect of the water droplets. That is, it is believed that a certain degree of the thickness and buoyancy of the smoke layer must exist within the compartment.

88.4

Conclusions and Future Research Directions

The following findings were obtained from this study as a result of the experimental investigation conducted by means of PIV for the smoke logging during sprinkler spray

88

An Experimental Study on the Smoke-Logging Phenomenon Using Sprinkler for. . .

activation. The objective is to provide data for performancebased evacuation safety design. 1. As a result of the investigation of the spray droplet characteristics by sprinkler heads to establish basic data to closely identify the smoke-logging phenomenon, valid data were obtained on the mean particle diameter, particle velocity, and spray distribution. In addition, it was found that the water spray mass and particle diameter had a linear relationship for three types of smoke-logging heads other than R50 (residential) head. 2. The analysis of the descending current of smoke identified by the PIV equipment and temperature and CO2 concentration measurements shows that the descending air current reaches to the bottom right after the spray and the smoke layer descends when the velocity reaches 0.6 m/s or faster. There is a relationship between CO2 concentration and descending air current. In addition, it was confirmed that stable descending air currents are generated only when the fire source is 100 kW or larger. This study was aimed to investigate the basic characteristics for analyzing the SL phenomenon; further research is deemed necessary on the model that may predict descending air current entering the lower layer, by analyzing the resistance to water droplets as well as calculating the buoyancy of the smoke on the basis of the basic data for the performance-based evacuation safe design. Acknowledgments This work was supported by grants of the National Emergency Management Agency that showcase our Fire Safety Technology Development (“NEMA-Next Generation-2011-30”). This work was supported by the Center for Fire Science and Technology, Research Institute for Science and Technology, Tokyo University of Science.

867

References 1. Bullen ML (1977) The effect of a sprinkler on the stability of a smoke layer beneath a ceiling. Fire Technol 13(1):21–34 2. Chu BG, Choi JW, Cha KS (2001) The study on characteristics of water spray and droplets from fire sprinkler head. J KIIS 16(2):13–21 3. Kim SC (2012) A numerical study of the effect of sprinkler spray on the flow characteristics induced by fire. J Korea Inst Fire Sci Eng 26(5):105–110 4. Kim DS, Park YI, Park HY (1993) An experimental study on the responsiveness of sprinkler heads. J Korea Inst Fire Sci Eng 7(2):3–12 5. Kim MB, Han YS (1996) Prediction of sprinkler activation time in compartment fire. J Korea Inst Fire Sci Eng 10(4):13–18 6. Kim JH, Kim WH, Lee SK (2001) A comparison of the prediction of sprinkler response time applying fire models. J Korea Inst Fire Sci Eng 15(2):46–52 7. Kim SC, Lee SW, Park WJ (2009) A sensitivity study of the number of parcels to the numerical simulation of sprinkler sprays. J Korea Inst Fire Sci Eng 23(1):48–54 8. Kwon YJ et al. (2012) Fire dynamics for PBD, Dong hwa Publisher, pp. 261–271 9. Lee BK, Tae SH (1991) A study on response time index and operating time for fusible link sprinkler head. J KIIS 6(4):34–44 10. Li KY, Hu LH, Huo R, Li YZ, Chen ZB, Li SC, Sun XQ (2009) A mathematical model on interaction of smoke layer with sprinkler spray. Fire Saf J 44(1):96–105 11. Li KY, Li MJ (2011) Simplified calculation method for determining smoke downdrag due to a sprinkler spray. Fire Technol 47 (3):781–800 12. NFSC 103 (2014) National fire safety code for sprinkler equipment. Notification of Ministry of Public Safety and Security 13. Tsuchiya M, Matsuyama K (2012) A study on smoke behavior affected on droplets of sprinkler, velocities of the droplets and smoke layer by PIV system, Ph. D. degree of Tokyo University of Science 14. Park WC, Lee MG, Park HS (2006) Simulation of a clean room fire II. Need of smoke control system and sprinkler system. J Korea Inst Fire Sci Eng 20(2):8–13 15. Park YH (2004) Directional water spray characteristics of sprinkler heads. J Korea Inst Fire Sci Eng 18(4):35–41 16. Zhang CF, Chow WK, Huo R, Zhong W, Li YZ (2010) Experimental studies on stability of smoke layer with a sprinkler water spray. Exp Heat Transf 23(3):196–216

Effect of Fire Detection Function on Fire Suppression in Home Stay Facilities in Taiwan

89

Chung-Hwei Su, Kuang-Chung Tsai, Ming-Hui Dai, and Chun-Chou Lin

Abstract

Home stay facility has been one of the main options of lodging in Taiwan. Besides the lodging supplies, the decorative articles inside the buildings are indeed inflammable. The home stay facilities are mostly located in remote scenic areas, where the firefighters cannot reach immediately in case of a fire. How to detect and extinguish the fire immediately at the initial stage of fire is an important topic for the home stay facility. Many studies have discussed the firefighting in large hotels, but seldom on firefighting in home stay facilities. Different detector types have varied detection lengths, thus affecting the timeliness of fire suppression. Referring to the “Regulations for the Management of Home Stay Facilities” enacted in 2001, this study analyzed the fire suppression timeliness of fire detectors. The Fire Dynamics Simulator was used to discuss the extent of harm of a fire on the persons attempting to suppressing fire. The changes of parameters, including temperature, visibility, and fire behavior parameters, due to dense smoke and fire over time were obtained. The activation time of two detectors, namely, smoke detector and rate of rise detector, in a case of fire was quantitatively analyzed. The results showed the scenarios the persons may be confronted with when suppressing a fire. Keywords

Home stay facility
Fire
Fire detector
Regulations for the Management of Home Stay Facilities
Fire Dynamics Simulator (FDS)

Nomenclature Cp D* G Q To

Specific heat (kJ/kg- C) Characteristic fire diameter Gravity acceleration (m/s2) Heat release rate (kW) Ambient temperature (K)

C.-H. Su (*)
K.-C. Tsai
M.-H. Dai Department of Safety, Health and Environmental Engineering, National Kaohsiung First University of Science and Technology, Kaohsiung, Taiwan e-mail: [email protected]; [email protected] C.-C. Lin Institute of Fire Science, WuFeng University, Chiayi, Taiwan

Greek Letters α ρ0

Fire growth parameter (kW/s2) Density of air (kg/m3)

89.1

Introduction

89.1.1 Popularity of the Home Stay Facility Industry in Taiwan In recent years, there is an increase of local and inbound tourists in Taiwan, including those from China, Japan, Korea, Europe, and the USA. Self-guided tour has also become a common mode of travel [1]. In terms of tourism

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_89

869

870

C.-H. Su et al.

Fig. 89.1 Turnover of home stay facilities in Taiwan

quality, safety is an important issue, and tourist safety is closely associated with tourism image. Many countries have paid considerable attention on maintaining travel safety because unsafe tourism quality will influence the tourists’ willingness of travel [2, 3]. When tourists are resting at lodging places, they tend to be less cautious, thus, ensuring the safety of lodging places is an important issue of tourist safety [4, 5]. Home stay facility is a popular lodging option for tourists in Taiwan. According to official statistics, there were 3309 home stay facilities, offering 13,946 guest rooms, in Taiwan, as of December 2009. The number of home stay facilities increased to 5565, with 23,182 guest rooms, as of November 2014 [6]. Moreover, in 2009, home stay facilities received 1,155,177 guests, with total revenue of 28.8 million dollars. In 2013, 2,504,553 guests were received, and the total amounted to 2282 million dollars. Figure 89.1 shows the increased number of home stay facilities.

There are many combustible materials in the rooms of home stay facilities, including wood, cotton fabrics, and foam decorations. The home stay facilities are mostly located in remote scenic areas where firefighters cannot reach immediately in case of fire. How to detect and extinguish fire at the initial stage is crucial. Different detector types affect the detection time, thereby influencing the effectiveness of disaster relief. Referring to the “Regulations for the Management of Home Stay Facilities,” this study aims to examine the fire safety in home stay facilities in using computer software for fire simulation. The results can be used to review the fire hazards to the home stay facilities.

89.2

The Existed Regulations for Fire Safety of Home Stay Facilities

89.2.1 Safety Management Measures 89.1.2 Significance of Early Fire Suppression for Home Stay Facility Fire At lodging places, the guests are mostly resting or sleeping, so they are less watchful of the fires. On November 21, 1980, the fire in MGM Grand Las Vegas killed 85 persons and injured about 700 persons; it is one of the most serious hotel fires in the American history [7]. Some studies have shown that the fire safety of hotels is of great importance [8, 9]. The local governments in Taiwan mandate all legally registered home stay facilities to comply with the “Regulations for the Management of Home Stay Facilities” enacted in 2001 [10]. Compared the regulation mandated for hotels in “Standard for Installation of Fire Safety Equipment Based on Use and Occupancy,” the required types and quantity of fire safety equipment are less for home stay facilities [11]. However, both hotels and home stay facilities face the equivalent fire hazards level.

Article 6 of the “Regulations for the Management of Home Stay Facilities” stipulates that the number of guest rooms shall be less than five and the total floor area shall not exceed 150 m2. Under special approval, the number of guest rooms can be increased to at most 15, but the total floor area shall be no more than 200 m2 [10]. The decorative materials, internal wall surface, ceiling decoration materials, structure of partition wall, corridor structure, and clear width shall conform to the “regulations on improvement on fire-prevention refuge facilities and firefighting equipment of existing building” [12]. This study reviewed the current regulations on home stay facilities and found that the indoor structure, material antiflaming requirement, fire safety equipment type, and periodic personnel training were specified loosely. The installation of fire safety equipment in home stay facilities is specified as follows [10]:

89

Effect of Fire Detection Function on Fire Suppression in Home Stay Facilities in Taiwan

• Each guest room, stairway, and corridor shall be equipped with emergency lighting. • There shall be automatic fire alarm equipment in the buildings, or each guest room shall be equipped with a residential fire alarm. • There shall be more than two fire extinguishers. Each story shall be equipped with at least one set for the storied buildings.

89.2.2 The Necessity of Self-Suppression of Fire in Home Stay Facilities Although the “Regulations for the Management of Home Stay Facilities” have rough measures for fire safety, whether the fire can be detected and extinguished as early as possible according to the existing regulations is a topic to be discussed. Some papers have conducted the computer simulation and analysis concerning the issue of the hotel fire safety by Fire Dynamics Simulator (FDS) and got a lot of research findings [13–15]. This study studied the difference in the fire detection time of different sensors and then the possible situations when suppressing a fire. This study used common computer software in fire safety for simulation to analyze the fire situation of home stay facility rooms. The temperature and visibility parameters inside the room on fire were used to evaluate the situation when the home stay facility operators enter the room on fire at the first second [16]. It is regarded as a condition that the personnel cannot enter and extinguish the fire safely if the inside of the room on fire does not meet the personal life safety criteria.

89.3

Research Method

89.3.2 The Description of Fire Scenario According to the regulations for home stay facilities, the total floor area of guest rooms shall be less than 200 m2 when the number of guest rooms is less than 15. Therefore, the room area is assumed to be about 15 m2. A room model in length and in width of 4 m, respectively, is built in this study. The partition wall is made of calcium silicate board, conforming to Article 7 of the Regulations for the Management of Home Stay Facilities. It is configured as a double suite, provided with wood desk, cabinet, bed frame, door, and floor, as well as foam-filled couch and mattress and cloth curtain, as shown in Fig. 89.2. The grid planning of FDS considered both the simulation efficiency and accuracy. According to the definition of Baum et al. [17, 18], the minimum length scale nearby the fire source is represented by characteristic fire diameter D*. The simulation grid is shown as the layout in Fig. 89.3.

Fig. 89.2 The configuration of simulated room

89.3.1 Fire Simulation Software This study used Fire Dynamics Simulator (FDS) software developed by the Building and Fire Research Laboratory (BFRL) of the National Institute of Standards and Technology (NIST). It is the most commonly used software for analyzing fire hazard around the world. FDS is a CFD (computational fluid dynamic) model for the dense smoke flow and heat transfer generated by fire. The simulation scope is split into geometric computational grid. The governing equations can be computed numerically. The fire simulation software of LES (large eddy simulation) equation is used to analyze spatial fire scenario, including the pressure, temperature, speed, and plume flow of fire scene. The Smokeview software is used to display the space situation in graphics. The user can observe the fire scene situation clearly and analyze the hazards in the fire development process.

871

Fig. 89.3 Grid frame in simulation model

872

C.-H. Su et al.

 D*

Q pffiffiffi ρ0 þ C p þ T 0  g



2=5 ¼

Q 1116

2=5 ð89:1Þ

The setup parameters of fire scene and fire source position in the home stay facility room are shown in Tables 89.1 and 89.2. The fire cannot be suppressed immediately since the condition is an unmanned room on fire. Table 89.1 Fire source position setting

Zone Zone D

Human activity frequency High

Zone A

High

Zone C Zone F

High Medium

Fire load Very low Very low Low High

Selection

Note Zone F is selected, and the fire source is set on the sofa

X

Table 89.2 Basic settings for FDS input file Item Simulated area Grid size Simulation time Fire source position Fire source area Heat release rate (HRR) Growth coefficient (α)

Set value 4.5 (L)  4.0(W)  2.5(H)m 0.1 m 600 s On sofa 0.2(L)  0.2(W) 0.57 MW 0.01127 Kw/s2

Table 89.3 Human life safety criteria Risk factor Smoke layer height Radiant heat Temperature CO concentration Visibility

Safety requirements Above 1.8 m Lower than 2.5 kW/m2 Below 60  C Below 800 ppm Above 10 m

The reference frame considered the fire occurrence probability and the consequence evaluation. It grew in time square model at medium speed, α ¼ 0.01127 kW/s2, provided that the heat release rate (HRR) reached the peak at 300 s [19]. The acceptable safety criterion for persons should be considered in the analysis of fire scene. The present life safety criteria of Taiwan are shown in Table 89.3 [20]. In terms of fire simulation time setting, according to the general detector types for home stay facility rooms, the actuation temperature for the rate of rise detector is room temperature plus 30  C, and the actuation concentration for residential smoke detector is 15 %m. It is actuated in 30 s after the set value is reached. The detection and confirmation of the fire take 10 s. When the fire is confirmed, the time for taking measures is set as 20 s [21] as shown in Table 89.4. In terms of the calculation of heat release rate, wood is used as fire source in this study, meeting the decoration characteristics of home stay facilities. If the wood burning rate per unit time is 30 g/s [22], the heat release rate is calculated as follows: Heat release rateðHRRÞQ ¼ Heat of combustion of unit wood  wood burning rate per unit time ¼ 19MJ  0:03 ¼ 19000  003

¼ 570 KW

89.4

ð89:2Þ

Results and Discussion

The burning process detected by different alarms is simulated. Figures 89.4 and 89.5 show the combustion regime of simulated indoor fire scene. Figure 89.6 shows the combustion heat release rate in the simulation process; the combustion is very severe. Figure 89.7 shows the indoor temperature variation curve with time. It is observed that the room temperature rises to 60  C after 255 s. Figure 89.8 shows the indoor visibility variation curve. It is observed that the indoor visibility has reduced to 10 m after 120 s.

Table 89.4 Time for reaching the room on fire Detector Fire confirmation Time for taking Case actuation time measures no. Actuated in 10 s 20 s SD-60 30 s 10 s 20 s HT-60 Actuation temperature: room temperature Actuated in 30 s +30  C 1. The response measure is that the person gets to the fire extinguisher and takes the fire extinguisher to the room on fire immediately to open the door 2. SD-60: the door is opened in 60 s after the residential smoke detector is actuated. The other symbols can be deduced by analogy 3. HT-60: the door is opened in 60 s after the rate of rise detector is actuated. The other symbols can be deduced by analogy

Detector type Residential smoke detector Rate of rise

Actuated setting Actuation concentration 15 %m

89

Effect of Fire Detection Function on Fire Suppression in Home Stay Facilities in Taiwan

(a) Analysis of sensing effect of rate of rise detector Table 89.5 shows the simulation results corresponding to time. If the home stay facility room is on fire, the personnel can enter the room in 60 s after the

Fig. 89.4 Combustion regime of indoor fire scene (at 60 s)

Fig. 89.5 Indoor state after door is opened (at 60 s)

Fig. 89.6 The curve of HRR

873

rate of rise detector is actuated. At the moment when the door is opened, the temperature at 180 cm above the floor at the door reaches 178  C, and the visibility is 0 m. It is obvious that the persons cannot enter the room to suppress the fire at this point. The results show that there are some problems in the rate of rise detector; the persons may be unable to enter the room on fire for fire suppression at the initial stage of fire. (b) Analysis of sensing effect of residential smoke detector The FDS simulated that the room on fire can be entered for fire suppression in 60 s after the residential smoke detector detects fire and sounds alarm. The simulation results are shown in Table 89.6. If the fire is not detected instantly or the action is delayed, the fire will not be extinguished. The visibility in the room 0 m at 240 s after the alarm is activated. As the temperature at the door is 125  C, the persons cannot enter the room for fire suppression. Table 89.7 shows the result of SD-60 simulated indoor visibility. The fire simulation shows that the smoke detector is actuated at 71 s of fire simulation, and the rate of rise detector is actuated at 278 s. The smoke detector detects the fire 207 s earlier than the rate of rise detector. The fire detection time varies with the detector type significantly. Therefore, the home stay facilities are recommended to use smoke detector, so as to detect fires early. The combustible materials in the home stay facility rooms are not always burning rapidly to increase indoor temperature quickly at the initial stage of fire, but these combustible materials generate a great deal of dense smoke. The home stay facility operators should strengthen fire-prevention education and fire drill regularly, so as to be familiar with the probable characteristics of the room on fire.

874

C.-H. Su et al.

Fig. 89.7 Time-temperature curve

Fig. 89.8 Visibility-time curve

Table 89.5 Simulation results of rate of rise detector Case no. HT-60 HT120

Activated time (s) 278 278

Door opened (s) 338 398

Visibility at 1.8 m (m) 0 0

Visibility at 1.5 m (m) 0 0

Temperature at 1.8 m ( C) 178 232

Evaluation Too high temperature, inaccessible

Zone D: door of room

Table 89.6 Simulation results of residential smoke detector Case no. SD-60 SD120 SD180 SD240

Activated time (s) 71 71

Door opened (s) 131 191

Visibility at 1.8 m (m) 7 2

Visibility at 1.5 m (m) 5 2

Temperature at 1.8 m ( C) 25 34

71

251

0.8

0.8

56

71

331

0

0

125

Zone D: door of room SD-N: the door is opened in N seconds after actuation

Evaluation Accessible to room for fire suppression The fire point cannot be identified Too high temperature, inaccessible

89

Effect of Fire Detection Function on Fire Suppression in Home Stay Facilities in Taiwan

Table 89.7 SD-60 simulation results (at the height of 1.8 m) Case SD-60 Visibility (m) Temperature ( C)

89.5

D zone 7

A zone 5

C zone 0

25

26

26

Evaluation Accessible to room for fire suppression

Conclusions

At lodging places, most guests are less aware of fires since resting or asleep. According to the statistical data, the home stay facility has become a popular alternative of lodging in Taiwan. Ensuring the safety of lodgings is very important for maintaining tourist safety. This study reviewed existing regulations on home stay facilities and found the terms to be loose. Firefighter cannot arrive at the scene immediately as many home stay facilities are located in remote scenic areas. The simulation results of this study showed that when a home stay facility room is on fire, the spread of fire is severe. The activation times of alarm systems permitted by the “Regulations for the Management of Home Stay Facilities” were analyzed. The results showed that the smoke detector has better detection effect and can detect fire earlier than the rate of rise detector. If the rate of rise detector is installed, the persons can hardly enter the room to suppress fire. Therefore, the home stay facilities are recommended to adopt the smoke detector. The home stay facility operators must suppress the fire by themselves to protect their properties and the guests.

References 1. Chen LC, Lin SP, Kuo CM (2013) Rural tourism: marketing strategies for the bed and breakfast industry in Taiwan. Int J Hosp Manag 32:278–286 2. Hall CM, Timothy DJ, Duval DT (2012) Safety and security in tourism: relationships, management, and marketing. Routledge, 6 Dec 2012. ISBN: 0789019167 3. Zheng XM, Wang XJ (2011) Research on the cooperation mechanism of tourism safety management in Fujian and Taiwan [J]. J Fujian Agric For Univ (Philos Soc Sci) 1:001 4. Ma´rquez Sierra FJ, Rubio-Romero JC, Rubio Ga´mez MC (2012) Status of facilities for fire safety in hotels. Saf Sci 50(7):1490–1494 5. Chow WK, Kot HT (1989) Hotel fires in Hong Kong. Int J Hosp Manag 8(4):271–281

875

6. Tourism Bureau, the Ministry of Transportation and Communications, Taiwan. Available: http://admin.taiwan.net.tw/ public/public.aspx?no¼237. Accessed 24 Dec 2014 7. Bryan JL (1983) A review of the examination and analysis of the dynamics of human behavior in the fire at the MGM Grand Hotel, Clark County, Nevada as determined from a selected questionnaire population. Fire Saf J 5(3):233–240 8. Buerk CA, Batdorf JW, Cammack KV, Ravenholt O (1982) The MGM Grand Hotel fire: lessons learned from a major disaster. Arch Surg 117(5):641–644 9. Chen YY, Chuang YJ, Huang CH, Lin CY, Chien SW (2012) The adoption of fire safety management for upgrading the fire safety level of existing hotel buildings. Build Environ 51:311–319 10. Tourism Bureau, the Ministry of Transportation and Communications (2001) Regulations for the management of home stay facilities, Taiwan. Available: http://law.moj.gov.tw/Eng/LawClass/ LawContent.aspx?PCODE¼K0110012. Accessed 12 Aug 2015 11. National Fire Agency, Ministry of the Interior (2013) Standard for installation of fire safety equipment based on use and occupancy, Taiwan 12. Construction and Planning Agency, the Ministry of Transportation and Communications (2012) Regulations on improvement on fireprevention refuge facilities and fire-fighting equipment of existing building, Taiwan 13. Shen TS, Huang YH, Chien SW (2008) Using fire dynamic simulation (FDS) to reconstruct an arson fire scene. Build Environ 43 (6):1036–1045 14. Shi W, Gao F (2014) Numerical simulation and evaluation of mechanical smoke exhaust in a loop corridor of a high-rise Hotel. In: Proceedings of the 8th International Symposium on Heating, Ventilation and Air Conditioning (pp 83–91). Springer, Berlin 15. Liu Y, Xu ZS, Yan L (2014) Study on hotel fire investigation based on FDS. In Intelligent Computation Technology and Automation (ICICTA), 2014 7th international conference on (pp 697–700). IEEE 16. Chang BL, Tang CH (2012) The simulating situation and the escaping strategy for disabilities in the huge underground space. Project report. Ministry of the Interior, Building Research Institute. Available: http://ir.lib.pccu.edu.tw/retrieve/46616/RRPG101010478-2732835[1].pdf. Accessed 12 Aug 2015 (in Chinese) 17. Baum HR, McCaffrey BJ (1989) Fire induced flow field – theory and experiment. In: Safety F (ed) Science – proceedings of the second international symposium. Hemisphere Publishing, Newport, pp 129–148 18. Su CH, Tsai KC, Xu MY (2015) Computational analysis on the performance of smoke exhaust systems in small vestibules of highrise buildings. J Build Perform Simul 8(4):239–252 19. Babrauskas V, Peacock RD (1992) Heat release rate: the single most important variable in fire hazard. Fire Saf J 18(3):255–272 20. Su CH, Lin YC, Shu CM, Hsu MC (2011) Stack effect of smoke for an old-style apartment in Taiwan. Build Environ 46(12):2425–2433 21. National Fire Agency, Ministry of the Interior (June, 2012) Fire equipment apparatus and equipment recognized standards, Taiwan. Available: http://law.ndppc.nat.gov.tw/GNFA/Chi/FLAW/ FLAWDAT0202.asp. Accessed 24 Dec 2014 (in Chinese) 22. Tanaka T, (2002) Introduction to fire safety engineering in buildings. Building Research Institute, pp 192 (in Japanese)

Full-Scale Experimental Study on Fire Suppression Performance of a Water Mist System for Large Shipboard Machinery Spaces

90

Xiaowei Wu and Shouxiang Lu

Abstract

As a clean fire extinguishing agent, water mist has been considered to be an alternative to Halon 1301 on shipboard. Recently, a new water mist system was designed for large shipboard machinery spaces. In order to test the fire suppression performance of this system, a full-scale shipboard machinery compartment having a volume of 1200 m3 was built. In this compartment, five fire scenarios, including diesel spray fires, diesel tray fires, and wood crib fires, were designed, and fire suppression parameters, including extinguishment time, compartment temperature, and gas concentrations, were measured. For all the scenarios, extinguishment times were typically less than 1 min, except for small fires. Average compartment air temperatures dropped to 70  C within seconds after mist was activated. The CO concentrations of all the scenarios were below 4000 ppm, which was not life-threatening for short-term exposures. The test results indicated that this water mist system had an excellent ability to extinguish both obstructed and unobstructed Class B tray fires, Class B spray fires, and Class A fires in large machinery spaces. Meanwhile, the system could minimize the thermal damage to the compartment and its infrastructure and provide favorable conditions for personnel evacuation and firefighting. In conclusion, this newly designed water mist system was a viable alternative to Halon 1301 system for protecting large machinery spaces. Keywords

Fire  Extinguishment  Halon alternatives  Water mist  Machinery space

90.1

Introduction

Fires on shipboard may lead to a loss of capability, function, and life and, in extreme situations, the loss of the ship. Machinery spaces are high fire risk areas where flammable liquids are pumped and piped under pressure in the presence of ignition sources. Thus, installed fire protection systems are required in these spaces. As a very effective extinguishing agent, Halon 1301 has been widely used for

X. Wu  S. Lu (*) State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei, Anhui 230027, China e-mail: [email protected]

protecting ship machinery spaces since the late 1970s [1]. But Halon 1301 has ozone-depleting effect, which could decompose into chlorine, fluorine, and carbon components that react with the ozone in the atmosphere to eliminate it. In 1987, the provision of the Montreal Protocol was adopted to ban the use of Halon 1301 [2]. Faced with the inability to utilize Halon 1301 for future designs, considerable efforts were expended on halon replacement. One excellent candidate of the halon alternatives is water mist. The abundance of water, its lack of toxicity, and its environmental acceptability offered attractive incentives to pursue a water mist system. The initial studies of water mist for fire protection applications started from 1978 to 1980 by the US Navy

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_90

877

878

X. Wu and S. Lu

[3, 4]. In this period, the research was focused on calculations that illustrated how small water droplets, as contained in “water fog” or “fine water spray,” could achieve fire extinguishment by gas phase cooling. Then some experiments were carried out to validate the theoretical calculations. From 1990 to 1994, water mist research entered into a prosperous period; the US Naval Research Laboratory conducted a series of small-scale testing in a 3  3  2.4 m steel compartment [5]. Parameters such as fire size and location, nozzle spacing, mist application rates, mist characteristics, ventilation, degree of obstructions, corner effects, and oxygen depletion were analyzed. Following small-scale tests, full-scale tests were conducted to develop fundamental water mist system design parameters for protecting ship machinery spaces. But most of these tests were conducted aboard the US Navy’s fire research vessel, the Ex-Shadwell [6–13]. The Shadwell tests were run in a two-level compartment having a gross volume of 960 m3. Thus, a question was posed: “if the shape of the test compartment changes and its volume is larger than 1000 m3, are the conclusions of the Shadwell tests still working?” Meanwhile, the performance of nozzles and the technique of piping works are developing. Recently, a new water mist system was designed, whose fire suppression performance needed to be tested. For the reasons above, a test compartment having a volume of 1200 m3 was built, which simulated a real shipboard machinery space. Five full-scale tests were conducted to study the suppression performance of the newly designed water mist system in the large compartment. In this article, the test compartment, the configuration of the newly designed water mist system, the test fire scenarios, and the measure instrumentation were introduced. Then, the impact of water mist injection on extinguishment time, compartment temperature, and concentrations of combustion products was studied. A series of full-scale test data was obtained, providing relating reference for such engineering applications.

90.2

Experimental

90.2.1 Experimental Compartment The tests were conducted in a full-scale compartment which was built to simulate a typical machinery space of ships (Fig. 90.1). The space was roughly 21  12  6 m (1200 m3) and was bounded by metal bulkheads. This space had three levels, i.e., a bilge level and two levels of catwalks, as shown in Fig. 90.1. These catwalks

allow easy access to critical areas in the space and, to some degree, serve as obstructions for the water mist systems. The catwalks located high in the space were constructed with metal grating, and the lower catwalks were constructed of steel plating. Typical machinery space equipments, such as engines, were simulated using sheet metal mock-ups as shown in Fig. 90.1. A 2  2 m door was added in the aft bulkhead on the lower catwalk level, and a similar door was added in the forward bulkhead on the lower catwalk level. Four 1.2  1.3 m vents and two 1.8  1.1 m vents were added on the roof.

90.2.2 Water Mist System Three types of nozzles were used in this water mist system. All the nozzles were researched and developed by self, and their specifications are listed in Table 90.1. In order to improve the suppression capabilities of the system, the piping network was divided into three layers. The upper layer consisting of 31 nozzles was installed under roof. The middle layer consisting of 23 nozzles was installed under catwalks of grating, i.e., higher catwalks. The lower layer consisting of 16 nozzles was installed under catwalks of plate, i.e., lower catwalks. The piping network with nozzle locations was shown in Fig. 90.2. The nozzle placement rules were followed. Under each platform of the compartment, a layer of nozzles should be installed. The spacing between nozzles should be less than the effective operating diameter of the nozzle. More nozzles should be considered in zones at high fire risk.

90.2.3 Fire Scenarios To select fire scenarios reasonably, on the one hand, general standards, such as MSC/Circ.848, MSC/Circ.1165, and so on, were adopted. On the other hand, the possible fire scenarios of the simulated machinery compartment were analyzed. Five types of fires were used in this study. The parameters of Fire #1–5 were listed in Table 90.2. Fire #1–4 were diesel fires to simulate Class B fires. Fire #1 was low-pressure horizontal spray with nominal fuel pressure 8 bar and fuel flow 0.16  0.01 kg/s. Fire #2 was an unobstructed small tray fire. Fire #3 was an unobstructed large tray fire. Fire #4 was obstructed by an engine mockup. Fire #5 was a wood crib fire to simulate Class A fire. The locations of Fire #1–5 were shown in Fig. 90.1. With the combination of Fire #1–5, five fire scenarios are concluded, as shown in Table 90.3.

90

Full-Scale Experimental Study on Fire Suppression Performance of a Water Mist. . .

879

Fig. 90.1 Machinery space mock-up: (a) elevation view, (b) plan view of the higher catwalk level, (c) plan view of the lower catwalk level

Table 90.1 Specifications of the water mist nozzles Nozzle type #1 #2 #3

DV99 (μm) 70–100 70–100 70–100

Operating pressure (MPa) 8–10 8–10 8–10

Kfactor 1.45 1.45 1.21

Cone angle ( ) 90 120 120

Flow rate (L/min) 13 13 13

In order to study the impact of water mist discharge on combustion products, gas analyzer was set up to measure the concentrations of carbon monoxide, carbon dioxide, and oxygen at three elevations in the compartment. The three measure points were poisoned at 1.0 m, 2.5 m, and 4.0 m above the lower catwalk level. The gas measure point (GM) layout was shown in Fig. 90.1.

90.2.4 Instrumentation 90.2.5 Experimental Procedures Three thermocouple trees were installed in the compartment. The thermocouple tree (TC Tree #1–3) layout was shown in Fig. 90.1. Each tree consisted of nine thermocouples. In order to measure the temperature at bilge, one thermocouple was set 0.5 m below the plates of the lower catwalks. The other eight thermocouples were poisoned at 0.5 m increments starting 0.5 m above the lower catwalk level. In order to determine the extinguishment time of each fire, thermocouples were located in the flame region of the fires. The thermocouples used in those tests were inconel-sheathed type-K thermocouples with 1.0 mm in diameter and could measure temperature up to 1200  C.

The data acquisition system was activated prior to the test. Then the test fires were ignited leaving the compartment vents open. Fires were allowed a certain preburn period before the water mist is discharged. A 120-s preburn time is required for the tray fire and 5–15 s for the spray fire. For the wood crib fire, the preburn time is 30 s for Fire Scenarios #4 and 360 s for Fire Scenarios #5. After the preburn, the compartment was then sealed, and the mist system was activated. The mist system remained activated for a period of 15 min during each test or until all of the fires had been extinguished, whichever came first. At the completion of the

880

X. Wu and S. Lu

Fig. 90.2 Piping network with nozzle locations: (a) higher catwalk level, (b) lower catwalk level, (c) bilge level Table 90.2 Parameters of fires Fire no. #1 #2 #3 #4 #5

Type Spray 0.5 m2 tray 2.0 m2 tray 4.0 m2 tray Wood crib

Table 90.3 Fire scenarios Fuel Diesel Diesel Diesel Diesel Spruce

Fire size (MW) 6.0 0.71 3.0 6.0 0.6

discharge, the mist system was secured marking the end of the test. The space remained off-limits until cleared by the safety officer and the test director.

90.3

Results and Discussions

90.3.1 Flame Temperature and Fire Extinguishment Time Extinguishment time was an important parameter to evaluate the suppression performance of a water mist system. In this article, the extinguishment time was defined as the time

Scenario no. #1 #2 #3 #4 #5

Fire combinations Fire #1 Fire #2 Fire #3+ Fire #4 Fire #5 Fire #1+ Fire #3+ Fire #4+ Fire #5

between the injection of the water mist and the instant of fire extinguishment. The extinguishment of fires was determined based on temperature measurements recorded in the flame during each test. The flame temperature of each fire for each fire scenario was shown in Fig. 90.3. Extinguishment time for each fire scenario was shown in Fig. 90.4. For Fire Scenario #1, the flame temperature of the spray fire was relatively low, about 160  C before the system activation. After water mist injection, the temperature decreased while oscillating. This indicated that the water mist formed a relatively sealed space around the

1000

200 150 100

700 600 500 400 300

Extinguishment

800

Mist System Activation

250

900 Flame Temperature (°C)

300 Flame Temperature (°C)

881

b

350 Extinguishment

a

Full-Scale Experimental Study on Fire Suppression Performance of a Water Mist. . .

Mist System Activation

90

200 50 100 0

0

20

40

60

80

100 120 Time (s)

140

160

180

200

c

10 20

30

40 50

60

70 80 90 100 110 120 130 140 150 Time (s)

1000

2.0m2 Tray 4.0m2 Tray

500

300 200 100 0

20

40

60

80

100

120

140

160

700 600 500 400 300 200

2

Mist System Activation

4m Tray Extinguishment

400

Flame Temperature (°C)

600

Extinguishment

800

2m2 Tray Extinguishment

700

900

Mist System Activation

800 Flame Temperature (°C)

0

d

900

0

0

220

100 180

200

0

0

40

80

120

160

Time (s)

200 240 Time (s)

280

320

360

400

e

1000

Extinguishment

900

700 600 500 Spray 2m2 Tray 4m2 Tray Wood Crib

400 300 200 100 0

0

40

80

Mist System Activation

Flame Temperature (°C)

800

120 160 200 240 280 320 360 400 440 480 Time (s)

Fig. 90.3 Time history of flame temperature: (a) Fire Scenario #1, (b) Fire Scenario #2, (c) Fire Scenario #3, (d) Fire Scenario #4, (e) Fire Scenario #5

spray fire and the flame burnt as in a poor ventilation enclosure. Thus, the spray fire turned to be a small fire which was more difficult to extinguish than larger fires. So the extinguishment time for this scenario was relatively longer, about 115 s.

For Fire Scenario #2, the flame temperature of the small tray fire was about 900  C before the system activation and dramatically decreased in a short time after water mist injection. The extinguishment time for this scenario was about 10 s.

882

X. Wu and S. Lu

Extinguishment Time (s)

240 Spray 0.5m2 Tray 2.0m2 Tray 4.0m2 Tray Wood Crib

200 160 120 80 40 0

#1

#2

#3 Fire Scenario

#4

#5

Fig. 90.4 Extinguishment time comparison for all the fire scenarios

For Fire Scenario #3, the 2.0 m2 tray fire was unobstructed, and the 4.0 m2 tray fire was obstructed by an engine mock-up. The extinguishment times of the two fires were both very short, 17 s for the 2.0 m2 tray fire and 22 s for the 4.0 m2 tray fire. For Fire Scenario #4, the flame temperature of the wood crib fire was about 900  C before the system activation. After water mist injection, the temperature decreased relatively slowly. The extinguishment time was about 241 s. Why Class A fires were so difficult to extinguish? Perhaps because the water mist could only put out the outer surface flame of the wood crib, and it was still high temperature in its inner part. For Fire Scenario #5, fire resources included spray fire, unobstructed tray fire, obstructed tray fire, and wood crib fire. After water mist injection, all the fires extinguished in a short time, approximately 30 s. This indicated that although fires like Fire #1 and Fire #5 were difficult to extinguish alone, in the presence of larger fires, they were much easier to extinguish.

90.3.2 Compartment Temperature Water mist could remove heat from a system because of its high specific heat and latent heat of vaporization. Thus, water mist systems could be used to thermally manage the conditions in machinery spaces. In order to analyze the temperature changes in the test compartment, Fire Scenario #1 was taken, for example. The time history of compartment temperature at different heights for Fire Scenario #1 was shown in Fig. 90.5. With regard to one thermocouple tree, before mist discharge, temperatures of higher thermocouples were higher

than those of lower thermocouples. It was in accordance with the fact that the upper room was filled with hot smoke and the lower room filled with cool air. After water mist activation, the temperatures of higher thermocouples decreased, and those of lower thermocouples increased. This illustrated that the upper smoke layer and down cool air were mixed because of water mist injection. After mist injection ending, the temperature gradient at different levels reappeared. Perhaps this phenomenon was caused by the air mixing effects of water mist injection. As shown in Fig. 90.5, for one thermocouple tree, the temperatures at different heights exhibited almost the same trend. Thus, average temperature was employed here. The time history of average compartment temperature of different thermocouple trees for all the fire scenarios was shown in Fig. 90.6. For all the fire scenarios, the compartment temperature reduced below 70  C. According to US naval ships’ technical manual, human would be incapable in 5 min and died in 30 min, and electronics would be permanent damaged when temperature was above 150  C. Thus, the compartment temperature in our tests would help manual intervention, minimize thermal damage, and prevent fire spread from the compartment of origin.

90.3.3 Gas Concentration The results of gas concentration measurement indicated that, for all cases, very slight variations were found in the O2 and CO2 concentrations, so only the changes in CO concentrations were analyzed in the following. The concentrations of CO at three measure points were shown in Fig. 90.7. The changes at three measure points exhibited almost the same trend for all the scenarios. The amounts of CO produced during the preburn remained fairly constant at low value, about 30 ppm, because the combustion was oxygen enriched in this period. During the extinguishment process, the amounts of CO were observed to dramatically increase in a certain time and then decreased. This indicated that water mist injection really brought incomplete combustion at the beginning, but it could then dilute CO with the continuing discharge. Compared to all the scenarios, the CO concentration was larger when fire HRR was higher. While the extinguishment of these fires produced significantly higher CO concentrations, the amount of CO for each scenario was below 4000 ppm. According to US naval ships’ technical manual, human would be incapable in 5 min and died when CO concentration was above 4000 ppm. Thus, the CO concentration in our tests was not life-threatening for short-term exposures. It would provide favorable conditions for personnel who remained staying in the space.

42 40

120 110

38 36 34 32 30 28

100 90

Elevation 4.0m 3.0m 2.0m 1.0m -0.5m

3.5m 2.5m 1.5m 0.5m Extinguishment

44

3.5m 2.5m 1.5m 0.5m

130

Mist System Activation

Comparterment Temperature (°C)

46

Elevation 4.0m 3.0m 2.0m 1.0m -0.5m

Comparterment Temperature (°C)

Mist System Activation

48

883

80 70 60

Discharge Ending

b

50

Discharge Ending

a

Full-Scale Experimental Study on Fire Suppression Performance of a Water Mist. . .

Extinguishment

90

50 40

26

30

24 22

0

20

40

60

80

100

120

140

160

180

200

20

220

0

20

40

60

80

100

Time (s)

140

160

180

200

220

40 38 36

Elevation 4.0m 3.0m 2.0m 1.0m -0.5m

3.5m 2.5m 1.5m 0.5m

34

Discharge Ending

Comparterment Temperature (°C)

42

Extinguishment

44

Mist System Activation

c

120

Time (s)

32 30 28 26 24 22 20

0

20

40

60

80

100

120

140

160

180

200

220

Time (s)

Fig. 90.5 Time history of compartment temperature at different heights (Fire Scenario #1): (a) thermocouple tree #1, (b) thermocouple tree #2, (c) thermocouple tree #3

90.4

Conclusions

Water mist fire suppression systems are being seriously considered to replace Halon 1301 total flooding systems in machinery space applications. In this article, a full-scale experimental facility was built to study the fire suppression performance of a newly designed water mist system for large shipboard machinery spaces. A series of full-scale tests was carried out with different fire scenarios. The tests contributed valuable experimental data for water mist systems installed in large machinery spaces. Conclusions can be drawn as follows:

1. These tests have demonstrated the ability of the newly designed water mist system to extinguish both obstructed and unobstructed Class B tray fires, Class B spray fires, and Class A fires in large machinery spaces. 2. The system required minutes to extinguish fires as opposed to seconds for the gaseous halon alternatives. 3. The parameters of the water mist system nozzles and piping networks, the extinguishment experimental data, and the results provided in this article may serve as an available reference for such engineering applications. 4. The water mist system can be used to thermally manage the conditions in the machinery spaces. For all of the fire

884

44

Average Compartment Temperature (°C)

#1 TC Tree #2 TC Tree #3 TC Tree

46

42 40 38 36 34 32

Extinguishment

Mist system Activation

Average Compartment Temperature (°C)

48

30 28 26 24 22

30.0

#1 TC Tree #2 TC Tree #3 TC Tree

29.5 29.0 28.5 28.0 27.5 27.0

Extinguishment

b

50

Mist System Activation

a

X. Wu and S. Lu

26.5 26.0 25.5 25.0

0

20

40

60

80

100

120

140

160

180

200

220

0

50

100

150

200

Time (s)

350

400

450

500

2

80 70 60 50 40

34

Extinguishment

90

36

Mist System Activation

100

Average Compartment Temperature (°C)

110

#1 TC Tree #2 TC Tree #3 TC Tree

38

#1 TC Tree #2 TC Tree #3 TC Tree 4m Tray Extinguishment

120

2

Mist System Activation

130

Average Compartment Temperature (°C)

300

d

140

2m Tray Extinguishment

c

250

Time (s)

32

30

28

30 20

26 0

50

100

150

200

250

0

300

100

200

300

500

600

700

800

900

160

Extinguishment

#1 TC Tree #2 TC Tree #3 TC Tree

140

120

100

Mist System Activation

Average Compartment Temperature (°C)

e

400

Time (s)

Time (s)

80

60

40

20 0

100

200

300

400

500

600

Time (s)

Fig. 90.6 Time history of average compartment temperature of different thermocouple trees: (a) Fire Scenario #1, (b) Fire Scenario #2, (c) Fire Scenario #3, (d) Fire Scenario #4, (e) Fire Scenario #5

scenarios, the system dramatically reduced the temperature of the space almost immediately after activation, and the space became well mixed with a uniform temperature below 70  C. This reduction in temperature will minimize the thermal damage to the compartment and its

infrastructure and provide favorable conditions for personnel evacuation and firefighting. 5. For all of the fire scenarios, CO concentrations were below 4000 ppm. It would provide favorable conditions for personnel who remained staying in the space.

b

3000 2800

Mist system Ativation

2000

100

1800 1600 1400

CO Concentration (ppm)

2200

Extinguishment

2400

CO Concentration (ppm)

120 110

Elevation 4.0m 2.5m 1.0m

2600

1200 1000 800 600

885

Extinguishment

a

Full-Scale Experimental Study on Fire Suppression Performance of a Water Mist. . .

Mist System Activation

90

90 80 70

Elevation 4.0m 2.5m 1.0m

60 50 40 30 20

400

10

200 0

0 0

50

100

150

200

250

0

300

50

100

150

200

250

d

4000

Elevation 4.0m 2.5m 1.0m

800 700 600

500

550

600

500 400 300 200

2

1000

500

Extinguishemt

2

1500

450

Mist System Activation

2000

CO Concentration (ppm)

2500

400

Elevation 4.0m 2.5m 1.0m

900

4m Tray Extinguishment

3000

350

1100 1000

Mist System Activation

CO Concentration (ppm)

3500

2m Tray Extinguishment

c

300

Time (s)

Time (s)

100

0

0 0

50

100

150

200

250

300

350

400

450

0

50

100

150

Time (s)

300

350

400

450

500

550

600

650

700

3500

Elevation 4.0m 2.5m 1.0m

2500

2000

Extinguishment

3000

CO Concentration (ppm)

250

Time (s)

Mist System Activation

e

200

1500

1000

500

0 0

50

100

150

200

250

300

350

400

450

500

550

600

650

700

Time (s)

Fig. 90.7 Time history of CO concentrations at different heights: (a) Fire Scenario #1, (b) Fire Scenario #2, (c) Fire Scenario #3, (d) Fire Scenario #4, (e) Fire Scenario #5.

In conclusion, this newly designed water mist system has an excellent ability to extinguish fires in the secured large compartment. However, ventilation was not considered in those tests. Thus, in the future, the ventilation effect on the fire suppression capabilities of the water mist system will be studied in the full-scale test compartment.

References 1. Darwin RL, Williams FW (2000) The development of water mist fire protection systems for U.S. navy ships. Navy Eng J 11:49–57 2. United Nations Environment Program (1987) The montreal protocol on substances that deplete the ozone layer, Nairobi

886 3. Fielding, GH, Williams EW, Carhart HW (1977) Suppression – why not water? NRL Memorandum Report 3435, Washington, DC 4. Lugar JR, Fornsler RO, Carhart HW Fielding GH (1978) Flame extinguishment by waterfogs and sprays, fifth quadripartite conference IEP ABCA-7 5. Hanauska CP, Back GG (1993) Halons: alternative fire protection systems, an overview of water mist fire suppression systems technology. Hughes Associates, Columbia 6. Leonard JT, Back GG (1994) Revised test plan: full-scale testing of total-flooding water mist system, NRL Ltr Rpt Ser 6180/0716.2 7. Leonard JT, Back GG, DiNenno PJ (1994) Full scale machinery space water mist tests: phase I – unobstructed space. NRL Ltr Rpt Ser 6180/0713.1. Naval Research Laboratory, Washington, DC 8. Leonard JT, Back GG, DiNenno PJ (1994) Full scale machinery space water mist tests: phase II – simulated machinery space. NRL Ltr Rpt Ser 6180/0868.2, Naval Research Laboratory, Washington, DC 9. Darwin RL, Leonard JT, Back GG (1995) Status report on the development of water mist systems for U.S. Navy shipboard

X. Wu and S. Lu machinery space, halon options technical working conference, final program/abstracts. Albuquerque 10. Leonard JT, Darwin RL, Back GG (1995) Full scale tests of water mist fire suppression systems for machinery spaces. In: Proceedings international conference on fire research and engineering, D.P Lund, Editor, Society of Fire Protection Engineers 11. Back GG, DiNenno PJ, Leonard JT, Darwin RL (1996) Full scale tests of water mist fire suppression systems for navy shipboard machinery spaces: phase I – unobstructed spaces. NRL/MR 618096-7830, Naval Research Laboratory, Washington, DC 12. Back GG, DiNenno PJ, Leonard JT, Darwin RL (1996) Full scale tests of water mist fire suppression systems for navy shipboard machinery spaces: phase II – obstructed spaces, NRL/MR/6180967831. Washington, DC 13. Back GG, Darwin RL, Leonard JT (1996) Full scale tests of water mist fire suppression systems for navy shipboard machinery spaces. In: Proceedings INTERFLAM 96. St. Johns College, Cambridge

Examination of Extinguishment Method with Extinguishing Powder Packed in a Spherical Ice Capsule

91

Miho Ishidoya, Hiroyuki Torikai, Akihiko Ito, and Yuji Shiibashi

Abstract

In the study, the extinguishment method with an extinguishing powder packed in the hollow capsule made of ice has been proposed and investigated experimentally. If a capsule can be used to deliver an extinguishing powder, it will be possible to transport the agent to the fire area over a long distance. The ice capsule is formed by using the double rotational axis casting machine. The amount of water of 1.8 cm3 is used to make the capsule which has 20 mm in outer diameter. ABC extinguishing powder is filled into the ice capsule. To clarify the characteristics of the ice capsule extinguishment, the extinguishing experiments of a methane-air diffusion flame have been performed, and also the extinguishing process has been observed with a high-speed camera. The ice capsule drops in free fall from the height of 1.8 m, and the impact velocity onto the metal plate, in which the burner is embedded, is 5.8 m/s. The extinguishment of the flame with the extinguishing powder of 3 g filled into the ice capsule has been succeeded, and the maximum effective range of the extinguishing method is 120 mm distance from the burner center to the dropping location of the ice capsule. Keywords

Extinguishment method
Ice capsule
Extinguishing powder
Diffusion flame

Nomenclature

91.1

L m P

In an ordinary portable powder fire extinguisher, the extinguishing powder is released from the extinguisher body through the hose toward the fire area by using a highpressure gas jet. As the jet flow with the powder progresses forward in the surrounding air, the amount of air entrained into the shear layer of the jet flow increases, and also the flow area spreads in lateral direction. As a result, the concentration of the extinguishing powder decreases, and the axial jet flow velocity is decelerated in the downstream. Therefore, there is the maximum effective distance of a portable extinguisher, which is usually less than ten meters. Thus, when a jet flow is used for the delivery method of powder extinguishing agent, it is difficult to transport and supply it to the fire area over a long distance. In particular, the smaller particles in the extinguishing powder agent,

Distance from impact location to burner center (mm) Quantity of extinguishing powder (g) Extinguishing probability

Subscripts ex

Extinguishing distance limit

M. Ishidoya
H. Torikai (*)
A. Ito
Y. Shiibashi Graduate School of Science and Technology, Hirosaki University, Hirosaki, Aomori 036-8561, Japan e-mail: [email protected]

Introduction

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_91

887

888

which are approximately less than 20 μm and have a higher extinguishing effectiveness [1, 2], are more difficult to be transported with the jet flow because several micro-order particles in diameter behave similar to extinguishing gaseous agents which are easy to diffuse into ambient air. On the other hand, after the occurrence of a large-scale earthquake, such as the Great East Japan Earthquake, multiple simultaneous fires often break out. Moreover, infrastructure, such as water utilization for firefighting, roads, and etc., is destroyed violently by the earthquake impact or tsunami. In that emergency situation, fire engines cannot reach the fire area, and local residents would have to fight the postearthquake fires by themselves using portable fire extinguishers. However, if local residents cannot approach to the fire area due to the damaged road or tsunami debris, it would be impossible to extinguish the fires with ordinary fire extinguisher because of the short effective range. Therefore, to mitigate and minimize the damage of post-earthquake fires, the development of a new firefighting method, which can be easily used by the general public and deliver extinguishing agents over a long distance more than several tens of meters, is needed. The authors have proposed and investigated the capsule extinguishing method with gaseous extinguishing agents, in which the capsule filled with extinguishing agents is transported from the application location to the fire area [3–6]. For example, when an inert gas is filled into a capsule such as a soap bubble and a rubber balloon, the capsule membrane inhibits mutual diffusion between the filling gas and the surrounding air and ruptures due to contact with the flame. Therefore, using the extinguishing capsule, the extinguishing gas can be easily transported through the atmosphere over a long distance and supplied directly to the fire zone without reduction of its concentration. Thus, the utilization of the capsule to fire extinguishing method will make the long-distance delivery of the extinguishing agents easier and the maximum effective range of the portable extinguisher wider. In this study, a hollow ball made of ice is used as a capsule to transport the extinguishing powder. The ice ball is a hard shell capsule and can be ejected easily at the high velocity with low-pressure gas gun like paintball gun whose firing range is over several tens of meters. Moreover, by using ice, it is easy to form various shapes and sizes of the capsule, and also when the ice capsule impacts on a ground or wall, it can be crushed easily and releases the extinguishing powder inside it to the surroundings. In the extinguishing experiment of this study, ABC extinguishing powder is filled into the ice capsule. A methane-air diffusion flame formed on a tube burner is used as the extinguished target. As the first step of the examination of the ice-capsule extinguishing method, the capsule is dropped in free fall and impacted onto a solid wall surface in the extinguishing experiment. From the

M. Ishidoya et al.

experiment result, we can clarify experimentally the extinguishing range from the impact location to the flame with ABC extinguishing powder packed in the ice capsule. The magnitude of the extinguishing range is considered to be a fundamental extinguishing ability of the capsule filled with the extinguishing agent. To grasp the extinguishing process, the extinguishing process is observed with a high-speed camera.

91.2

Experimental Setup and Method

91.2.1 Formation of Spherical Ice Capsule The hollow ice ball to encapsulate an extinguishing powder was formed using the rotary molding method. Figure 91.1 shows the molding machine of the ice capsule with semirandomized two-axis motion. Frames 1 and 2 in Fig. 91.1 rotated vertically and horizontally, respectively. The driving force was given by the electric geared motor, and its torque was transmitted with a rubber belt from the pulley 1 which was fixed to the main frame of the rotating machine to the pulley 2 which was a movable one. Figure 91.2 shows the ice capsule mold, which was made of acrylic material. The outer and inner diameters of the acrylic hemispherical bowl were 25 and 20 mm, respectively. To form the hole in the ice capsule, the external thread shown in Fig. 91.2 was used. Figure 91.3 shows the formed capsule with a hole, through which the extinguishing powder was fed into the capsule. The diameter of the hole in the ice capsule was approximately 3 mm. The ice capsule had almost spherical shape and the almost uniform wall thickness. The outer diameter of the ice capsule was approximately 20 mm. To separate easily

Fig. 91.1 Double rotational axis casting machine

91

Examination of Extinguishment Method with Extinguishing Powder Packed in a Spherical Ice Capsule

889

Fig. 91.2 Hemispherical plastic bowls to form the ice capsule

Fig. 91.4 Experimental setup for the extinguishing experiment with the ice capsule filled with extinguishing powder

Fig. 91.3 Ice capsule (black circle on the ice capsule shows the hole for filling ABC extinguishing powder)

the formed ice capsule from the acrylic bowl, silicone oil was used as a releasing agent, which was sprayed on the inner surface of the ice mold before molding. The volume of water to make one capsule was 1.8 cm3. When the rotary molding machine with the ice mold at the rotating rate of about 72 rpm was placed in a freezer at about minus 10  C, it took approximately 35 min to form one ice capsule. In order to keep the movement of the electrical motor stable under the low-temperature condition, a heater was set in the motor section, and its section was covered with a thermal insulating material. ABC extinguishing powder (Yamato Protec Co., Ltd.), which is a mixture of ammonium dihydrogen phosphate (NH4H2PO4), ammonium sulfate ((NH4)2SO4), and silicon dioxide (SiO2) and is a very popular extinguishing powder, was used. The powder agent was put into the ice capsule through the hole, and then the hole was closed with a small piece of ice. To suppress the ice capsule to melt during the preparation of the experiment, the capsule was always cooled with a Peltier cooling device.

91.2.2 Extinguishing Experiments Figure 91.4 shows the experimental setup of the flame extinguishment with ABC extinguishing powder packed in the ice capsule. A methane-air jet diffusion flame was used as the extinguishing target, and formed with the aluminum tube burner, whose inner and outer diameters were 9 mm and 10 mm, respectively. The burner was embedded in the iron plate, and the top of the tube burner was set at the same height of the iron plate surface. The fuel gas was supplied to

the burner at the volumetric flow rate of 3.4 l/min, and the measured flame height was 325  47 mm. From the calculation using the lower heating value of methane, the thermal size of the flame was 1.84 kW. As the first step to examine the ice-capsule extinguishing method, the ice capsule was dropped in free fall at the height of 1800 mm from the iron plate surface. The measured impact velocity of the ice capsule filled with the extinguishing powder was 5.8 m/s. When the ice ball collides on the iron plate, the wall of the ice capsule was broken into small pieces, and also the extinguishing powder was scattered around the impact location. In the extinguishing experiments, the weight of ABC extinguishing powder, m [g], and the distance from the center position of the tube burner to the impact location of the ice capsule on the iron plate, L [mm], were varied as parameters. To reveal the effective distance of the ice capsule filled with an extinguishing agents, the distance, L, was changed from 20 to 240 mm, and the weight of the extinguishing powder, m, was also changed from 0 to 3 g. At m ¼ 3 g, the ice capsule was full of ABC extinguishing powder. The weight of the extinguishing agent was measured with a digital weighing scale (A&G, GF-610 g). To evaluate the extinguishing effectiveness of the extinguishing ice capsule, the extinguishing probability, P, was measured. The extinguishment experiments were performed in the following way. First, the location of the iris to release the ice capsule was set at a certain distance from the burner center, and the stable jet diffusion flame was formed. Second, the ice capsule dropped and impacted on the iron plate, and then the extinguishing powder was scattered around. Third, we checked whether the flame was extinguished or not visually. When the flame was extinguished perfectly, we recorded it as a success of the extinguishment. The probability was computed as the ratio of the number of successful extinguishments to the number of total experiments of ten.

890

91.3

M. Ishidoya et al.

Results and Discussion

91.3.1 Impacting Process of Ice Capsule Figure 91.5 shows a series of sequential images in the typical impacting process of the ice capsule onto the iron plate. The images were recorded with a high-speed camera (Nac, GX-8; frame rate, 20,000 fps and exposure time, 10 μs), and the scattered fragments produced from the broken ice capsule was illuminated with a metal halide lamp (Photron, HVC-UL, 250 W). The time in images of Fig. 91.5 indicates the elapsed time from when the lowest part of the ice capsule contacts on the surface of the plate. As seen in Fig. 91.5, the ice capsule is completely crushed and is deformed from the ball shape to the flat one up to 3 ms. The fragments of the broken ice capsule fly all over the plate from 8 ms to 20 ms, and the relatively large sizes of the fragments are about several millimeters. While some pieces of the broken capsule fly in upward direction, the most part of the ice fragments has momentum in the lateral direction in the image and moves along the plate surface. Figure 91.6 shows a series of sequential images in the typical releasing process of the extinguishing powder from

the crushed ice capsule. The images were recorded with a high-speed camera (Nac, HX-3; frame rate, 10,000 fps and exposure time, 100 μs) and a metal halide lamp (Photron, HVC-UL, 250 W). From the images in Fig. 91.6, the ice capsule is crushed up to 3 ms as well as in Fig. 91.5. Then the extinguishing powder is released and flies from the crushed ice capsule to diagonally upward and lateral direction. To understand the behavior of the extinguishing powder released from the broken ice capsule, the lateral position of the leading edge of the moving ice fragment in the images of Fig. 91.5 and the scattering extinguishing powder in images of Fig. 91.6 were measured, and the results are shown in Fig. 91.7. The horizontal axis of the graph shows the elapsed time, and the origin of the vertical axis is set at the impacting location of the ice capsule. From Fig. 91.7, it is found that the ice fragment and the extinguishing powder released from the crushed ice capsule indicate the same behavior and displacement speed, which is computed from the slop of the straight line in Fig. 91.7. As a result, it can be said that the scattering behavior of the extinguishing powder follows the motion of the ice fragments produced from the crushed ice capsule.

Fig. 91.5 Sequential images of the impact process of the empty ice capsule onto the iron plate

Fig. 91.6 Sequential images of the extinguishing powder scattering around from the broken ice capsule

91

Examination of Extinguishment Method with Extinguishing Powder Packed in a Spherical Ice Capsule

91.3.2 Extinguishing Process with Ice Capsule Figure 91.8 shows the sequential images of the process of the flame extinguishment with the extinguishing powder packed in the ice capsule. The images were recorded with a digital camera (CASIO, EX-F1; frame rate, 300 fps; exposure time, 1/320 s). The time in the images of Fig. 91.8 indicates the elapsed time from when the ice capsule contacts on the plate. For the experiment, the distance from the dropping location of the ice capsule to the flame is 120 mm, and the weight of the extinguishing powder filled into the capsule is 3 g.

891

From Fig. 91.8, it can be seen that after the ice capsule crushes, the extinguishing powder is scattered radially but not uniformly and there are some trajectories drown with the extinguishing powder. As shown in Fig. 91.7, the extinguishing powder released from the crushed ice capsule follows the motion of the ice fragment. Therefore, the trajectories of the extinguishing powder in Fig. 91.8 mean the path along which the ice fragments move over the plate. Under the free-fall condition of this study, the impact velocity is too small to break the ice capsule into small pieces. The nonuniformity of the crushed ice ball is considered to lead the nonuniformity of the distribution of the scattered extinguishing powder. If the higher impact velocity is given to the ice ball by using a gas gun, it would be possible to spread the extinguishing powder more uniform and wider from the impact location of the ice ball. Although the powder is not released uniformly from the broken ice capsule under this experimental condition, some extinguishing powder reaches the flame region, and then the flame is extinguished. Moreover, from Fig. 91.8, it is found that the diffusion flame starts to be blown off from the bottom region, that is, the flame base. The flame base is the most important part of the flame because the stability of a diffusion flame formed over a burner, which determines whether the flame blows off completely or not, is controlled by the flame base behavior.

91.3.3 Flame Base Behavior in the Extinguishing Process Fig. 91.7 Motions of the fragment produced and the extinguishing powder released from the broken ice capsule due to impact onto the iron plate Fig. 91.8 Sequential images of the extinguishing process by using the ice capsule filled with extinguishing powder (m ¼ 3 g, L ¼ 120 mm)

By using a digital camera (CASIO, EX-F1; frame rate, 300 fps and exposure time, 1/320 s), the motion of the

892

M. Ishidoya et al.

Fig. 91.9 Extinguishment process. (a) Success case (m¼3 g, L¼120 mm), (b) Failure case (m¼1 g, L¼150 mm)

flame base in the blowoff process with the extinguishing powder have been observed. Figure 91.9 shows a series of sequential images of the flame base behavior in the extinguishing process. At 0 ms in Fig. 91.9, the extinguishing powder arrives at the flame base. Figure 91.9a shows the success case of the flame extinguishment with extinguishing powder (m ¼ 3 g and L ¼ 120 mm). At 30 ms, the scattered extinguishing powder from the broken ice capsule is supplied to the bottom region of the flame, and at the same time, the local extinction at the flame base occurs due to interfering with the chemical chain reaction at the reaction zone of the flame. Then, at 57 ms, the flame starts to lift off from the burner because of losing the flame base. Finally, the flame is blown off completely at 103 ms. Figure 91.9b indicates the failure process of the flame blowoff. In this case, the distance from the flame to the impact location is 30 mm larger than that in the success case shown in Fig. 91.9a, and also the amount of the extinguishing powder in the ice capsule is 1 g. Therefore, the concentration of the extinguishing powder to reach the combustion region is considered to decrease more than that in the success case. At 60 ms in Fig. 91.9b, it is seen that the local extinction occurs at the flame base region and lifts off from the burner as well as the blowoff case. However, the

flame does not blow away to the downstream at 120 ms, and then the flame moves back to the burner at 227 ms. Eventually, the flame is restabilized on the burner. The flame motion observed at 227 ms in Fig. 91.9b is caused by the propagation of the leading edge flame into the premixing layer between methane and air formed in front of the lifted flame observed at 120 ms. Based on the above examination, it is said that when the extinguishing powder having the concentration enough to suppress the chemical reaction in the combustion zone can be delivered from the crushed ice capsule to the flame base region attached on the burner and then lifted from the burner, the extinguishment of the whole flame is achieved.

91.3.4 Extinguishing Probability To evaluate quantitatively the extinguishing effectiveness of the ice capsule filled with ABC powder agent, the blowoff probability, P, was measured by varying the amount of the extinguishing powder, m, and the distance between the flame and the impacting location, L. Figure 91.10 shows the profiles of the extinguishment probability. At m ¼ 0 g in Fig. 91.10, that is, air packed in the ice capsule, the extinguishing probability always shows zero

91

Examination of Extinguishment Method with Extinguishing Powder Packed in a Spherical Ice Capsule

Fig. 91.10 Extinguishing probability profiles as a function of the distance between the impacting location and the flame

independent of the distance, L. Although small fragments produced from the crushed ice capsule sometimes contact with the flame, the flame was not extinguished by the ice fragments. Therefore, the ice capsule without the extinguishing agent has no extinguishing effectiveness. On the other hand, for m ¼ 1 g, the extinguishing probability shows unity when the distance from the impact point of the ice capsule is less than 50 mm. As the distance increases, the extinguishing probability decreases and then becomes zero at 150 mm. For m ¼ 3 g, the distribution of the extinguishing probability indicates the similar tendency to that for m ¼ 1 g. From the probability profiles in Fig. 91.10, we can define the limitation value of the extinguishable distance, Lex, within which the probability shows unity and the flame is always extinguished perfectly. In other words, Lex means the maximum value of the effective range of the ice capsule filled with the extinguishing powder. From Fig. 91.10, it is found that Lex in m ¼ 3 g is 120 mm and 2.4 times larger than that in m ¼ 1 g.

91.4

Concluding Remarks

In the present study, the extinguishment method with extinguishing powder packed in the hollow capsule made of ice has been proposed and investigated experimentally. If a capsule can be used to deliver an extinguishing powder, it will be possible to transport the agent to the fire area over a

893

long distance. Therefore, the extinguishing method is considered to be useful to mitigate and minimize the damage of post-earthquake fires because after a large-scale earthquake occurs, fire engines and also local residents cannot approach to the fire area due to the damaged road or tsunami debris. In this experiment, a hollow ball made of ice is used as a capsule to transport the extinguishing powder. The ice ball could be ejected at the high velocity with low-pressure gas gun like paintball gun. ABC extinguishing powder is filled into the ice capsule. A methane-air diffusion flame formed on a tube burner is used as the extinguished target. As a first step of the examination of the ice-capsule extinguishing method, the capsule is dropped in free fall from the height of 1.8 m in the extinguishing experiment. The ice capsule is formed by using the double rotational axis casting machine and with water of 1.8 cm3. The ice capsule has 20 mm in outer diameter. ABC extinguishing powder can be filled into the ice capsule up to 3 g. The impact velocity onto the metal plate, in which the burner is embedded, is 5.8 m/s. The extinguishment of the flame which is 345 mm high has been succeeded using the extinguishing powder filled into the ice capsule. The maximum effective distance in the extinguishing method, that is, the distance from the burner center to the dropping location of the ice capsule, is 120 mm. Within the circle of the radius of 120 mm from the dropping location, the extinguishing probability is always thought to show unity. Under free-fall condition of this study, the impact velocity is too small to break the ice capsule into small pieces uniformly. The nonuniformity of the crushed ice ball is considered to lead the nonuniformity of the distribution of the scattered extinguishing powder. If the higher impact velocity is given to the ice ball by using a gas gun, it would be possible to spread the extinguishing powder more uniform and wider from the impact location of the ice ball and increase the maximum effective distance of the extinguishing method.

References 1. Ewing CT, Faith FR, Hughes JT, Carhart HW (1989) Fire. Technology 25(2):134–149 2. Fischer, G., Leonard, JT (1995) Effectiveness of fire extinguishing powders based on small scale suppression tests (in publication) 3. Watanabe K, Torikai H., Ito A (2013) 24th international on transport phenomena. pp 664–670 4. Torikai H, Narita M, Ito A (2013) The 24th international on transport phenomena. pp 682–688 5. Torikai H, Murashita T, Ito A, Metoki T (2011) Fire safety scienceproceeding of the tenth international symposium. pp 557–568 6. Murashita T, Torikai H, Ito A (2012) Visualization of mechanical processes. doi:10.1615/VisMechProc.vl.i4.70

Experimental Study on Transformer Oil Pool Fire Suppression by Water Mist

92

Pei Zhu, Xishi Wang, Zhigang Wang, Haiyong Cong, and Xiaomin Ni

Abstract

Oil-insulated transformer is the main source of power substation fire. It is important therefore to investigate the fire suppression of transformer oil pool fires. Experiments were conducted in a 3
3
3 m room to study the transformer oil pool fire suppression by water mist. The square pool fire with dimensions of 17, 25, and 30 cm and a downwarddirected single-injector nozzle with operating pressures of 1.0, 2.5, and 4 MPa were considered in the experiments. The flame shape during water mist application was recorded by a video camera. The temperature, CO concentration, and flame thermal radiation were measured to evaluate the fire suppression process by water mist. The results show that the flame would be intensified due to the injection of water mist at the initial period, and the intensification phenomenon is related to the fire size and the injection pressure of water mist system. The flame intensification occurred obviously under low injection pressure and large-size fire conditions. In addition, the fire extinguishment time increases with the decrease of the injection pressure and the increase of pool size. The results of this work would be valuable for optimizing the water mist system in the application of transformer substation fire suppression. Keywords

Transformer oil  Pool fire  Fire suppression  Water mist

92.1

Introduction

With the development of social economy and electricity industry, the coverage of transformer substations expanded incrementally, and oil-immersed transformer played an essential part in large transformer substations. The fire risk of transformer oil fire has been increased due to overheat, leaking out, etc. The damage and loss of transformer oil fire are extremely very large, due to the fire releasing lots of heat and toxic gases [1, 2]. So it is important and necessary to

P. Zhu  X. Wang (*)  Z. Wang  H. Cong  X. Ni State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei 230026, China e-mail: [email protected]

study the protection and suppression technologies for transformer oil fire. Water mist being considered as a clean and efficient fireextinguishing agent has been widely studied and applied [3– 8]. Especially, water mist systems have been proven to be effective in suppressing transformer fires in certain situations [9]. Fire suppression using water mist is a complex problem, which involves several combined physical phenomena [10, 11]. The main effects usually observed in water mist action are gas-phase cooling, oxygen displacement, fuel vapor dilution, and wetting and cooling of the fuel surface [12]. Wang et al. [13] conducted tests with kerosene, alcohol, and heptane pool fires and found that in the cases of heptane and ethanol pool fires, water mist suppressed and extinguished them quickly through cooling and oxygen displacement. But in the case of kerosene, combustion was

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_92

895

896

P. Zhu et al.

enhanced at the beginning of water mist application due to impediment of soot particles, evaporation expansion, and the chain reactions. The studies about water spray suppression and intensification of high-flash-point hydrocarbon pool fires by Ho [14] indicated that flame intensification occurred to the mineral seal oil and cooking oil pool fires, which are class IIIB liquids with a flash point at or above 93  C. Cong et al. [15] investigated extinction limit of diesel pool fire and found that the fuel surface cooling is the dominant mechanism at extinction limit and fire intensification was dependent on the effective water flux and the plume-spray thrust ratio. However, the suppression of transformer oil pool fire by water mist has been seldom studied. So experimental study on water mist suppression of transformer oil pool fire was conducted. The results would provide some guidance for water mist applied on transformer substation facilities.

92.2

Experimental Apparatus and Materials

The schematic diagram of the experimental apparatus is shown in Fig. 92.1. A downward-pointing solid-cone water mist nozzle was installed 2.0 m above the fuel pan. The flame and plume temperatures were measured with seven K-type thermocouples with bead diameters of 1 mm and a resolution of 0.1  C, which are located between 0.0 m and 1.05 m above the fuel pan with an interval of 0.15 m. A radiative heat flux meter was installed 0.45 m above the fuel pan and at a horizontal distance of 0.6 m from the center of the fuel pan. The flue gas analyzer to measure the CO concentration was installed 1.05 m above the fuel pan and

at a horizontal distance of 0.25 m from the center of the fuel pan. An electric balance with sampling intervals of 0.1 s and a resolution of 0.01 g was positioned below the fuel pan to measure the mass loss rate without water mist. The square pool fires with dimensions of 17, 25, and 30 cm were tested in the experiments. The 25# transformer oil was used as the fuel and whose properties can be seen in Table 92.1. Because the transformer oil is a high-flash-point hydrocarbon fuel, it is not easy to be ignited, so a small amount of N-heptane fuel was used to ignite it. The experimental test cases were shown in Table 92.2. In each test, the water mist activated after the fuel pre-burn for about 120 s.

92.3

Water Mist Characteristics

Water mist was discharged from a single fluid-type nozzle prototype with seven heads made of stainless steel. Only the downward head was used in the experiments. The diameter of the orifice is 0.8 mm. The droplet size and velocity were obtained by a PDA (phase Doppler anemometry) system at the cross section 1.0 m away from the nozzle exit. Typical droplet size distribution at the radial Table 92.1 The properties of transformer oil (25#) Property Appearance Density (20  C), kg/m3 Pour point,  C Condensation point,  C Flashing point,  C

Fig. 92.1 Schematic diagrams of the experimental apparatus

Characteristics Light-yellow transparent liquid 895 22 25 140

Water mist nozzle gas analyzer K-Thermocouple

Radiation heat fLux meter

camera

oil pump

fuel pan thermal baffle Computer

Data acquisition system

92

Experimental Study on Transformer Oil Pool Fire Suppression by Water Mist

Pool size (m) 17
17 25
25 30
30 17
17 25
25 30
30

Initial fuel mass (g) transformer oil + N-heptane 150 + 15 300 + 20 500 + 25 150 + 15 300 + 20 500 + 25

Injection pressure (MPa) Without Without Without 1.0/2.5/4.0 1.0/2.5/4.0 1.0/2.5/4.0

20 15 10 5 0 0 15

Size distribution (%)

2

4

6

v

5 0 –1

0

1

2

20 20

Injection pressure 1.2 MPa d32= 267.25 mm

15

10

%

a

u

10 %

Case code 1 2 3 4 5 6

%

Table 92.2 The experimental test cases

897

3

w

10 0 –1

0

1 v(m/s)

2

3

Fig. 92.3 Droplet velocity distribution measured at the centerline and 1.0 m below the nozzle

5 Table 92.3 Water mist characteristics

0 0

100

200

300

400

500

Size (µm)

Size distribution (%)

b 10

Injection pressure:1.2MPa 8 d32 = 276.40 mm

Injection pressure (MPa) 1 2 3 4

92.4

6

Flow rate (L/min) 1.5 2.2 2.8 3.2

Spray angle ( ) 79 75 73 72

Results and Discussion

92.4.1 Transformer Oil Fire Without Water Mist

4 2 0 0

100

200

300

400

500

Size (µm) Fig. 92.2 Droplet size distribution measured in different radial locations (a) 0 cm from the centerline and (b) 15 cm from the centerline

locations of 0 cm and 15 cm under an injection pressure of 1.2 MPa is shown in Fig. 92.2. The droplet velocity distribution is shown in Fig. 92.3, and it shows that the mean velocities of U-, V-, and W-direction are 3.82, 0.77, and 0.86 m/s, respectively. Its other characteristics can be seen in Table 92.3.

The transformer oil fire combustion and flame characteristics were measured. Figure 92.4 shows the typical temporal evolution of temperature measured above the fuel pan of a 25
25 cm pool fire, and it can be seen that the maximum flame temperature is about 770  C. For about 120 s after the fuel ignited, the burning was almost kept in the steady state. So in the latter cases with application of water mist, the activation time of water mist system was also at this time. Figure 92.5a–c shows the temporal evolution of mass loss rate, CO concentration, and radiative heat flux for different pool fires without water mist. It can be seen that the mass loss rate, smoke production, and thermal radiation from transformer oil fire have large difference for fires with different pool size, i.e., the larger the pool size is, the higher the CO concentration and the radiative heat flux would be.

898

P. Zhu et al.

The fire suppression or flame intensification phenomenon occurred for different fire sizes and water mist work

No mist L=25cm

900 800

Distance above the fuel pool 0 cm 15 cm 30 cm 45 cm 65 cm 85 cm 105 cm

700

T( )

600 500 400 300 200 100 0 0

100

200

300 400 500 Time (s)

600

pressures. Figure 92.6 shows the results of experiment for fire extinguishment time. It can be seen that for pool fire with a dimension of 30 cm and water mist injection pressure of 1.0 MPa, the fire suppression was failed, while the others were successful. The fire extinguishment time increased

25 20 14

15 9

10

9 Failure 5

5

5 3

0 17cm

30cm

25cm

Fig. 92.6 The results of fire extinguishment time

17cm 25cm 30cm

3.0 2.5 2.0 1.5 1.0

co concentration(ppm)

b

4.0 3.5

m'(g/s)

13

700

Fig. 92.4 Typical temporal evolution of temperature above the fuel pan (L ¼ 25 cm)

a

30

4MPa 2.5MPa 1MPa

30 Fire extinguishment time (s)

92.4.2 Fire Suppression and Flame Intensification with Water Mist

17cm 25cm 30cm

100 80 60 40 20 0

0.5 0.0 100

200

400 Time(s)

Radiative heat flux(kw/m2)

c

500

600

0

100

200 300 Time(s)

400

500

6 5

17cm 25cm 30cm

4 3 2 1 0 0

100 200 300 400 500 600 700 800 Time(s)

Fig. 92.5 Typical temporal evolution of measured parameter for transformer oil fires with different pool size: (a) mass loss rate, (b) co concentration, and (c) radiation heat flux

92

Experimental Study on Transformer Oil Pool Fire Suppression by Water Mist

899 4

Radiative heat flux(kw/m2)

4

3

1.0MPa

1

2.5MPa

110 115 120 125 130

4.0MPa

2

1 L=25cm Water mist activation 0

a

No mist

2 0

0

Fig. 92.7 Typical temporal evolution of flame image after water mist activation for 25 cm pool fire with injection pressures of (a) 1.0 MPa and (b) 4.0 MPa

3

40

80

120 Time(s)

160

200

Fig. 92.9 Temporal evolution of radiative heat flux for 25 cm pool fire under different injection pressures

with the increase of the pool size as well as the decrease of the injection pressure of water mist. For the cases of fire suppression, the flame size and 700 Distance above behavior after water mist activation were affected by the fuel pool 600 injection pressure. Figure 92.7 shows the flame image of 0 cm 500 25 cm pool fire. For an injection pressure of 1.0 MPa, as 15 cm 400 30 cm shown in Fig. 92.7a, the flame sizes are always suppressed 45 cm by the water mist through the whole fire extinction process. 300 65 cm But for 4.0 MPa case, the flame size enlarged rapidly at the 200 85 cm initial period of water mist injection, as shown in 100 105 cm Fig. 92.7b, especially in lateral direction, which may be 0 also deemed as a flame intensification, although the fire 0 20 40 60 80 100 120 140 160 180 200 extinction time is short. Figures 92.8 and 92.9 show the Time(s) temporal evolution of temperature and radiative heat flux for the fire under different injection pressures. Since the b Fire extinction thermocouple trees were put at the centerline of flame Water mist activation 800 L= 25 cm plume above the fuel pan, the temperature shows no clear Distance above increase after water mist activation (shown in Fig. 92.8), 700 P L=4MPa fuel pool although the flame intensification occurred. From Fig. 92.9, 0 cm 600 it can be seen that the radiative heat flux suddenly 15 cm 500 30 cm increased as soon as the water mist activated. The reason 400 45 cm for the slight flame intensification is that the air entrain65 cm 300 ment was enhanced and the flame was deflected due to the 85 cm 200 high velocity of mist droplets. So the enlargement of flame 105 cm by water mist should be avoided even though fire can be 100 suppressed. 0 For the failed case of fire extinction, the flame was 0 20 40 60 80 100 120 140 160 180 200 usually intensified due to water mist activation. To the Time(s) relative large pool size of 30 cm and relative small spray Fig. 92.8 Temperature variation after water mist activation with injec- injection pressure of 1 MPa, the flame intensification obvition pressures of (a) 1.0 MPa and (b) 4 MPa ously occurred which is mainly due to the spray-induced oil T(°C)

T(°C)

800

L= 25 cm Water mist activation P L=1MPa

Fire extinction

900

P. Zhu et al.

a

a

12

12

0s

1s

b 900 800

L=30cm P L=1MPa

700 T(°C)

600 500 400 300

3s

12s

29s

71s

Water mist activation Distance above the fuel pool 0 cm 15 cm 30 cm 45 cm 65 cm 85 cm 105 cm

200

Radiative heat flux(kw/m2)

10 8

9

No mist

6

1.0MPa

3

2.5MPa 4.0MPa

0 115

120

125

130

6 4 2 L=30cm

0

Water mist activation –2 0

20

40

60

80

100 120 140 160 180 200 Time (s)

b

100

350 0

50 100 150 200 250 300 350 400 450 500 Time (s)

Fig. 92.10 Typical characteristic of 30 cm pool fire with water mist (PL ¼ 1.0 MPa). (a) Temporal evolution of flame image after water mist activation. (b) Temperature variation

splattering. In the observation, oil splash phenomenon was very obvious, and the flame height increased largely after water mist activation, as shown in Fig. 92.10a. Similarly, from Fig. 92.10b, it can be seen that the flame temperatures all increased first after water mist activation and fluctuated at relative high temperature. Figure 92.11 shows the radiative heat flux and CO concentration for 30 cm pool fire under different water mist injection pressures. Figure 92.11a shows that the flame intensification can be reflected by the variation of radiative heat flux and for the large injection pressure, the flame intensification also occurs due to the enlargement of flame in lateral direction. From Fig. 92.11b, it can be seen that the CO concentration increased slightly when water mist activated. But for the failed case, it increased largely due to the incomplete combustion by water mist. From the above results, it can be seen that for larger pool fire and lower injection pressure water mist, the flame intensification occurred easily. So the flame intensification phenomenon in the practical water mist system applied on the transformer oil fire should be avoided. In addition, the appropriate water mist characteristic should be carefully selected for suppressing transformer oil fire.

CO concentration(ppm)

0

Water mist activation

300

Failed

No mist 1.0MPa 2.5MPa

L=30cm

250

4.0MPa

200 150 100 50 0 0

50

100

150 200 Time (s)

250

300

350

Fig. 92.11 Temporal evolution of radiative heat flux (a) and CO concentration (b) under different injection pressures for 30 cm pool fire

92.5

Conclusions

The experiments about transformer oil fire suppression by water mist have been conducted. The following conclusions can be drawn: 1. Fire extinguishment time increased with the increase of pool size and the decrease of injection pressure of water mist system. 2. The flame size and behavior of transformer fire after water mist activation are related to the pool size and water mist injection pressure. With small injection pressure, the flame size would be enlarged obviously.

92

Experimental Study on Transformer Oil Pool Fire Suppression by Water Mist

3. Strong flame intensification phenomenon occurred to large pool size fire and low injection pressure of water mist system. All of the flame height, CO concentration, and radiative heat flux increased as the intensification phenomenon occurred. 4. Appropriate water mist characteristics should be carefully optimized for suppressing transformer oil like liquid pool fires.

Acknowledgment The authors appreciate the Natural Science Foundation of China (Grant No. 51323010), the Fundamental Research Funds for the Central Universities (WK23200000035), and the Anhui Provincial Natural Science Foundation (Grant No. 1408085MKL95).

References 1. Petersen A, Blanc R, Carrander K et al (2012) Guide for transformer fire safety practices, Working Group A2.33 2. Duarte D (2012) Aspects of transformer fires in Brazil. Open J Saf Sci Technol 2:63–74 3. Gupta M, Pasi A, Ray A, Kale SR (2013) An experimental study of the effects of water mist characteristics on pool fire suppression. Exp Thermal Fluid Sci 44:768–778

901

4. Huang X, Wang XS, Liao GX (2011) Characterization of an effervescent atomization water mist nozzle and its fire suppression tests. Proc Combust Inst 33:2573–2579 5. Husted BP, Petersson P, Lund I, Holmstedt G (2009) Comparison of PIV and PDA droplet velocity measurement techniques on two high-pressure water mist nozzles. Fire Saf J 44:1030–1045 6. Mawhinney JR, Richardson JK (1997) A review of water mist fire suppression research and development. Fire Technol 1:54–90 7. Wang XS, Liao GX, Yao B, Fan WC, Wu XP (2001) Preliminary study on the interaction of water mist with pool fires. J Fire Sci 19:45–61 8. Wang XS, Zhao XD, Zhang Y, Cai X, Gu R, Xu HL, Liao GX (2009) Experimental study on the interaction of a water drop impacting on hot liquid surfaces. J Fire Sci 27:545–559 9. Bureau, Hydroelectric Research and Technical Services Group (2005) FIST 3-32, Transformer fire protection 10. Grant G, Brenton J, Drysdale D (2000) Fire suppression by water sprays. Fire Saf J 26:79–130 11. Liu Z, Kim AK (2000) Review of water mist fire suppression systems—fundamental studies. J Fire Prot Eng 10:32–50 12. Mawhinney JR, Back GG (2002) Water mist fire suppression systems. In: SFPE Handbook of fire protection engineering, pp 4.311–4.337 13. Wang XS, Liao GX, Qin J, Fan WC (2002) Experimental study on the effectiveness of the extinction of a pool fire with water mist. J Fire Sci 20:279–295 14. Ho San-Ping (2003). Water spray suppression and intensification of high flash point hydrocarbon pool fires, doctoral thesis. Worcester Polytechnic Institute 15. Cong BH, Liao GX, Huang Z (2009) Extinction limit of diesel pool fires suppressed by water mist. J Fire Sci 27:5–26

Inhibition of Propane/Air Premixed Flame by Water Mist

93

Toichiro Okawa, Wataru Ebina, Hiroyoshi Naito, and Akira Yoshida

Abstract

The effects of fine water mist on flame temperature and laminar flame speed of propane-air mixtures were investigated both experimentally and numerically. In experiments, flame temperature and laminar flame speeds were measured using a single jet-plate configuration for the cases with and without water mist. The numerical simulation was also performed using the OPPDIF code in CHEMKIN package. To include the phase change with evaporation, the evaporation process was assumed as a chemical reaction of which rate constant follows the Arrhenius law. For the case without water mist, experiments showed that the flame speeds increase with the stretch rate toward the limit of extinguishment. This tendency was fairly reproducible by the numerical simulation with Davis-Law-Wang kinetic mechanism. The flame temperature was measured by a fine wire thermocouple. The flame temperature decreases with the water mist addition toward the limit of extinguishment within the range of φ from 0.8 to 1.2. The reduction of the flame temperature is more enhanced for lean and rich mixtures than the stoichiometric one. On the other hand, in terms of inhibition of the flame speed, the water mist is more effective for lean mixture than rich one. Reduction of flame temperature in rich mixture does not contribute to decrease the flame speed. Additionally, the stretch rate also decreases the flame temperature. Therefore, for lean mixtures, an appropriate combination of water mist and stretch rate can enhance the suppression effectiveness of water mist. Keywords

Water mist
Laminar flame speed
Flame stretch
CHEMKIN

93.1

Introduction

Since the production of ozone-depleting fire suppressing agents, e.g., Halon 1301 (CF3Br) and Halon 1211 (CF2ClBr), has been banned, effective Halon replacements

T. Okawa
W. Ebina
A. Yoshida (*) Tokyo Denki University, 5 Senju-Asahicho, Adachi-ku, Tokyo 120-8551, Japan e-mail: [email protected] H. Naito Fire and Disaster Management Agency, 2-1-2 Kasumigaseki, Chiyoda-ku, Tokyo 100-8927, Japan

are being searched [1]. Then, C2HF5, C6F12O, and 2-BTP (CF3CBrCH2, 2-bromo-3,3,3-trifluoropropene) were proposed as alternatives of Holon. However, they were found to enhance the premixed flame burning depending on the equivalence ratio of the flammable mixture, instead of inhibiting, by increasing the laminar flame speed or extending the flammability limit [2], compared to the base agent, CF3Br, of which inhibition characteristics have been clearly identified [3]. Water mist is a favorable substitute for typical halogenated hydrocarbon fire suppressants, because water is ubiquitous, inexpensive, nonelectrically conductive, and environmentally acceptable. In addition, water mist is fairly effective to suppress fires and to mitigate explosions

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_93

903

904

[4–7]. Adding water mist in a reactive mixture causes significant changes in flame properties by the three following mechanisms: (a) thermal effect due to the absorption of heat, (b) dilution effect caused by the reduction in reactant concentrations, and (c) chemical effect owing to the activity of water vapor that may alter some reaction paths. Fine water mist enhances these effects due to significant increase in surface area available for heat absorption and evaporation. These three mechanisms are concomitant and closely linked with each other. In addition, the flame stretch also affects the flame properties and extinguishing process. However, few studies exist in the literature relating to the effect of water mist on stretched flame from the point of view of fire suppression and explosion mitigation. Gaseous water vapor has been long recognized to be effective to suppress a fire. The inhibiting effects of gaseous water vapor on the laminar flame speed of methane flames were investigated [8], and the numerically predicted reduction in flame speed was in good agreement with the experiments. In addition, the water vapor was found to have chemical inhibiting effect on the combustion reactions of H2/CH4/Air mixtures; however, it was found to be small but not negligible [9]. Effects of elevated temperatures and pressures on the laminar flame speed of H2/O2/water vapor system were investigated both experimentally and computationally [10], and the addition of water vapor was found to cause a significant reduction in the flame speed. In the course of our project of water mist fire suppression carried out in our laboratory, the effect of water mist and its inhibition mechanism are being investigated in various situations including premixed and diffusion flames [11– 17]. Anticipated results will be of help for an actual design of an effective water mist suppression system. Liquid water has a more favorable thermal property for fire suppression than gaseous vapor, because it has a high latent heat of evaporation and can absorb a significant quantity of heat from flames. Therefore, water mist should be more effective in reducing the flame speed of premixed flames than other gaseous thermal agents (N2 and CF4) or chemical agents (CF3Br) and also more effective than the same mass of gaseous water vapor [18]. Furthermore, the flame speeds of propane-air premixed flame stabilized in the stagnation flowfield and influenced by water mist were measured [14], and the dependency of flame speed on stretch rate was found to change from positive to negative by the addition of water mist. In the diverging flowfield, the mist droplets accumulate around the stagnation streamline due to the Stokes number effect acting on mist droplets, so that the increased local mass loading of water mist will make easier to extinguish the flame than the uniformly dispersed water mist. Laminar flame speed is a fundamental property of a flammable gaseous mixture describing the overall reaction rate, heat release, and heat and mass transport in the flame,

T. Okawa et al.

and as such many efforts have been devoted to measure or predict the precise laminar flame speeds of various kinds of fuels. For the flame speed measurement, the counterflow and opposed-jet techniques were well documented [19–26] and have been traditionally implemented with the use of twin opposed-nozzle or single jet-plate configurations. In a twinflame or single-flame configuration, the velocity minimum is identified in the velocity profile of unburned mixture as a reference upstream flame speed SL, and the velocity gradient a ahead of the minimum point is identified as the stretch rate K (¼ a for an axisymmetric flame) experienced by the flame. Then the unstretched laminar flame speed SL0 is obtained by systematically determining the dependence of the reference flame speed SL on the stretch rate K and extrapolating SL to zero K. The reduction in laminar flame speed is frequently used as an indicator of the fire suppression effectiveness of an inhibiting agent [27–29]. The impact of gaseous water vapor or liquid water mist on stretched laminar flame has been the subject of a relatively limited number of studies. In the present study, the effects of water mist on the laminar flame speed and flame temperature of propane/air mixtures were investigated both experimentally and numerically. In the experiments, stretched laminar premixed flames were established in the stagnation flowfield produced by a mixture flow emerging from a nozzle impinging on a flat plate. The unstretched laminar flame speed S0L is obtained by a linear or nonlinear extrapolation to zero stretch. The effect of water mist on flame speed and flame temperature of highly stretched premixed flame was also simulated numerically by using OPPDIF code in CHEMKIN package, modified to include the evaporation process, which is assumed to be a chemical reaction.

93.2

Numerical Simulation

The OPPDIF code in the CHEMKIN package was utilized to simulate the flame speed and flame temperature of stretched, adiabatic, laminar propane/air premixed flames stabilized in the stagnation flowfield. The flame speeds estimated by CHEMKIN depend on the reaction mechanism adopted. At present, there are three mechanisms applicable to propane combustion. These are the NUI Galway mechanism [30], the San Diego mechanism [31], and that proposed by Davis, Law, and Wang [32]. In the NUI Galway mechanism, 230 chemical species and 1326 elementary reactions are involved. On the other hand, in the San Diego mechanism, 46 chemical species and 235 elementary reactions are considered, whereas 71 species and 469 reactions are included in the DLW mechanism. In preliminary calculation, the DLW mechanism was found to give the same flame speeds as the NUI Galway mechanism. In addition, the San Diego mechanism provided lower flame speeds for rich mixtures

93

Inhibition of Propane/Air Premixed Flame by Water Mist

than those experimentally obtained [16]. Consequently, the DLW model was adopted in the present simulations of flame speed and flame temperature. As the reference, case , the flame speed and flame temperature were calculated for the case without water mist. In the calculation with water mist, case , the latent heat of evaporation was included in the detailed kinetic mechanism. Since the phase change in evaporation process cannot be treated by the OPPDIF program, the liquid water mist was assumed to be an imaginary ideal gas, following the model proposed by Takahashi and Katta [33]. This imaginary ideal gas was identified as the water mist gas. The conversion of this water mist gas into gaseous water vapor was chemically treated as an Arrhenius reaction, of which rate was expressed by k ¼ Aexp(-E/RT), where A is the pre-exponential factor and E is the activation energy. Thereby, the evaporation process was described by a chemical reaction, water mist gas !1354H2O. Since the water mist gas, as well as the water vapor, was assumed to be an ideal gas, the expansion due to molar change was included in this chemical reaction. The thermodynamic data of water mist gas was obtained by scaling the corresponding data of liquid water with the ratio of molecular masses. Activation energy, E, and pre-exponential factor, A, should depend on the evaporation process, i.e., the water mist mass loading and the mist diameter. In Ref. [33], these two parameters were obtained while calibrating the global reaction based on the criterion that water should fully evaporate by the time when the gas temperature reaches Tev ¼ 500 K. It was found that the increase of Tev from 500 to 1500 K induced no significant changes in the flame structure and flame speeds [34], and therefore Tev ¼ 500 K was assumed throughout the present study. Since liquid water is actually getting converted into H2O vapor, latent heat of evaporation is automatically included. Thus, the effects of expansion, latent heat, and chemical kinetics of water mist can be considered by this evaporation model.

93.3

Experimental

The implementation of the counterflow, opposed-jet techniques for the determination of laminar flame speeds, and stretch rates at extinguishment has been reported in detail [19–24, 35–40]. So far, a twin flame [19–23] or a single flame [24] has been used for the determination of laminar flame speeds and stretch rates at extinguishment of H2 and C1-C3 hydrocarbon fuels. In the present study, experiments were performed, using the single jet-plate configuration, to measure the flame speed and flame temperature for atmospheric propane/air flames at different equivalence ratios with and without water mist in the mixture. The experimental apparatus, shown in Fig. 93.1, consisted basically of an axisymmetric burner and an

905

Fig. 93.1 Experimental apparatus

axisymmetric stagnation plate. The axis of the burner was vertical, and the upward jet impinged on the horizontal stagnation plate, which was cooled by water to retain the plate temperature constant. To ensure that the experimental results were not affected by the specific nature of the stagnation plate, the cooled plate was replaced by an uncooled ceramic plate. Then, the minimum velocity at the flame and the burned gas temperature immediately after the flame were not changed by the plate replacement. Boundary layer effects were also found to be negligible as long as the flame was not too close to the stagnation plate. Water mist diameters and their distributions were measured by a phase Doppler particle analyzer (PDPA) and the flow velocity by a laser Doppler velocimetry (LDV). For these measurements, the plate was chamfered with a slant of 3 to enable the laser beams to make the closest approach on the flame near the plate. The plate was mounted on a traverse system to allow easy and accurate movement relative to the burner. The burner plate assembly was mounted on another traverse system, whereas the LDV and PDPA systems were fixed on the ground. The inside diameter of the nozzle exit was 45 mm. The mixture flow was surrounded by a shrouding airflow to protect the flame from the disturbances caused by the entrainment of surrounding air. Propane was used as a fuel throughout. For PDPA and LDV measurements, the mixture flow should be seeded with light-scattering particles. For the case without water mist, the mixture flow was seeded with aluminum oxide (Al2O3) particles of a nominal diameter of 1 μm, generated by a fan-stirred particle generator. When the water mist was added to the mixture flow, the water mist itself played a role of light-scattering particles. In the present study, the pure water was used for generation of water mist to avoid the flame reaction of metal compounds in tap water. To generate fine water mist, six

906

piezoelectric atomizers were installed in the mist chamber. These atomizers could be operated separately to control the mist flow rate. The water mist flow rate was controlled by the number of piezoelectric units activated. The mixture gas flow was separated into six paths in the mist chamber. Each path was bent so that the mixture gas flow swept through the hovering water mist produced on each piezoelectric unit and entrained mist that was then carried up to the exit of the mist chamber. Nominal mist supply rate for one piezoelectric unit was 575 ml/h at 1.6 MHz. For the piezoelectric atomizers, the mist size distribution and the supply rate depend in general on the water level above the unit and the water temperature. The water level was maintained constant automatically by a level controller, 40 mm above the piezoelectric plate. The mechanical vibration of the piezoelectric plate generates heat that is transferred into the surrounding water, resulting in a slow rise in water temperature. To avoid this effect, water was heated by an electric heater, and its temperature was kept constant at 313 K, 20 K above the room temperature. We confirmed that, in our experiments, the change of water temperature was statistically insignificant. Water mist diameters and their distributions were measured by a PDPA, and the number mean diameter D10 was 11.5 μm, and the Sauter mean diameter D32 was 18.4 μm with a wide range of the size distribution ranging from 1 to 60 μm. Accounting for the loss trapped by the damping screen and the burner wall, the maximum mist flow rate actually delivered to the flame was 20.3 ml/min. Temperature measurements were performed using a silica-coated 50 μm thermocouple of Pt-PtRh13% or PtRh6%–PtRh30% elements depending on the temperature range measured. The thermocouple configuration was similar to that used in Ref. [41]. The thermocouple wire was extended between the prolonged supports with 40 mm span. Weak tension was given to the wire by two springs to prevent the deformation of wire, when inserted in hightemperature gas flow of which momentum is not negligibly small. The thermocouple junction and wire element were silica coated to minimize the catalytic effect of platinum. Radiation loss was corrected by a common manner [42]. The lifetime of a thermocouple was relatively short, typically 20 min, especially when it was used in the flame zone or in high-temperature burned gas region, and therefore two or three thermocouples were used in a series of experiments. Even though carefully treated, each thermocouple afforded a slightly different flame temperature from the other due to small differences in junction diameter and thickness of coating, which resulted in the temperature difference of about 50 K.

T. Okawa et al.

To minimize the perturbation of flame due to the insertion of a thermocouple, the temperature measurement was made along 10 mm off axis of the axisymmetric mixture flow, and therefore strictly speaking, the temperature distribution reported here was not along the stagnation streamline. However, the off-axis placement was found not to induce a significant difference from that on the stagnation streamline (within 5 % uncertainty).

93.4

Results and Discussion

93.4.1 Flame in the Stagnation Flowfield Figure 93.2 shows a typical photograph of the propane/air premixed flame stabilized in the stagnation flowfield. The equivalence ratio ϕ is 1.0, the uniform velocity of the mixture at the nozzle exit u is 2.6 m/s, and the distance between the burner exit and the stagnation plate L is 20 mm, i.e., the velocity gradient a ¼ 100 s1, which is equal to the stretch rate K for the axisymmetric stagnation flow. The presence of the stagnation plate modifies the pressure field and as such the velocity profile, producing a slightly dish-shaped flame when the flame was not close to the stagnation plate. The center portion of the flame, however, can be assumed to be planar and perpendicular to the stagnation streamline. When ϕ > 1.4, the cellular instability appeared at small velocity gradient K for large L and small u. Additionally, when ϕ < 0.7, the flame was deformed significantly and not stationary at small K. The experimental conditions, therefore, were limited within the range of equivalence ratio of 0.8 < ϕ < 1.3.

93.4.2 Stagnation Flowfield with and Without Water Mist The velocity profile along the stagnation streamline shown in Fig. 93.3 can be considered to be the superposition of the

Fig. 93.2 Direct photograph of a premixed flame (ϕ ¼ 1.0, K ¼ 100 s1)

93

Inhibition of Propane/Air Premixed Flame by Water Mist

Fig. 93.3 Velocity profiles along the stagnation streamline with and without water mist for ϕ ¼ 1.0

effects of the flame and the stagnation flowfield. When approaching the flame zone, the velocity decreases almost linearly with the distance from the stagnation plate z. The velocity gradient a ¼ du/dz (equal to the stretch rate K ) was obtained from the velocity profile along the stagnation streamline. Hence, the flame stretch rate K ¼ a was determined. The velocity abruptly increases in the flame zone due to thermal expansion and then decreases again toward the stagnation plate. The velocity at the point of initial temperature rise is the point where the curve starts to depart from the descending line due to thermal expansion, and the minimum point was defined as a reference upstream flame speed of a stretched flame SL, similarly to previous investigations by Law and co-workers [19–24]. The unstretched laminar flame speed SL0 can be subsequently determined by systematically extrapolating SL to zero K. By increasing the nozzle exit velocities or decreasing the separation distance, K increases, flames are pushed toward the stagnation plate, and extinguishment eventually occurred when a critical value Kext is reached. When the water mist is added to the mixture, the flame moves toward the stagnation plate to accommodate to the decrease of the flame speed. Similar velocity profiles were obtained for methane/air premixed flames with and without water mist [43]. For the case without water mist, the calculated velocity profile agrees well with experiment on the unburned gas side and in the flame zone. However, on the burned gas side, the measured velocity is significantly lower than the calculated value. In the present calculation, we used the 1-D plug flow assumption. However, the 1-D calculation was shown to overestimate the velocity in the high-temperature region [44]. Additionally, the thermal radiation of CO, CO2, and

907

H2O was not included in the simulation, resulting in the overestimation of the burned gas temperature and hence the higher velocity of burned gas. Furthermore, in the actual flame, the burned gas is cooled by the heat loss to the watercooled stainless steel stagnation plate, which should reduce the gas velocity. It is clear that the further accurate simulation needs the 2-D calculation. Nevertheless, the velocity minimum, which is the subject of the present study, can be well simulated by the 1-D calculation. When adding the water mist, the velocity profile shifts to downstream, whereas the velocity gradient in the upstream unburned mixture remains constant. Water mist makes the flame move apparently closer toward the stagnation plate than the simulation. In the flame zone, the decelerating flowfield changes to accelerating one through a velocity minimum. Inertial force acting on mist droplets is definite, and mist droplets cannot follow the change of flow from deceleration to acceleration due to the Stokes number effect as described in Ref. [11]. Therefore, the location of the velocity minimum where the equilibrium in velocity between mist droplets and gas phase was recovered due to local zero Stokes number moves toward the stagnation point compared to the prediction [15]. Of note is that the minimum velocity ahead of the flame zone is lower than that without water mist, suggesting that the laminar flame speed SL decreases when the water mist is added in the unburned mixture.

93.4.3 Stretched Laminar Flame Speeds Without Water Mist Stretched laminar flame speed, SL, is shown in Fig. 93.4a, b for lean and rich mixtures, respectively. Present experimental data and calculated values are concurrently presented in both figures. Calculated values are in fairly good agreement with experiments, indicating the nonlinear dependence of flame speed on stretch rate. Experimental data for ϕ ¼ 0.95 [20] are also shown in Fig. 93.4a and are in good agreement with the present study. Noteworthy is that the laminar flame speed increases with the stretch rate for all equivalence ratios tested.

93.4.4 Effect of Water Mist on the Temperature Distribution Temperature distributions along the stagnation streamline are shown in Fig. 93.5 for the cases with and without water mist added for K ¼ 100 s1. For comparison, calculated temperature distribution is also shown for the case without water mist. The difference between the calculated and measured flame position should be caused partly by the

908

T. Okawa et al.

Flame Temperature [K]

Laminar Flame Speed, SL [cm/s]

a

b

Stretch Rate, K [s-1] Fig. 93.4 Effect of stretch rate on laminar flame speeds without water mist. a Lean flames. b Rich flames

Equivalence Ratio, j Fig. 93.6 Flame temperatures measured experimentally and simulated numerically for K ¼ 100 s1 without water mist

Temperature [K]

flame settles down to the equilibrium position between the mixture velocity and the flame speed. Therefore, the flame zone moves closer to the stagnation plate, where the axial velocity of mixture decreases due to the flow divergence. Comparison of the flame temperatures measured experimentally and calculated numerically without water mist is shown in Fig. 93.6. In the calculation, the saturation vapor was considered. Temperature measurement includes intrinsically some error due to catalysis, radiation, and conduction. Catalysis and radiation losses were corrected using a common manner [42], whereas the conduction loss was not corrected, which should result in a lower temperature than the true one. Calculated temperature is at most 100 K higher than measured, resulting in 5 % uncertainty.

Distance from Stagnation Plate [mm] Fig. 93.5 Temperature distributions along the stagnation streamline with and without water mist for K ¼ 100 s1 and ϕ ¼ 1.0

perturbation of flame due to insertion of thermocouple and partly by the 1-D effect in calculation. For both cases, temperature increases sharply in the flame zone and gradually in the burned gas region due to the slow oxidization of carbon monoxide. Temperature far in the burned gas can be assumed to be uniform, indicating that the heat loss to the water-cooled stagnation plate is negligibly small. The uniform temperature in the burned gas region is defined as the flame temperature in the present study. With increase of the amount of water mist added, the flame temperature decreases due to the cooling effect of water mist, which in turn reduces the flame speed. In the stagnation flowfield, the

93.4.5 Effect of Water Mist on the Flame Temperature Effect of water mist on the flame temperature is shown in Fig. 93.7. With increase of the amount of water mist added, the flame temperature decreases up to 150 K for all the equivalence ratios due to the cooling effect of water mist. It should be noticed that the flame temperature is increased by the addition of the typical Halon replacements, i.e., 2-BTP and C2HF3, because each suppressant itself has its intrinsic burning nature [2], whereas water mist acts only as cooling substance. In this regard, water mist may be superior to other Halon replacements. The flame temperature normalized by that without water mist is shown for K ¼ 100 s1 in Fig. 93.8 as a function of the amount of water mist added. Flame temperature decreases gradually

Inhibition of Propane/Air Premixed Flame by Water Mist

909

Twith / Tw/o

Flame Temperature [K]

93

Equivalence Ratio, j Fig. 93.7 Effect of water mist on flame temperatures measured for K ¼ 100 s1

Fig. 93.9 Effect of stretch rate on reduction of flame temperature with water mist addition

Twith / Tw/o

SL,with / SL,w/o

Water Mist Flow Rate, Qw [ml/min]

Water Mist Flow Rate, Qw [ml/min] Fig. 93.8 Reduction of flame temperature by the addition of water mist for K ¼ 100 s1

with the amount of water mist and it suddenly drops near the limit of extinguishment. The decrease is larger for off-stoichiometric mixtures (ϕ ¼ 0.8 and 1.2), whereas it is rather small for near-stoichiometric mixtures (ϕ ¼ 0.9, 1.0, and 1.1). For both cases, the extinguishment occurs at a rather small temperature decrease, at most 0.92. The effect of stretch rate K on the normalized flame temperature Twith/Tw/o is shown in Fig. 93.9. With increase of K, the normalized flame temperature is decreased, similarly to the amount of water mist added. The effect is larger for lean mixture (ϕ ¼ 0.8) than rich mixture (ϕ ¼ 1.2). For

Equivalence Ratio, j Fig. 93.10 Effect of stretch rate on the reduction of flame speed for Y0 ¼ 0.03

ϕ ¼ 0.8, the highly stretched flame is easily extinguished at a smaller amount of water mist than weakly stretched flame, whereas this trend is not so clear for ϕ ¼ 1.2.

93.4.6 Effect of Stretch Rate on the Reduction of Flame Speed Figure 93.10 shows the effect of stretch rate on the flame speed for water mist mass fraction of Y0 ¼ 0.03. The flame speed SL,with was normalized by that without water mist SL,w/o. Of note is that the flame speed without water

910

T. Okawa et al.

mist increases with the stretch rate as shown in Fig. 93.4. The reduction of the flame speed monotonously decreases with the equivalence ratio, whereas the reduction of flame temperature is larger for both lean and rich mixtures than the stoichiometric one. Therefore, in terms of flame speed inhibition, the water mist is more effective for lean mixture than rich one. This inhibition effect is enhanced by the stretch rate. On the other hand, both of water mist and stretch rate can hardly reduce the flame speed for rich mixture (ϕ ¼ 1.2). However, this may not fall in estimation of effectiveness of water mist, because the lean flame should be in general encountered in the real-fire or explosion incidences.

93.5

Conclusions

The effect of water mist on flame temperature and laminar flame speed of propane/air mixtures was investigated both experimentally and numerically. Experiments were performed using a single jet-plate configuration, and the OPPDIF code in CHEMKIN package was used in the numerical simulation. For the case without water mist, experiments show that the flame speeds increase with the stretch rate toward the limit of extinguishment. This tendency is fairly reproducible by the numerical simulation with DLW kinetic mechanism. The flame temperature decreases with the water mist addition toward the limit of extinguishment within the range of ϕ from 0.8 to 1.2. The reduction of the flame temperature is more enhanced both for lean and rich mixtures than stoichiometric one. Stretch rate also decreases the flame temperature. In terms of flame speed inhibition, water mist is more effective for lean mixture than rich one, of which dependency is different from temperature.

References 1. Gann RG (2008) Guidance for advanced fire suppression in aircraft. Fire Technol 44:263–282 2. Babushok VI, Linteris GT, Burgess DR Jr, Baker PT (2014) Hydrocarbon flame inhibition by C3H2F3Br (2-BTP). Combust Flame 162:1104–1112 3. Osorio CH, Vissotski AJ, Petersen EL, Mannan MS (2013) Effect of CF3Br on C1-C3 ignition and laminar flame speed: numerical and experimental evaluation. Combust Flame 160:1044–1059 4. Zhang P, Zhou Y, Cao X, Cao X, Bi M (2014) Mitigation of methane/air explosion in a closed vessel by ultrafine water fog. Saf Sci 62:1–7 5. Xu H, Li Y, Zhu P, Wang X, Zhang H (2013) Experimental study on the mitigation via an ultra-fine water mist of methane/coal dust mixture explosions in the presence of obstacles. J Loss Prev Process Ind 26:815–820 6. Grant G, Brenton J, Drysdale D (2000) Fire suppression by water sprays. Prog Energy Combust Sci 26:79–130

7. Liu Z, Kim AK (2001) A review of water mist fire suppression technology: part II-application studies. J Fire Prot Eng 11:16–42 8. Mazas AN, Fiorina B, Lacoste DA, Schuller T (2011) Effects of water vapor addition on the laminar burning velocity of oxygenenriched methane flames. Combust Flame 158:2428–2440 9. Go¨ckeler K, Albin E, Kr€ uger O, Paschereit CO (2013) Burning velocities of hydrogen-methane-air mixtures at highly streamdiluted conditions. In: Proceedings of 4th international conference on jets, wakes and separated flows, p 1–6 10. Kuznetsov M, Redlinger R, Breitung W, Grune J, Friedrich A, Ichikawa N (2011) Laminar burning velocities of hydrogenoxygen-steam mixtures at elevated temperatures and pressures. Proc Combust Inst 33:895–903 11. Naito H, Uendo T, Saso Y, Kotani Y, Yoshida A (2011) Effect of fine water droplets on extinguishment of diffusion flame stabilized in the forward stagnation region of a porous cylinder. Proc Combust Inst 33:2563–2571 12. Yoshida A, Uendo T, Takasaki R, Naito H, Saso Y (2011) Water droplets behavior in extinguishing the methane-air counterflow diffusion flame. In: Fire Safety Science-Proceedings of 10th international symposium, p 569–582 13. Sakurai I, Suzuki J, Kotani Y, Naito H, Yoshida A (2013) Extinguishment of propane/air co-flowing diffusion flames by fine water droplets. Proc Combust Inst 34:2727–2734 14. Yoshida A, Udagawa T, Momomoto Y, Naito H, Saso Y (2013) Experimental study of suppressing effect of fine water droplets on propane/air premixed flames stabilized in the stagnation flowfield. Fire Saf J 58:84–91 15. Yoshida A, Takasaki R, Kashiwa K, Naito H, Saso Y (2013) Extinguishment of counterflow methane/air diffusion flame by fine water droplets. Combust Flame 160:1357–1363 16. Yoshida A, Okawa T, Ebina W, Naito H (2015) Experimental and numerical investigation of flame speed retardation by water mist. Combust Flame 162:1772–1777 17. Yoshida A, Kashiwa K, Hashizume S, Naito H (2015) Inhibition of counterflow methane/air diffusion flame by water mist with varying droplet diameters. Fire Saf J 71:217–225 18. Fuss SP, Chen EF, Wang WH, Kee RJ, Williams BA, Fleming JW (2002) Inhibition of premixed methane/air flames by water mist. Proc Combust Inst 29:361–368 19. Wu CK, Law CK (1984) On the determination of laminar flame speeds from stretched flames. Proc Combust Inst 20:1941–1949 20. Law CK, Zhu DL, Yu G (1986) Propagation and extinction of stretched premixed flames. Proc Combust Inst 21:1419–1426 21. Zhu DL, Egolfopoulos FN, Law CK (1988) Experimental and numerical determination of laminar flame speeds of methane/(Ar, N2, CO2)-air mixtures as function of stoichiometry, pressure, and flame temperature. Proc Combust Inst 22:1537–1545 22. Egolfopoulos FN, Zhu DL, Law CK (1990) Experimental and numerical determination of laminar flame speeds: mixtures of C2-hydrocarbons with oxygen and nitrogen. Proc Combust Inst 23:471–478 23. Vagelopoulos CM, Egolfopoulos FN, Law CK (1994) Further considerations on the determination of laminar flame speeds with the counterflow twin-flame technique. Proc Combust Inst 25:1341–1347 24. Vagelopoulos CM, Egolfopoulos FN (1998) Direct experimental determination of laminar flame speeds. Proc Combust Inst 27:513–519 25. Egolfopoulos FN, Hansen N, Ju Y, Kohse-Hoinghaus K, Law CK, Qi F (2014) Advances and challenges in laminar experiments and implications for combustion chemistry. Progr Energy Combust Sci 43:36–67 26. Niemann U, Seshadri K, Williams FA (2014) Accuracies of laminar counterflow flame experiments. Combust flame 162:1540–1549

93

Inhibition of Propane/Air Premixed Flame by Water Mist

27. Linteris GT, Truett L (1996) Inhibition of premixed methane-air flames by fluoromethanes. Combust Flame 105:15–27 28. Reinelt D, Linteris GT (1996) Experimental study of the inhibition of premixed and diffusion flames by iron. Proc Combust Inst 26:1421–1428 29. Noto T, Babushok V, Hamins A, Tsang W (1998) Inhibition effectiveness of halogenated compounds. Combust Flame 112:147–160 30. Donato N, Aul CJ, Petersen E, Zimmer C, Curran H, Bourque G (2010) Ignition and oxidation of 50/50 butane isomer blends. J Eng Gas Turb Power 132:051502, also see http://c3.nuigalway.ie/ butane.html 31. San Diego Mechanism Homepage, http://web.eng.ucsd.edu/mae/ groups/combustion/mechanism.html. Cited 1 Dec 2005 32. Davis SG, Law CK, Wang H (1999) Propene pyrolysis and oxidation kinetics in a flow reactor and laminar flame. Combust Flame 119:375–399, also see http://ignis.usc.edu/Mechanisms/C3/c3.html 33. Takahashi F, Katta VR (2009) Extinguishment of diffusion flames around a cylinder in a coaxial air stream with dilution or water mist. Proc Combust Inst 32:2615–2623 34. Yoshida A, Yukawa A (2013) Numerical simulation of the effect of water mist on the flame speed and structure of propane-air premixed flame. J Combust Soc Jpn (in Japanese) 55:403–410 35. Davis SG, Law CK (1998) Determination of and fuel structure effects on laminar flame speeds of C1 to C8 hydrocarbons. Combust Sci Tech 140:427–449

911 36. Law CK (1988) Dynamics of stretched flames. Proc Combust Inst 22:1381–1402 37. Sohrab SH, Ye ZY, Law CK (1984) An experimental investigation on flame interaction and the existence of negative flame speeds. Proc Combust Inst 20:1957–1965 38. Liu GE, Ye ZY, Sohrab SH (1986) On radiative cooling and temperature profiles of counterflow premixed flames. Combust Flame 64:193–201 39. Egolfopoulos FN, Zhang H, Zhang Z (1997) Wall effects on the propagation and extinction of steady, strained, laminar premixed flames. Combust Flame 109:237–252 40. Williams FA (1985) Combustion theory, 2nd edn. Benjamin/ Cummings, Menlo Park, pp 415–423 41. Skovorodko PA, Tereshchenko AG, Knyazkov DA, Paletsky AA, Korobeinichev OP (2012) Experimental and numerical study of thermocouple-induced perturbations of the methane flame structure. Combust Flame 159:1009–1015 42. Fristrom RM, Westenberg AA (1965) Flame structure. McGrawHill, New York, pp 170–174 43. Chen ZH, Lin TH, Sohrab SH (1988) Combustion of liquid fuel sprays in stagnation-point-flow. Combust Sci Tech 60:63–77 44. Bouvet N, Davidenko D, Chauveau C, Pillier L, Yoon Y (2014) On the simulation of laminar strained flames in stagnation flows: 1D and 2D approaches versus experiments. Combust Flame 161:438–452

Part XXIII Tunnel Fires

The Influence of Train on the Smoke Movement in Subway Tunnel with Longitudinal Ventilations

94

Shaogang Zhang, Xudong Cheng, Ruifang Zhang, Kaiyuan Li, Song Lu, Hui Yang, and Heping Zhang

Abstract

The smoke movement in the subway tunnel with train under different longitudinal ventilations was simulated using Fire Dynamics Simulator (FDS) program in this paper. The influence of subway train on the characteristics of smoke flow, including the smoke temperature underneath the tunnel ceiling and back-layering length, were investigated. The results showed that the upstream smoke gas temperature was much lower with train than without when the ventilation velocity was lower than the critical value and the downstream dimensionless smoke gas excess temperature decayed exponentially. The subway train could contribute to reduce the back-layering length at low ventilation velocities. Meanwhile, the FDS simulated smoke back-layering length agreed well with the calculated values of the proposed model. However, the difference between train lengths of 20 m and 40 m was insignificant. The mechanism of the train influence was discussed in depth using hydrodynamics. Keywords

Subway tunnel
Train length
FDS
Longitudinal ventilation
Smoke temperature
Backlayering length

Nomenclature A A1 A2 C0 Cs D* g H k L

An amplitude coefficient of equation (3) Cross-sectional area of subway tunnel (m2) Cross-sectional area of subway train (m2) Specific heat of capacity (J/(kg K)) Smagorinsky constant Dimensionless grid size The acceleration of gravity (m/s2) Tunnel height (m) The decay coefficient of ceiling temperature Back-layering length of smoke (m)

L* Prt Q Q* Sct T0 Tmax Tx V V*

Dimensionless back-layering length of smoke Turbulent Prandtl number Heat release rate (kW) Dimensionless heat release rate Turbulent Schmidt number Temperature of ambient air (K) Maximal smoke gas temperature (K) Smoke gas temperature x m away downstream of fire source (K) Ventilation velocity in tunnel (m/s) Dimensionless ventilation velocity

Greek Symbols S. Zhang
X. Cheng
R. Zhang
K. Li
S. Lu
H. Yang H. Zhang (*) State Key Laboratory of Fire Science, University of Science and Technology of China, No.96, Jinzhai Road, Hefei 230026, China e-mail: [email protected]

ρ0 φ

Density of ambient air (kg/m3) Blockage ratio

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_94

915

916

94.1

S. Zhang et al.

Introduction

To resolve the severe traffic jams, more and more subway transportation systems are constructed in modern cities. Fire is therefore a great threat to the occupants’ safety inside the subways. Due to the special building structure, the hot smoke gas induced by fires will flow along the ceiling and carry much hazardous combustion products to a long distance, causing catastrophic results, especially when the train on fire stops in the tunnel. A notable example is the subway fire accident in Baku, Azerbaijan, leading to 558 fatalities and 269 people injured [1]. Most of the deaths were in fact caused by the hot smoke gas. So far, many researches have been carried out on the smoke characteristics induced by tunnel fires, including full-scale tests and a significant amount of reduced scale experiments. The full-scale tests, such as Memorial Tunnel Fire Ventilation Test Program, Second Benelux Tunnel Fire Test Program, EUREKA, and Runehamar Tunnel Fire Test, have provided valuable information [2]. The reduced scale experiments also played an important role in the study of tunnel fire. Many key features of the tunnel fire were studied based on the model tunnel experiments [3–7], and many theories were developed using the experimental results [8– 15]. Besides, CFD as a fire safety engineering tool has been also widely used in the tunnel fire researches. Chow [16] once analyzed the smoke movement using the fire zone model “CFAST,” and Karaaslan [17] studied the effects of longitudinal ventilation on the temperature distribution in a tunnel using FLUENT. Tilley [18] performed a qualitative analysis on the effectiveness of different extraction strategies in a semi-transverse ventilated tunnel. Chen and Leong [19] conducted numerical simulations in a 400 m tunnel using Fire Dynamics Simulator (FDS). It can be concluded from the literature review that most of the previous studies on tunnel fire are focused on the road tunnels, which has different aspect ratio compared to the subway tunnels. Generally, the width of the road tunnel is

Fig. 94.1 The difference between (a) road tunnel and (b) subway tunnel in longitudinal ventilation

10 ~ 14 m and the height is 7 ~ 9 m, while the subway tunnel is usually 4.5 m wide and 5 m high for rectangular cross section and around 6 m in diameter for the circular cross section. So the aspect ratio of the road tunnel is approximately 0.8, which is lower than that of subway tunnel over 1.0. The subway tunnel is much narrower. According to the results of Ryou and Lee’s experiments, the aspect ratio affects the growth and development of smoke in tunnel fires [9]. When the fire occurs in a narrow tunnel, the air entrainment of smoke plume will be restrained by the sidewalls, and the heat feedback from the heated boundaries will increase. Ji et al. [20] have also found that the restriction effect of tunnel sidewalls would increase the maximum smoke temperature significantly when the distance between the fire and the sidewalls decreases to a certain threshold value. Hence, the fire plume characteristics in a narrow subway tunnel are quite different from the ones in the wide road tunnel [21]. On the other hand, the subway train is about 3 m wide and 3.8 m high, which is larger than the heavy goods vehicles (2.4  2.7 m) and the small cars (2  1.5 m) in the road tunnels. As a result, the blockage ratio in a narrow subway tunnel is higher than that in the wider road tunnel. Similar to the aspect ratio, the blockage ratio also has a great influence on the smoke behaviors in tunnel fires. More importantly, as a most commonly used method to control the fire smoke, the longitudinal ventilation in the subway tunnel is different from that in the road tunnel. In the road tunnel, the longitudinal ventilation is achieved by booster fans installed at the tunnel ceiling, which can cause uneven distribution of longitudinal ventilation along the height in practice. However, in the subway tunnel, evenly distributed longitudinal ventilation at the whole cross section will be generated using the station ventilation facilities. That implies that the subway train is “immersed” into the longitudinal ventilation, while the car or truck is “separated” from the ceiling ventilation in the road tunnel, as shown in Fig. 94.1. Meanwhile, the train length is much longer than the truck or car, and the influence

94

The Influence of Train on the Smoke Movement in Subway Tunnel with Longitudinal Ventilations

of vehicle length on the smoke movement has not been revealed yet. So, in the present study, the influence of train length on the smoke ceiling temperature distribution and backlayering length in subway tunnel with different longitudinal ventilation velocities are investigated using FDS 5.3.

94.2

Numerical Simulation

94.2.1 FDS Model The Fire Dynamics Simulator (FDS) is a powerful CFD model developed by the National Institute of Standards and Technology (NIST). In FDS, a form of the Navier-Stokes equations is solved numerically for low-speed, thermally driven flows with an emphasis on smoke and heat transport from fires. Based on the conservation laws of mass, momentum, species, and energy, FDS models the density, velocity, temperature, pressure, and species concentration of smoke. FDS has been widely used in the numerical simulations of fire and has achieved a great success. The simulated full-scale subway tunnel is 360 m long, 4.0 m wide, and 5.0 m high with a rectangular crosssectional area of 20 m2, as shown in Fig. 94.2. The materials of tunnel surface including walls, ceiling, and floor are specified as “concrete,” and their thermal properties are listed in Table 94.1. A simplified steel train model with 3 m wide and 3.8 m high is inside the tunnel. The length of train is varied to study the effect on the characteristics of fire and smoke. Every tunnel has two opening portals. The left portal of tunnel is set as an air “supply” vent to achieve the constant longitudinal ventilation velocity on the whole cross section at the inlet opening in each simulation case, and the right one is set as “open.” The smoke gas temperature along the tunnel ceiling under different longitudinal velocities is measured using the thermocouples function in FDS, and all the Fig. 94.2 The schematic diagram of subway tunnel with a train

917

thermocouples are set 5 cm below the tunnel ceiling. The intervals between two adjacent thermocouples in the region above the fire source are 0.5 m, while it is 1 m in other regions. It is assumed to be smooth for the tunnel wall surface, and the default 1D heat transfer process is applied to the walls. The mixture fraction combustion model is used in FDS, and the convective heat is 65 % of total heat release rate of fire. In all the simulation cases, the LES is adopted as the turbulence model, and the Smagorinsky constant Cs is the default value of 0.20 with the turbulent Prandtl number Prt of 0.5 and the turbulent Schmidt number Sct of 0.5, respectively.

94.2.2 Fire Scenario A large number of statistical data on subway and railway fires showed that about 48 % of the fires resulted from the electrical and mechanical failures of the train which are usually located at the electrical and brake systems at the bottom of the trains [1]. In order to mimic the electronic equipment fire in the locomotive, the fire source is set in the front of the train in this study. It is simplified as an obstruction with top surface of a “burner,” which is 2 m (L)  2 m (W)  1.5 m (H) and located at the origin of coordinates. The heat release rate per unit area (HRRPUA) is assigned for the fire source. According to the results of correlative fire experiments conducted in Hong Kong, the size of subway train fire in a tunnel is 5 ~ 10 MW. Therefore, we set the

Table 94.1 The thermal properties of material Model Ceiling Walls Floor Train

Material

Density kg/m3

Species heat kJ/(kg ∙ K)

Concrete

2280

1.04

Steel

7850

0.46

Conductivity w/(m ∙ K) 1.80 45.8

918

S. Zhang et al.

average value of 7.5 MW simultaneously at the start of the calculation as the current fire size in the simulations. Based on the relevant regulations, the subway train has to keep running at low speed until it reaches the adjacent station for passenger evacuation, if the fire is not very severe. However, investigations on the train fire accidents show that in about 45 % of the fires, the train failed to reach the next stations [1]. Therefore, this paper mainly focuses on the condition that the train stops in the tunnel connecting two adjacent subway stations. In order to study the influence of train length and longitudinal ventilation, in total, 12 cases are simulated in this research, and Table 94.2 lists the detailed information about them. In all the simulations, the ambient temperature is 20  C with the pressure of 101.325 kPa, and the simulation time is 900 s. Table 94.2 The detailed information about the simulated cases

0

7.5

1

7.5

2

7.5

3

Fig. 94.3 The temperature distribution in vertical direction simulated with four different grids

∗



D ¼

Q pffiffiffi ρ0 C0 T 0 g

2=5 ð94:1Þ

where Q is the heat release rate (kW) and ρ0, C0, and T0 are the density, specific heat, and temperature of ambient air, respectively. As the heat release rate is 7.5 MW, D* is calculated to be 2.155 m for this fire size; thus, 0.1D* is approximately 0.2155 m. Therefore, the grid size of 0.20 m can be used in the numerical simulation. In fact, numerical simulations are carried out using four grid sizes, which are 0.2 m, 0.15 m, 0.12 m, and 0.10 m, to evaluate the grid independence. Figure 94.3 presented the simulation results of temperature distribution in the vertical direction under different grid sizes. The measured location is 40 m away from the fire source. From Fig. 94.3, the two curves are rougher with bigger grid size of 0.20 m and 0.15 m, while they become relatively smoother when the grid sizes are reduced to

a

b 5.0

5.0 dd=0.20

4.5

dd=0.15

4.5 Height/m

7.5

Train length (m) 0 20 40 0 20 40 0 20 40 0 20 40

It is well known that the simulation results are directly affected by the grid size. Ma and Quintiere [22] used FDS to simulate axially symmetric flames and claimed that 0.05D* produced the best simulation results for the fire simulations, while the simulated velocity and temperature distribution of the fire plume would meet Baum and McCaffrey’s experimental curve when the grid size was taken as 0.1D* [23, 24]. According to these studies, the grid size in this study was determined by the following equation:

4.0 3.5 3.0

4.0 3.5 3.0

0

50

100 150 200 Temperature/⬚C

250

c

0

50

100 150 200 Temperature/⬚C

250

d 5.0

5.0 dd=0.12

4.5

dd=0.10

4.5 Height/m

Ventilation velocity (m/s)

Height/m

Fire size (MW)

Height/m

Case 1 2 3 4 5 6 7 8 9 10 11 12

94.2.3 Grid Resolution Analysis

4.0 3.5 3.0

4.0 3.5 3.0

0

50

100 150 200 Temperature/⬚C

250

0

50

100 150 200 Temperature/⬚C

250

94

The Influence of Train on the Smoke Movement in Subway Tunnel with Longitudinal Ventilations

0.12 m and 0.10 m. Most importantly, previous studies [25] have confirmed that there is a temperature gradient within the hot smoke layer beneath the ceiling. In other words, the upper surface temperature of the smoke layer is lower due to the heat transfer through the ceiling, and the lower surface temperature of the smoke layer is not high because of the air entrainment, and the maximum temperature occurs typically around 1 % of the distance from the fire source to the ceiling. As shown in Fig. 94.3, this temperature gradient cannot be simulated well using the grid of 0.20 m and 0.15 m; however, the 0.12 m and 0.10 m grid is workable. Therefore, in order to save the computation time, the grid of 0.12 m is selected to simulate the subway tunnel fire.

919

Figure 94.4 presented the smoke gas temperature distribution beneath the tunnel ceiling at the upstream and downstream of the fire source. As expected, the maximum

smoke gas temperature occurred around the fire source, and the temperature decreased significantly along the tunnel due to the heat transfer to the cold boundaries and the entrained fresh air as the smoke moved underneath the ceiling [26]. As shown in Fig. 94.4a, the smoke gas temperature was symmetrically distributed under the ceiling of the tunnel without longitudinal ventilation, and there was no noticeable difference in the temperature values measured at the same position in the tunnel with different train lengths. However, this symmetric distribution no longer existed due to the longitudinal ventilation. The smoke gas temperature decayed to the ambient value faster at the upstream than at the downstream of the fire source, especially with the train in the tunnel. The longitudinal ventilation effectively prevented the hot smoke spreading at the upstream, which was enhanced by the train as the temperature decreased to ambient value more easily with lower ventilation velocities when there was a train (as shown by (b) and (c) in Fig. 94.4). There was no noticeable difference in the temperature distribution between train lengths of 20 m and 40 m under the condition of same longitudinal ventilation velocity.

a

b

94.3

Results and Discussion

94.3.1 Smoke Temperature Under the Tunnel Ceiling

1000

L=0 L=20 L=40

600 400

600 Temperature/⬚C

Temperature/⬚C

800

V=0m/s

200

–240 –200 –160 –120 –80 –40 x/m

0

40

300 200

V=1m/s

100

–240 –200 –160 –120 –80 –40 x/m

L=0 L=20 L=40

240 V=2m/s

80 0

0

40

200

80 120

L=0 L=20 L=40

160 Temperature/⬚C

320 Temperature/⬚C

400

80 120

d

400

160

500

0

0

c

L=0 L=20 L=40

700

120 80

V=3m/s

40 0

–240 –200 –160 –120 –80 –40 x/m

0

40

80 120

Fig. 94.4 The smoke temperature distribution along the tunnel ceiling

–240 –200 –160 –120 –80 –40 x/m

0

40

80 120

920

S. Zhang et al.

94.3.2 Smoke Gas Excess Temperature

From Table 94.3, a uniform expression for the exponential fitting equations is given as

In order to normalize all the temperature distributions in the downstream of fire source under different conditions, a dimensionless smoke gas excess temperature was defined as

y ¼ y0 þ A*expðk*xÞ

ð94:3Þ

The downstream dimensionless temperature distributions along the tunnel ceiling for all simulations were shown in Fig. 94.5. In the previous studies, Hu [27] has found that the above dimensionless excess temperature fell into an exponential decay by investigating the ceiling temperature recorded in full-scale experiments, and Ingason [12] has claimed that the dimensionless smoke gas excess temperature could be correlated using an exponential decay model in theory; hence, the exponential fitting was applied. From Fig. 94.5, it could be seen that all the distributions were well fitted by an exponential line. And the detailed fitting results of each case were listed in Table 94.3. The proposed dimensionless correlation could be used to predict the temperature distribution along the downstream of the fire source.

where y is the dimensionless smoke excess temperature, y0 is the constant parameter, A is amplitude coefficient, and k is the decay coefficient. The constant parameter y0 denotes the ratio of excess temperature of smoke gas reaching the exit of the tunnel to the maximum excess temperature in the tunnel. And the decay coefficient k is related with both ventilation conditions and ceiling heat transfer process [12]. Analyzing the fitting results of each case in Table 94.3, it could be known that the constant parameter y0 and the decay coefficient k of fitting equations increased as the ventilation velocity increases, which indicated that the dimensionless excess temperature decayed faster under the condition of better ventilation. With the increase of ventilation velocity, the influence of the train became more obvious for the curves separated from each other. Besides, the difference in constant parameter y0 became larger between the cases with a train and the case without a train, which implied the train

a

b

0.35 L=0 L=20 L=40 Exp fitting of L=0 Exp fitting of L=20 Exp fitting of L=40

0.30 (Tx-T0)/(Tmax-T0)

ð94:2Þ

0.25 0.20 0.15 V=0m/s

0.10 0.05

20

40

60 X/m

80

100

0.5 0.4 0.3 0.2

120

V=1m/s 0

d

0.7 L=0 L=20 L=40 Exp fitting of L=0 Exp fitting of L=20 Exp fitting of L=40

0.6 0.5 0.4 0.3 V=2m/s

0.2 0

20

60 X/m

40

60

80

100

120

0.8 L=0 L=20 L=40 Exp fitting of L=0 Exp fitting of L=20 Exp fitting of L=40

0.7 0.6 0.5 0.4 V=3m/s

0.3 40

20

X/m

(Tx-T0)/(Tmax-T0)

(Tx-T0)/(Tmax-T0)

L=0 L=20 L=40 Exp fitting of L=0 Exp fitting of L=20 Exp fitting of L=40

0.1 0

c

0.7 0.6

(Tx-T0)/(Tmax-T0)

Tx  T0 ΔT ¼ T max  T 0

80

100

120

0

20

40

60 X/m

Fig. 94.5 The dimensionless smoke gas excess temperature distributions in the downstream of fire source

80

100

120

94

The Influence of Train on the Smoke Movement in Subway Tunnel with Longitudinal Ventilations

921

Table 94.3 The exponential fitting results of smoke gas excess temperature in the downstream of fire source No 1 2 3 4 5 6 7 8 9 10 11 12

Ventilation velocity (m/s) 0

1

2

3

Train length (m) 0 20 40 0 20 40 0 20 40 0 20 40

350

contributed to reduce the temperature of smoke gas reaching the exit of the tunnel. But for the train length of 20 m and 40 m, the difference was found to be insignificant.

R2 0.94 0.94 0.93 0.95 0.96 0.94 0.93 0.93 0.94 0.94 0.96 0.95

Exponential fitting equation y ¼ 0:045 þ 0:331*expð0:023xÞ y ¼ 0:044 þ 0:329*expð0:021xÞ y ¼ 0:045 þ 0:344*expð0:022xÞ y ¼ 0:137 þ 0:471*expð0:024xÞ y ¼ 0:103 þ 0:573*expð0:023xÞ y ¼ 0:106 þ 0:574*expð0:024xÞ y ¼ 0:221 þ 0:478*expð0:023xÞ y ¼ 0:197 þ 0:531*expð0:025xÞ y ¼ 0:201 þ 0:530*expð0:026xÞ y ¼ 0:364 þ 0:486*expð0:027xÞ y ¼ 0:310 þ 0:492*expð0:030xÞ y ¼ 0:297 þ 0:490*expð0:032xÞ

Back-layering length=33 m

300

94.3.3 Back-Layering Length Influence of Subway Train When the longitudinal ventilation velocity was relatively low, the hot smoke could spread into the upstream along the tunnel ceiling, which was defined as the back-layering. The furthest distance of hot smoke gas propagation in the upstream direction was the back-layering length. The backlayering affecting the clean evacuation environment at the upstream tunnel and threatening the evacuating people should be prevented by the longitudinal ventilation in tunnel fire. The temperature measurements are usually used to identify the back-layering length. Figure 94.6 presents the typical longitudinal temperature distribution with two adjacent thermocouples labeled as thermocouple A and B in the figure. The temperature of thermocouple B is at ambient value; meanwhile, that of thermocouple A has a significant temperature rise reading. Thus, the horizontal back-layering smoke front is identified as position of thermocouple A with its distance from the fire source taken as the back-layering length [28], as shown in Fig. 94.6. From Table 94.4, which summarized the smoke-layering length in the simulations, we can know that the backlayering length was reduced as the ventilation velocity increased and the train was very useful in terms of reducing the back-layering length of hot smoke gas. For example, with the ventilation velocity of 2 m/s in the tunnel, the back-layering length decreased to be around 2 ~ 3 m in cases with a train, while it was over 30 m without a train. However, the difference between two train lengths of 20 m and 40 m was unnoticeable.

Temperature/⬚C

250 200 150 100 Thermocouple A 50

Thermocouple B Back-layering length

0 –55 –50 –45 –40 –35 –30 –25 –20 –15 –10 –5

0

5

10 15

X/m

Fig. 94.6 Longitudinal temperature distribution and the indication of back-layering length Table 94.4 The back-layering length of smoke gas in each case

No. 1 2 3 4 5 6 7 8 9 10 11 12

Ventilation velocity (m/s) 0

1

2

3

Train length (m) 0 20 40 0 20 40 0 20 40 0 20 40

Back-layering length (m) Simulated Calculated – – – – – – 108 97.5 19 19.4 18 19.4 33 33.5 3 0 2 0 0 0 0 0 0 0

Based on the Froude modeling, Li and Ingason [28, 29] has derived the following dimensionless model to predict the back-layering length in the tunnel fire through dimensionless

922

S. Zhang et al.

analysis method, which was found to well comply with the experimental data:   ( 18:5 ln 0:81 Q*1=3 =V * , Q*  0:15 * L ¼ ð94:4Þ   18:5 ln 0:43=V * , Q* > 0:15 where Q* ¼

Qpffiffiffiffiffiffi

ρ0 c0 T 0

gH 5

, V * ¼ pVffiffiffiffiffi, L* ¼ HL . gH

In order to investigate the influence of blockage on the critical ventilation velocity and back-layering in the road tunnel fire, Hu [28] has studied the tunnel fire with blockage and revised the above model equation as follows: n (

o 18:5 ln 0:81 Q*1=3 = ð1  φÞV * , Q*  0:15 * L ¼ 

18:5 ln 0:43= ð1  φÞV * , Q* > 0:15 ð94:5Þ where φ ¼ AA21 ¼ 33:8 45 ¼ 0:57 in this paper is the blockage ratio of train in the subway tunnel. The calculated backlayering lengths based on the above models were also listed in Table 94.4. Comparing the FDS simulated and model calculated values, it was concluded that the FDS simulation results agreed well with the model-predicted back-layering lengths under different ventilations. The reason was that these models were proposed based on the results of reduced scale experiments where the ventilation method was the same with that used in this simulation.

94.3.4 Influence of Subway Train The existence of a subway train has a great influence on the smoke temperature distribution under the tunnel ceiling and

back-layering length, as discussed before. These influences were caused by the change of local ventilation flow field around the fire source. Lee and Tsai [15] studied the tunnel fire behavior with vehicular blockage and found that the change of fire behavior was primarily owing to the change of local ventilation velocity and upstream blockage would prevent the ventilation airflow from reaching the fire region directly. Hu [28] and Li [29] have also identified the similar behaviors in their studies. When there was no train in the tunnel, the longitudinal ventilation flow interacted with the fire plume directly. The ventilation velocity and volume flow rate did not change before reaching the fire plume. However, the ventilation flow had to go across the train vehicle due to the blockage effect when there was a train in the upstream of fire. Assuming ventilation flow in the tunnel was a steady incompressible flow, the velocity would increase with the reduction of cross-sectional area [15], which is


m ¼ ρAV 1 ¼ ρð1  φÞAV 2 ¼ const V2 ¼

1 V1 1φ

ð94:6Þ ð94:7Þ

According to the above equations, the velocity of ventilation flow after going across the train blockage would became 1=ð1  φÞ times higher than before, which was illustrated in Fig. 94.7. Figure 94.7 presented that the velocity distribution contour of ventilation flow field around the fire source along the tunnel center line when the ventilation velocity is 2 m/s. From Fig. 94.7, the longitudinal ventilation kept stable as 2 m/s before reaching the fire plume when there was no train in the upstream. Similarly, the ventilation velocity also did not change before reaching the train front in

Fig. 94.7 The comparison of ventilation flow field around fire source along the tunnel center line: (a), (b) and (c) without fire; (A), (B) and (C) with fire

94

The Influence of Train on the Smoke Movement in Subway Tunnel with Longitudinal Ventilations

the cases with a train. However, the velocity of flow between tunnel ceiling and train upper surface increased to 4 m/s due to great reduction of sectional area, and it would further increase to 5 ~ 6 m/s before reaching the fire plume. The direct interaction between ventilation flow with higher velocity and fire plume would cause a large amount of turbulence with high intensity, leading to the intensive mixing of incoming fresh air and smoke gas, as shown in Fig. 94.7b, c. As a result, the tunnel in the downstream of the fire source would be full of smoke gases, and the backlayering length was much reduced. On the other hand, the length of space between tunnel ceiling and the train’s upper surface was long enough for the ventilation flow velocity to increase and then become stable. Thus, the increase of train length from 20 to 40 m only prolonged the time of flow traveling the space between tunnel ceiling and train, which could not change the outlet velocity of ventilation flow. The interaction between ventilation flow with high velocity and fire plume was little changed by the train length. That was the reason why the difference was found to be insignificant between train lengths of 20 m and 40 m.

94.4

Conclusion

In order to investigate the influence of train on the smoke characteristics in the subway tunnel with different longitudinal ventilation velocities, a series of FDS simulations were carried out in this study. Two train lengths of 20 m and 40 m were examined with the longitudinal ventilation velocity varying from 0 to 3 m/s. The results showed that: 1. The train contributed to the decrease of smoke gas temperature under the tunnel ceiling at the upstream with low longitudinal ventilation velocity. 2. The dimensionless smoke gas excess temperature in the downstream decayed exponentially, and it decayed faster with the increase of ventilation velocity. The subway train contributed to reduce the temperature of smoke gas reaching the exit of the tunnel. 3. Compared to the cases without train, the back-layering length of smoke was reduced significantly with train when the longitudinal ventilation velocity was lower than the critical value. The simulated back-layering length agreed well with the calculated values using proposed model in the literatures. 4. These influences were mainly caused by the change of the longitudinal ventilation flow field around the fire source. With a train in the upstream tunnel, the velocity of ventilation flow increased to be about 1=ð1  φÞ times higher than before, and it became stable before interacting with

923

fire plume, resulting in little difference between the train lengths of 20 m and 40 m.

Acknowledgment This work was mainly supported by National Natural Science Foundation of China (No. 51323010) and Fundamental Research Funds for the Central Universities (No. WK2320000033). We deeply appreciate it.

References 1. Zhou D, Yan X, Zheng J eds (2014) Study on fire characteristics of subway train running with fire. In: CICTP 2014@ Safe, Smart, and Sustainable Multimodal Transportation Systems; ASCE, pp 19–23 2. Lin P, Lo S, Li T (2014) Numerical study on the impact of gradient on semi-transverse smoke control system in tunnel. Tunn Undergr Space Technol 44:68–79 3. Ji J, Gao Z, Fan C, Zhong W, Sun J (2012) A study of the effect of plug-holing and boundary layer separation on natural ventilation with vertical shaft in urban road tunnel fires. Int J Heat Mass Transf 55(21):6032–6041 4. Fan C, Ji J, Sun J (2014) Influence of longitudinal fire location on smoke characteristics under the tunnel ceiling. Fire Mater 39:132–147 5. Harish R, Venkatasubbaiah K (2014) Effects of buoyancy induced roof ventilation systems for smoke removal in tunnel fires. Tunn Undergr Space Technol 42:195–205 6. Ura F, Kawabata N, Tanaka F (2014) Characteristics of smoke extraction by natural ventilation during a fire in a shallow urban road tunnel with roof openings. Fire Saf J 67:96–106 7. Yi L, Niu J, Xu Z, Wu D (2013) Experimental studies on smoke movement in a model tunnel with longitudinal ventilation. Tunn Undergr Space Technol 35:135–141 8. Ko GH, Kim SR, Ryou HS (2010) An experimental study on the effect of slope on the critical velocity in tunnel fires. J Fire Sci 28 (1):27–47 9. Lee SR, Ryou HS (2005) An experimental study of the effect of the aspect ratio on the critical velocity in longitudinal ventilation tunnel fires. J Fire Sci 23(2):119–138 10. Hu L, Tang W, Chen L, Yi L (2013) A non-dimensional global correlation of maximum gas temperature beneath ceiling with different blockage–fire distance in a longitudinal ventilated tunnel. Appl Therm Eng 56(1):77–82 11. Hu L, Chen L, Wu L, Li Y, Zhang J, Meng N (2013) An experimental investigation and correlation on buoyant gas temperature below ceiling in a slopping tunnel fire. Appl Therm Eng 51 (1):246–254 12. Ingason H, Li YZ (2010) Model scale tunnel fire tests with longitudinal ventilation. Fire Saf J 45(6-8):371–384 13. Yi L, Xu Q, Xu Z, Wu D (2014) An experimental study on critical velocity in sloping tunnel with longitudinal ventilation under fire. Tunn Undergr Space Technol 43:198–203 14. Chen L, Hu L, Tang W, Yi L (2013) Studies on buoyancy driven two-directional smoke flow layering length with combination of point extraction and longitudinal ventilation in tunnel fires. Fire Saf J 59:94–101 15. Lee Y-P, Tsai K-C (2012) Effect of vehicular blockage on critical ventilation velocity and tunnel fire behavior in longitudinally ventilated tunnels. Fire Saf J 53:35–42

924 16. Chow W (1996) Simulation of tunnel fires using a zone model. Tunn Undergr Space Technol 11:221–236 17. Karaaslan S, Hepkaya E, Yucel N (2013) Cfd simulation of longitudinal ventilation systems in a scaled short tunnel. J Therm Sci Technol 33:63–77 18. Tilley N, Rauwoens P, Merci B (2011) Verification of the accuracy of CFD simulations in small-scale tunnel and atrium fire configurations. Fire Saf J 46(4):186–193 19. Chen F, Leong J (2011) Smoke flow phenomena and turbulence characteristics of tunnel fires. Appl Math Model 35(9):4554–4566 20. Ji J, Fan CG, Zhong W, Shen XB, Sun JH (2012) Experimental investigation on influence of different transverse fire locations on maximum smoke temperature under the tunnel ceiling. Int J Heat Mass Transf 55(17-18):4817–4826 21. Guo X, Zhang Q, Simone E, Astore G, Xu S, Grasso P (2012) The critical condition of longitudinal emergency tunnel ventilation – comparison of theoretical prediction with experimental data. Tunn Undergr Space Technol 32:78–86 22. Ma T, Quintiere J (2003) Numerical simulation of axi-symmetric fire plumes: accuracy and limitations. Fire Saf J 38(5):467–492 23. Baum H, McCaffrey B (eds) (1989) Fire induced flow field–theory and experiment. Fire safety science–proceedings of the second

S. Zhang et al. international symposium. Hemisphere Publishing, Newport, pp 451–468 24. Yang P, Tan X, Xin W (2011) Experimental study and numerical simulation for a storehouse fire accident. Build Environ 46 (7):1445–159 25. Karlsson B, Quintiere J (2002) Enclosure fire dynamics. CRC press, p 145 26. Lee SR, Ryou HS (2006) A numerical study on smoke movement in longitudinal ventilation tunnel fires for different aspect ratio. Build Environ 41(6):719–725 27. Hu L, Huo R, Wang H, Li Y, Yang R (2007) Experimental studies on fire-induced buoyant smoke temperature distribution along tunnel ceiling. Build Environ 42(11):3905–3915 28. Tang W, Hu L, Chen L (2014) Effect of blockage-fire distance on buoyancy driven back-layering length and critical velocity in a tunnel: an experimental investigation and global correlations. Appl Therm Eng 60(1):7–14 29. Li YZ, Lei B, Ingason H (2010) Study of critical velocity and backlayering length in longitudinally ventilated tunnel fires. Fire Saf J 45(6):361–370

The Effect of Aspect Ratios on Critical Velocity in Tunnel Fires

95

Changcheng Liu, Song Lu, Ruifang Zhang, Hui Yang, Xudong Cheng, and Heping Zhang

Abstract

The influence of aspect ratio, As (height/width for a tunnel of rectangular cross section), on the critical velocity in tunnel fires was studied by computational fluid dynamics (CFD) simulation. According to previous researches, the influence of aspect ratio on critical velocity presents different tendencies. Someone believes the critical velocity increases with As; others think that the critical velocity decreases when As is larger than unity. According to previous experimental results and CFD results, in this paper, it is inappropriate to use aspect ratio as the only factor which affects the critical velocity. Both As and the tunnel width should be considered. By introducing the characteristic fire diameter, the difference of relationship between critical velocity and aspect ratio is unified, and a new critical velocity model is developed. The critical velocity varies with the one-fifth power of the heat release rate for under-medium fires. Keywords

Tunnel fire
Aspect ratio
Critical velocity
Smoke control

Nomenclature H ΔT 

V A As Cp D D* Fr g H hc L

Hydraulic diameter of the tunnel (m) Temperature difference from ambient (K) Volumetric flow rate (m3/s) Tunnel cross-sectional area (m2) Tunnel aspect ratio Specific heat (J/kg K) Diameter of fire source (m) Characteristic fire diameter (m) Froude number Gravitational force (m/s2) Tunnel height (m) Heat transfer coefficient (W/m2 K) Characteristic length in Eq. 95.5 (m)

l Pr Q Re T Uc V Vp W

Length (m) Prandtl number Heat release rate (kW) Reynolds number Smoke temperature (K) Critical ventilation velocity (m/s) Ventilation velocity (m/s) Velocity of fire plume (m/s) Tunnel width (m)

Greek Symbol ρ0

Ambient air density (kg/m3)

Subscript C. Liu
S. Lu
R. Zhang
H. Yang
X. Cheng
H. Zhang (*) University of Science and Technology of China, Jinzhai 96, Hefei 230026, China e-mail: [email protected]

F M

Values used in full-scaled models Values used in scaled models

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_95

925

926

95.1

C. Liu et al.

Introduction

Toxic smoke with high temperature produced in fire is the most important cause of casualties [1]. Tunnel can be considered as a long narrow confined space, and when fire occurs, there are no vents but only two exits. This makes the combustion inadequate and smoke production increased, which leads a serious threat to personnel safety. Therefore, the control of smoke is a very important endeavor during a tunnel fire. Especially, it is significant to stop the backflow of hyperthermal toxic gases to guarantee the evacuation of people from the tunnel. However, the ventilation may strengthen the fire and lead severe results [2, 3]. The critical ventilation velocity Uc which defined as the minimum airflow velocity to prevent smoke back layering has been widely used for smoke control in tunnel fires. It already becomes a crucial parameter in tunnel ventilation system design. Researchers have studied critical velocity for many years including experiments and numerical modeling. Critical velocity varies with many factors such as fire heat release rate, tunnel physical dimension, gradient, etc. Thomas first presented using longitudinal ventilation to control smoke in tunnel fires and gave a critical velocity model as follows [4, 5]:


gHQ Uc ¼ k ρ0 Cp AT

1=3 ð95:1Þ

where k is a constant approach to unity and must be confirmed by experiments under different fire scenarios. According to Eq.1, the critical velocity (Uc) has a one-third power relation with the heat release rate (Q). Many researchers have studied how tunnel width and height influence the critical velocity as they are two of the fundamental Fig. 95.1 Side view and front view of the two series tunnel models. The top shows the arrangement of the fire source and the velocity-sensor tree. The bottom shows the tunnel cross sections and aspect ratios. The unit of length is centimeter

parameters of a tunnel. Wu et al. [6] described the influence of the tunnel’s geometrical shape with introducing the hydraulic diameter H. Lee et al. [7] pointed that the tunnel’s aspect ratio was an important factor. They modified a model by introducing tunnel aspect ratio As (height H to width W ), and the new model showed that the critical velocity always increased with the aspect ratios. Nevertheless, in the study by O. Vauquelin and Wu [8], influence of tunnel width on smoke control was studied. It was found that the critical velocity increased with tunnel aspect ratio less than unity and decreased with aspect ratio larger than unity. For this reason, an approach has been carried out in order to investigate the influence of the tunnel aspect ratios on the critical velocity. The results can provide a reference during ventilation system design. The present work tries to make a numerical analysis to find the relationship between tunnel aspect ratio and critical velocity of various fire intensities along the tunnels. Two past experimental studies were employed to establish reduced-scale numerical models to examine the relationship.

95.2

Numerical Modeling

Fire Dynamics Simulator (FDS, version 6.0) was employed to conduct the present numerical study. The numerical models here simulate the reduced-scale experiments conducted by Lee and O. Vauquelin. The hydraulic diameters of tunnels in experiments of O. Vauquelin were different, while all the tunnels had same hydraulic diameter in experiments of Lee [7, 8]. In order to exclude the effect of hydraulic diameter, two different series models with invariant hydraulic diameter were built to simulate the two groups’ experiments, respectively, as shown in Fig. 95.1. The positions of the velocity sensors in the sensor tree were

95

The Effect of Aspect Ratios on Critical Velocity in Tunnel Fires

927

Table 95.1 Fire sizes and HRRs of the fires at the quasi-steady state [Lee] Pan size, cm 8 12 14 18

Q_ M ðkWÞ 2.47 5.63 8.27 12.30

Q_ F ðMWÞ 4.40 10.10 14.80 22.00

same as those of the vertical thermocouples used in the experimental study by Lee [7]. The heat release rate (HRR) of the fire source at quasisteady state varies from 2.47 to 12.3 kW by changing the fire sizes from 0.08  0.08 m to 0.18  0.18 m, as shown in Table 95.1. The corresponding HRR values for the prototype by Froude scaling are also shown behind reduced fire HRRs. According to Lee [7], the tunnel used for experiments was reduced to a full-scale tunnel by the length of 1/20. The fullscale tunnel was 208 m long. The scaling relationships for velocity, volumetric flow rate, and heat release rate are [7] rffiffiffiffiffi lM VM ¼ VF ð95:2Þ lF
5=2 lM V_ M ¼ V_ F lF

ð95:3Þ


5=2 lM QM ¼ QF lF

ð95:4Þ

The fires are designed by the method that assigns the heat release rate per unit area (HRRPUA) on the surface of the fire source provided by FDS [9]. For all simulation cases, the left tunnel end is assigned as an air supply condition, and the right end is an open condition. The ambient and initial ceiling temperature is 20  C. The left inlets of the tunnels are velocity inlets which are used to simulate the longitudinal ventilation, and the right side of the tunnels are free outlets. The simulation time is set as 200 s because all the fires achieved their quasi-steady state before this time [7, 10]. Models of series 1 are same with Lee’s experiments, and the models of series 2 are similar with O. Vauquelin’s experiments [7, 8]. The grid in this paper is the same as models of Li [11]. For independence exam of mesh in this paper, please check the previous work [11]. The size of grids near the fire sources is 0.01  0.01  0.01 m, and the size of the other parts is 0.02  0.02  0.02 m. The total cells are between 183,375 and 468,000. Under lengthways ventilation temperature of the forward of the back layering under the ceiling of the tunnel will have a significant decrease and can be used for judging if the velocity reaches the critical

Fig. 95.2 Temperature distribution centerline slice near the fire source at the time of 200 s

condition [12] as shown in Fig. 95.2. As mentioned above, the fire will achieve their quasi-steady state within 200 s; the smoke movement shown by the smokeview would reach to a steady state in which the position of the forward of the back layering would not change. If the ventilating velocity reached the critical velocity, the forward of the back layering and the fire source would have a same vertical position as shown in Fig. 95.2. If the ventilating velocity was too large, the location of the forward would be some distance downstream from the fire source. Oppositely, the location of the forward would be in the left of the fire source with an insufficient ventilating velocity. We increased or decreased the velocity by 0.01 m/s each time. The critical velocity would be determined after several times of increase or decrease. The accuracy of the velocity is 0.01 m/s. At the inlet the homogeneous velocity distribution was employed. The thermal exchange modeling on the roof was accounted for by the default convective heat transfer model in FDS. By default in LES simulation in FDS, the heat transfer coefficient to the ceiling surface is expressed as [9]
 k hc ¼ max C1 jΔT j1=3 , C2 Re4=5 Pr 1=3 ð95:5Þ L where C1 accounts for the natural convection, which by default is 1.52 for a horizontal surface and 1.31 for a vertical surface; C2 accounts for the forced convection, which is 0.037; and L is taken to be 1 m for most calculations in FDS [9]. The characteristic length for conductive and radiative heat transfers in this word was also accounted for by the default models in FDS. Many models are available as SGS model in FDS ver.6; default model was chosen in the present work.

928

95.3

C. Liu et al.

Results and Discussion

95.3.1 CFD Results Figure 95.3 gives the simulation results of two series of tunnel fires. As the tunnels of series 1 are consistent with the tunnels of experiments carried out by Lee [7], the critical velocities of series 1 are contrasted with experimental results to verify the reliability of the CFD results. The maximum relative error between the experiment and the simulation is 4.52 %, and the minimum value is 0.15 %, which indicates a good agreement between the numerical and experimental results. From Fig. 95.3, we can see that the critical velocity increases with As in tunnels of series 1 and first increases with As and then decreases when As is larger than unity in tunnels of series 2. With the hydraulic diameters kept invariant, the results have the same trends with the experimental results of O. Vauquelin and Lee [7, 8]. As tunnels of each

series have the same hydraulic diameter, it reflects that hydraulic diameter is not the key reason that makes the difference. As critical velocities of tunnels with an aspect ratio of 2.0 in series 2 were smaller than that of tunnels with an aspect ratio of 1.0 which had a different trend with series 1, tunnels with an aspect ratio of 2.0 of the two series were selected for further analysis. Figure 95.4 shows the velocity contours and pressure contours of tunnels on a cross section 0.3 m downstream from the fire seat in order to make a deeper insight about the relationship between aspect ratio and critical velocity. As shown in Fig. 95.4, a tunnel with an aspect ratio of 2.0 is the narrowest tunnel. Although the fire intensity is the largest, the sidewalls have little restriction of air entrainment near the fire zone in the tunnel of series 1. Edges of the plume still have some distance to the sidewall. Compared with the tunnel of series 2, the sidewalls have significant effect on the fire development with the plume occupying the whole width of the tunnel which is similar with the situation reported by O. Vauquelin [8]. According to Wu [6], the critical velocity becomes independent of the heat release rate at higher heat release rate as the flame reaches the ceiling. In the higher part of the tunnel, the velocity of the back layering under the ceiling is close to zero and smaller than that of the plume and air in other parts of the tunnel cross section. It is noticed that thickness of the back layering under the ceiling of the tunnel of series 1 is thicker than that of the tunnel of series 2. It means that the flame of tunnels of series 2 is more close to the ceiling which may have an effect on the critical velocity. This may be the reason of the decrease of critical velocity of tunnels of series 2 in Fig. 95.3. Through analysis of Fig. 95.4, we can find the tunnel width had an important effect on the critical velocity. Thus, it is the relative size of the fire source and the tunnel width which becomes the key parameter which have effect on the critical velocity in this work. In view of that, the two groups did not consider too much about the relationship between the fire source and the tunnel width; further analysis of the experimental results of the two groups may concentrate on this aspect. So the following section will analyze the experimental results of Lee and O. Vauquelin in the respect of the relative size of the fire source and the tunnel width and try to find some new information about the critical velocity.

95.3.2 Data Analysis According to Lee [7], the critical velocity model is as follows: Fig. 95.3 Critical velocity against aspect ratio of model series 1 and model series 2 from top to bottom

0

Q ¼

Q qffiffiffiffiffiffiffiffiffiffiffiffiffi 5 ρa CP T a As gH

ð95:6Þ

95

The Effect of Aspect Ratios on Critical Velocity in Tunnel Fires

929

Fig. 95.4 The CFD-predicted pressure and velocity distribution at 0.3 m downstream of the fire source of 12.30 kW fire in the tunnel of series 1 and 2.47 kW fire in the tunnel of series 2 from top to bottom with an aspect ratio of 2.0

0

Vc ¼

Uc pffiffiffiffiffiffi gH

A0:2 s

ð95:7Þ

The hydraulic tunnel height was used as the characteristic length in the analysis. The aspect ratio was also included. In this model, the tunnel width and tunnel height had same weight. The previous section indicated that tunnel width and tunnel height had different extent of impact on the critical velocity. Based on experimental results of Lee and O. Vauquelin [7, 8], we develop a new critical velocity model to unify the results. The dimensionless analysis method was employed to analyze the two groups’ data. Considering the two groups used different shapes of fire pools (circle and rectangle), the characteristic fire diameter from FDS user’s guide is introduced to represent the fire source [9]. It is defined as follows: D ¼ *

Q_ pffiffiffi ρa Cp T a g

!2=5 ð95:8Þ

According to the previous section, tunnel width has more effect on the critical velocity than tunnel height, so the tunnel width was employed as the characteristic length in the analysis. The dimensionless heat release rate is defined as the relative size of the fire source and the tunnel width.  5=2 0 Q ¼ D* =W ¼

Q pffiffiffi ρa CP T a gW 5=2

ð95:9Þ

Equation 95.9 has the same form with the model of Wu and Bakar [6]. The dimensionless critical velocity is as follows, which has the same form with the model of Oka [13] with the aspect ratio included for the first time. Uc 0 V c ¼ pffiffiffiffiffiffiffiffiffiffiffiffi gAs W

ð95:10Þ

Based on these two definitions, experimental results of the two groups can be repainted as Fig. 95.5. From Fig. 95.5, it can be seen that experimental data scale of O. Vauquelin is bigger than Lee’s. What we must pay attention to is that both

930

C. Liu et al.

0

00

Fig. 95.5 Dimensionless critical velocity Vc against 2/5 power of dimensionless heat release rate Q0 of experimental results by Lee and O. Vauquelin [7, 8]

Fig. 95.6 Dimensionless critical velocity Vc against 2/5 power of dimensionless heat release rate Q00 of experimental results by Lee and O. Vauquelin [7, 8]

the two groups’ data have a rising trend. However, this dimensionless analysis cannot correlate the experimental results into a simple form as the data is scattered. The present work proposed to introduce the effect of the aspect ratio of the tunnel on the critical velocity. As the first introduction of the aspect ratio still cannot correlate the experimental results into a simple form, the two group’s experimental data was analyzed with different power of the aspect ratio again by data analysis software (Origin 9.0) and finally unified to a simple form. The new dimensionless critical ventilation velocity and new dimensionless heat release rate are defined as

work. Similar situation has been discussed by Hwang [14]. In his work, the critical ventilation velocity was plotted against the fire heat generation rate and was roughly proportional to the 1/5 power of Q. By referring to the study of Quintiere [15], the reference velocity for a fire plume is

00

Q ¼

00

Q pffiffiffi ρa Cp T a gW 5=2

Vc ¼

Uc 1=10 pffiffiffiffiffiffiffi gW As

ð95:11Þ

Vp ¼

00

The plot of Vc against Q00 using the tunnel width and the aspect ratio is shown in Fig. 95.6. A simple one-dimensional correlation to predict the critical velocity for tunnels can be proposed as Eq. 95.13.   00 0:5*Q0015, Q00 < 0:27 Vc ¼ ð95:13Þ 0:385jQ00  0:27 From Figs. 95.5 and 95.6, we can see that the influence of tunnel width, tunnel height, and tunnel aspect ratio on critical velocity is in different degrees. So they must be considered, respectively. Dimensionless critical velocity increases with the one-third power of the dimensionless heat release rate by previous studies [7, 8], while it increases with the one-fifth power of the dimensionless heat release rate in this




Q pffiffiffi ρa gCp T a

1=5 ð95:14Þ

This may because the velocity of fire plume occupies the major section of the smoke velocity. Combining Eqs. 95.14 and 95.12, the relationship between the reference velocity for a fire plume and the dimensionless critical ventilation velocity can be expressed as follows: Vp 1 V c ¼ 1=10 pffiffiffiffiffiffiffi ¼ 1=10 gW As As 00

ð95:12Þ

pffiffiffi g

Q pffiffiffi ρa Cp T a gW 5=2

!1=5 ð95:15Þ

From Eq. 95.15, we can see that the content in the bracket has the same form with Eq. 95.11. The aspect ratios of the tunnels in this paper vary from 0.25 to 2.0 which lead to the values of As1/10 varying from 0.87 to 1.07. For a rough analysis, it can be argued that the dimensionless critical ventilation velocity varies with 1/5 power of Q. In this study, the critical velocity does not vary with the heat release rate when the dimensionless heat release rate is larger than 0.27. The corresponding heat release rates and tunnel widths are shown in Table 95.2. Combining Fig. 95.6 and Table 95.2, we can see that the three inverted triangle points are 3.0, 7.5, and 15 kW fires with As ¼ 1.83 (W ¼ 0.136) of O. Vauquelin and the other triangle point is 15 kW fire with As ¼ 1.0 (W ¼ 0.25) of O. Vauquelin. Fire HRRs of all these four points are larger than the HRRs

95

The Effect of Aspect Ratios on Critical Velocity in Tunnel Fires

Table 95.2 Tunnel widths and the corresponding heat release rates of Q00 ¼ 0.27 Tunnel width, m 0.136 0.25 0.30 0.33 0.40

HRR, kW 2.19 10.05 15.85 20.12 32.55

931

between critical ventilation velocity and tunnel cross-section aspect ratio. According to simulated results, hydraulic diameter is not the key reason which makes the different trends of critical velocity vary with cross-section aspect ratio. The relative size of tunnel width and characteristic fire diameter has the most important effect on the critical velocity. By introducing the characteristic fire diameter and considering the influence of tunnel width and tunnel aspect ratio, the difference of the two groups’ opinion is unified. A new critical ventilation velocity model is presented. Acknowledgment This work was mainly supported by National Natural Science Foundation of China (No. 51323010) and Fundamental Research Funds for the Central Universities (No. WK2320000033). The authors deeply appreciate the support. The authors thank Dr. Liming Li of Shenyang Fire Research Institute of China.

References

00

Fig. 95.7 Dimensionless critical velocity Vc against 2/5 power of dimensionless heat release rate Q00 of experimental results by Li [12]

in Table 95.2. What we must pay attention to is that the HRRs in Table 95.2 are appropriate for tunnels with a certain range of aspect ratios (0.136–2.0 in this paper). Figure 95.7 is the comparison between Eq. 95.14 and experimental results of Li [12]. It can be seen that all the experimental data of Li are found to be little above the line proposed in this paper. This indicates that the predicted critical velocity from Eq. 95.14 is lower than that of the experimental data of Li especially at a dimensionless heat release rate above 0.14. In other words, all the experimental data of Li are larger than data of Lee and O. Vauquelin. It may be caused by the different materials of tunnels. The tunnels of Lee used gypsum board as the material of tunnel section near the fire source, and O. Vauquelin used PMMA as the tunnel material. Both these two materials had small thermal conductivity, while tunnels of Li which are made of stainless steel had a quite larger thermal conductivity.

95.4

Conclusion

Critical velocity is an important factor of smoke control in a tunnel fire, and tunnel cross-section aspect ratio has a crucial impact on it. In this paper, the impact of tunnel aspect ratio on critical velocity was studied using numerical modeling. Some researchers have different opinions on the relationship

1. Babrauskas V, Gann RG, Levin BC, Paabo M, Harris RH, Peacock RD, Yusa S (1998) A methodology for obtaining and using toxic potency data for fire hazard analysis. Fire Saf J 31(4):345–358 2. Chow WK (1998) On smoke control for tunnels by longitudinal ventilation. Tunn Undergr Space Technol 13(3):271–275 3. Ingason H, Li YZ (2010) Model scale tunnel fire tests with longitudinal ventilation. Fire Saf J 45(6–8):371–384 4. Thomas PH (1958) The movement of buoyant fluid against a stream and the venting of underground fires, Fire Research Note No. 351. Fire Research Station, Borehamwood 5. Thomas PH (1968) The movement of smoke in horizontal passages against an air flow, Fire Research Station Note No. 723. Fire Research Station, Borehamwood 6. Wu Y, Bakar MZA (2000) Control of smoke flow in tunnel fires using longitudinal ventilation systems – a study of the critical velocity. Fire Saf J 35:363–390 7. Lee SR, Ryou HS (2005) An experimental study of the effect of the aspect ratio on the critical velocity in longitudinal ventilation tunnel fires. J Fire Sci 23:119–138 8. Vauquelin O, Wu Y (2006) Influence of tunnel width on longitudinal smoke control. Fire Saf J 41:420–426 9. McGrattan K, Hostikka S, Floyd J (2013) Fire dynamics simulator (version 6) user’s guide. National Institute of Standards and Technology, Washington, DC 10. Lee SR, Ryou HS (2006) A numerical study on smoke movement in longitudinal ventilation tunnel fires for different aspect ratio. Build Environ 41:719–725 11. Li L, Xie Q, Cheng X, Zhang H (2011) Temperature distribution of fire-induced flow along tunnels under natural ventilation. J Fire Sci 30(2):122–137 12. Li YZ, Lei B, Ingason H (2010) Study of critical velocity and back layering length in longitudinally ventilated tunnel fires. Fire Saf J 45(6-8):361–370 13. Oka Y, Atkinson (1996) Control of smoke flow in tunnel fires. Fire Saf J 25(4):305–322 14. Hwang CC, Edwards JC (2005) The critical ventilation velocity in tunnel fires – a computer simulation. Fire Saf J 40(3):213–244 15. Quintiere JG (1989) Scaling applications in fire research. Fire Saf J 15(1):3–29

Influence of Different Longitudinal Wind on Natural Ventilation Efficiency with Vertical Shaft Under Different Fires in Tunnel

96

Haiyong Cong, Xishi Wang, Pei Zhu, Zhigang Wang, Tonghui Jiang, and Qiong Tan

Abstract

The influence of longitudinal wind on the nature of ventilation with a vertical shaft in tunnel fires was investigated numerically by large eddy simulations (LES). The smoke flow characteristics under the coupled effects of different longitudinal wind and different heat release rates were analyzed. The dimensionless value φ has been introduced to define the exhaust efficiency. The results show that under the same longitudinal wind velocity, the bigger the fire can lead to a better exhaust efficiency. In a similar manner, an optimal wind speed will achieve best performance in shaft exhaust efficiency and toxic gas emission. The traditional simple wind speed calculation is not suitable for all of the cases; it is necessary to adjust the ventilation speed according to different fire situations. Keywords

Tunnel fire
Shaft
Natural ventilation
Stack effect
Exhaust efficiency

Nomenclature

96.1

Cp D* g Q_ : T1 ωshaft ωo

In recent years, large numbers of subway tunnels have been under construction in cities as strategic infrastructure to solve the traffic problems. Owing to the special structure of underground tunnels, when a fire occurs, smoke and toxic gases will accumulate in the tunnel and not be easily discharged [1, 2]. According to statistics, 85 % of deaths in fires were the consequence of toxic gas inhalation [3]. There are two typical ventilation systems adopted in tunnels. One is natural ventilation and the other is mechanical ventilation. The installation of mechanical ventilation systems needs large installation space and is not always able to meet the requirements, in particular when there is poor management or maintenance. In contrast, natural ventilation with a shaft is easy to build and more stable in daily operations, because it nearly does not require maintenance. Many researchers have conducted studies on tunnel fires. The maximum temperature beneath the ceiling, the critical velocity, and the heat release rates have been widely considered [4–10]. Nevertheless, research on the optimization of ventilation efficiency of a shaft under the coupled effects of

Specific heat capacity Characteristic fire diameter Gravity acceleration The total heat release rate of the fire Temperature of ambient air CO outlet mass fraction CO ceiling mass fraction without wind

Greek Symbols φ δx ρ1

Stoichiometric coefficient The mesh size Air density

H. Cong
X. Wang (*)
P. Zhu
Z. Wang
T. Jiang
Q. Tan State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei 230026, China e-mail: [email protected]

Introduction

# Springer Science+Business Media Singapore 2017 K. Harada et al. (eds.), Fire Science and Technology 2015, DOI 10.1007/978-981-10-0376-9_96

933

934

H. Cong et al.

different heat release rates and longitudinal flows are relatively few. Several researchers have studied the fire-induced smoke flow in tunnels under natural ventilation with a shaft. Yoon et al. [11] studied the pressure of nature ventilation in the shaft and tunnel and found that the natural ventilation pressure induced by the shaft had a remarkable impact on the efficiency of the ventilation. Bi et al. [12] investigated the effect of natural ventilation within shafts by using numerical simulations. Wang et al. [13] carried out a series of experiments in a full-scale tunnel with roof opening. Ji et al. [14] studied the stack effect in a tunnel and obtained the critical shaft height that avoided plug-holing. Zhong et al. [15] studied the smoke flow characteristics of the tunnel under different longitudinal wind velocities. However, none of the above studies perform an analysis of the influence of the coupling effect of different fire release rates and longitudinal wind. In this paper, a series of numerical studies were conducted to investigate the couple effects of fire and longitudinal wind on smoke ventilation efficiency of shaft in a tunnel.

96.2

CFD Numerical Modeling

96.2.1 On Fire Dynamics Simulator As we know, CFD simulation has become an important tool to study fire and fluid problems. Fire Dynamics Simulator (FDS) was developed by the US National Institute of Standards and Technology (NIST), and now being used widely in fire research. FDS numerically solves a form of the Navier–Stokes equations for thermally driven flow. It includes both direct numerical simulation (DNS) model and large eddy simulation (LES) model. The LES model is widely used in the study of fire-induced smoke flow behavior and selected in this study. The model has been subjected to numerous validations, calibrations, and studies on the temperature and velocity fields in fires. The critical ventilation velocity which was just sufficient to prevent smoke backflow was examined numerically by Hwang et al. [16] and experimentally by Roh et al. [6]. Both

Fig. 96.1 Arrangement of measure points inside the tunnel

studies showed an almost similar result when calculating the backflow distance. The highest temperature under the ceiling of tunnel was simulated with FDS by Hu et al. [17], and the results match up with the experimental values. Jing [18] conducted simulations on flow field with a jet fan, and the simulation results accord with the experiment carried by Mutama et al. [19]. These proved that the FDS can effectively simulate the gas movement and the heat transfer process in a fire. FDS (version 5.5.3) was used in this study.

96.2.2 Fire Scenario Analysis As is shown in Fig. 96.1, the primary region of the tunnel is 120 m long, 9 m wide, and 6 m high. The fire source is located in the tunnel axle wire longitudinally and 50 m away from the left side. The shaft is 8 m high with a cross section of 1.5  1.5 m, 15 m away from the left edge of the tunnel shaft. The T-square fires with a peak heat release rate of 3 MW, 5 MW, and 8 MW were considered, which represent the typical scenarios of car fires inside tunnels, especially inside the road tunnels, where 3 MW represents car fire, 5 MW represents passenger car fire, and 8 MW represents van fire [20]. The air inlet is set at the left entrance of the tunnel model, and the two outlets are at the right side and the shaft, respectively. Both ends of the tunnel are set open. In the full-scale experiments conducted by Yan et al. [21], the ambient longitudinal wind velocity was measured between 0.8 m/s and 1.5 m/s. Therefore, in this simulation, the longitudinal wind velocity range was enlarged and set to be 0–3 m/s in current physical model. Based on the former study, the exterior environment is a significant parameter in CFD simulations. A model without an exterior environment will result in different data. Therefore, the ambient temperature is set as 20 C, the ambient pressure is set as 101,325 Pa, and the simulation time is 300 s. A series of thermocouples are arranged 0.15 m below the tunnel ceiling to study the gas distribution along the tunnel central line. Gas velocity, temperature, and pressure were tested inside the shaft. Three rows of test points were arranged inside the shaft as is shown in Fig. 96.2.

96

Influence of Different Longitudinal Wind on Natural Ventilation Efficiency with. . .

935

Fig. 96.2 Arrangement of measurement points inside the shaft

96.2.3 Mesh Sensitivity Study For FDS simulation, one of the most important parameters is mesh size, which may affect the accuracy of the numerical simulation results. According to FDS User’s Guide, an FDS input file using a relative coarse mesh should be built, and then gradually refine the mesh until it satisfies your requirement, i.e., mesh sensitivity study [22]. For simulations involving buoyant fire plume, the D*/δx has been widely used to estimate how well the flow field is resolved, where δx is the mesh size and D* is a characteristic fire diameter which is calculated by Eq. 96.1: ∗

D ¼

Q_ pffiffiffi ρ1 c P T 1 g

!2=5 ð96:1Þ

In this formula, Q_ is heat release rate, ρ1 is ambient gas density, cP is specific heat capacity, T 1 is ambient temperature, and g is gravity acceleration. In the guidebook, it was noted that the value of D*/δx should be in the range of 4–16. So the mesh size for 3 MW fire is 0.09–0.37 m, for 5 MW fire is 0.114–0.456 m, and for 8 MW fire is 0.137–0.551 m. Generally, the smaller the mesh is, the higher the accuracy would be. But if the mesh is too small, there will be no significant improvement of the measurement results but is more time consuming. In this study, five different mesh sizes between 0.1 and 0.3 m were selected to compare mesh sensitivity by the vertical temperature distribution in the tunnel. The vertical temperature with different mesh sizes is shown in Fig. 96.3. With the mesh density increase, the temperature curve tends to be uniform. In this paper, a smaller and more precise mesh system with grid size of 0.15 m is adopted.

96.3

Fig. 96.3 Vertical temperature distribution in the tunnel

Results and Discuss

96.3.1 Temperature Distribution A shaft with 8 m height and 1.5  1.5 m square cross section was tested. The study was carried out under different heat release rates and longitudinal wind velocities to study the exhaust capacity under different conditions.

Fig. 96.4 Temperatures in the tunnel shaft (HRR ¼ 3 MW)

The ventilation speed was set from 0 to 3 m/s, and the fire sizes were 3 MW, 5 MW, and 8 MW. Figure 96.4 shows the temperature curves in the shaft, and Fig. 96.5 shows the temperature distribution in the shaft and the tunnel. These results were obtained under the heat release rate of 3 MW. As is shown in Fig. 96.4, temperatures in the shaft almost remain same from the bottom to the top of the shaft with the effects of longitudinal wind. It indicates that the flow temperature inside the shaft is steady, which is differing from the windless condition. Figure 96.5 was taken at 260 s when the flow in the tunnel achieves steady state. Figure 96.5a was obtained for a windless condition; the flow is stable in most part of the tunnel except the shaft area. Ceiling gases obtained a great deal of heat from the fire. When the hot gases enter the shaft, a

936

H. Cong et al.

Fig. 96.5 Temperature distribution inside the tunnel and the shaft with fire heat release rate of 3 MW (a represents the case of windless nature ventilation; b–f represent the condition under longitudinal wind with 1 m/s, 1.5 m/s, 2 m/s, 2.5 m/s, 3 m/s)

special effect called stack effect would emerge in this special structure and result in the increase of the vertical inertial force.

If the vertical inertial force is too big and the horizontal inertial force is not strong enough for gas to escape, large amounts of fresh air beneath the smoke layer will be sucked

96

Influence of Different Longitudinal Wind on Natural Ventilation Efficiency with. . .

into the shaft, this phenomenon is called plug-holing. During this process, the entrained cold air will cool the hot gas which will in turn restrain the stack effect. As shown in Fig. 96.5a, there is a sunken area; the smoke temperature in this area is about 25–30  C, lower than any other regions in the shaft; and this is a typical phenomenon controlled by plug-holing. Figure 96.5b, c was taken under the conditions of wind velocity of 1 m/s and 1.5 m/s, respectively. During the increase of wind speed, as is shown in Fig. 96.5b, c, a thicker and relatively stable smoke layer forms, the formation of plug-holing will be restrained under the combination of the increased horizontal inertial force and smoke layer, and less cold air can be directly sucked. In this condition, the smoke temperature inside the shaft will increase, which in turn will strengthen the stack effect. Figure 96.5d, f shows the temperature distribution in the shaft with wind velocity from 2 to 3 m/s. The temperature inside the shaft decrease significantly with the increase of longitudinal wind velocity. As shown in Figure 96.5d, f, the temperature decrease sharply with an interval of about 10  C owning to the large horizontal inertia force induced by the high longitudinal wind, which further enhanced the convection heat transfer of smoke layer and cold air beneath. Vast amounts of cold air were supplied by a tunnel jet fan with a given velocity; heat transfer convection was strong, leading to a lower average temperature of smoke layer, which would in turn weaken the stack effect, and to a certain degree plugholing would disappear. So there must be a proper longitudinal wind velocity – a moderate wind velocity that can restrain the formation of plug-holing and induce a relatively strong stack effect.

96.3.2 Smoke Stratification From Fig. 96.5, ceiling gas temperature in front of the shaft decrease with the increase of ventilation velocity. In other studies, explanations given for this question are air entrainment induced by tunnel ventilation. This conclusion is arbitrary. This paper has checked the strength of the fire-induced stratification at the place between fire and shaft with Newman’s model [23, 24], which were placed at 5 m and 10 m away from fire on the right side central axis. Table 96.1 gives the Froude number calculated using Newman’s model under heat release rate of 3 MW with different ventilation velocities. Uavg ffi, where the velocity In Newman’s model, Fr ¼ qffiffiffiffiffiffiffiffi ΔT gH cf T avg

Uavg is the “hot” velocity across the cross section on the downstream side of the fire, ΔTcf is the temperature

937

Table 96.1 Froude number calculated for different test conditions Position X¼0 X ¼ 5

Mode V¼0 0.03 0.02

V¼1 0.13 0.18

V ¼ 1.5 0.19 0.212

V¼2 0.25 0.28

V ¼ 2.5 0.21 0.34

V¼3 0.35 1.04

difference between the ceiling temperature (0.88 H) and the floor temperature (0.12 H), and Tavg is the average temperature across the cross section. Newman proposes three smoke stratification region defined by the Froude number and temperature quotients: Region I (Fr