現今的火災模擬通常被用做評估或預測火災發生的情況，火災模擬讓我們得以火災預測特定環境中火災的發展情況以及各項數據。但為了進行與現實情況接近的模擬，涉及了許多複雜的計算過程，這需要花費非常大量的計算資源。計算量大帶來的大量耗時，也因此難以在分秒必爭的火災發生時用來進行模擬火災的發展情況。
火災發生時規劃逃生路徑需要對火災的現況與未來的發展有所了解，才能夠規劃出安全的逃生路徑，避免逃生時發生未預期的危害。因此若有快速的火災模擬可以在火災初期完成，將可以使逃生規劃更加完善與安全。
本文提出了兩種即時火災危險區域預測的方法。第一種為網格屬性預測法，透過記錄並學習傳統火災模擬方法的模擬結果，使用簡單的計算方法在數秒內輸出火災發生時未來危險區域的可能擴散情形；第二種為圖像序列預測法，將學習時間序列的模擬資料，透過火災發生時的即時危險區域資料推估後續危險區域擴散的情形。兩種方案將透過實驗了解其性能及實際應用上的可行性。
此項研究的輸出可做為火災發生時迅速進行決策的依據，例如進行火場逃生路徑規畫時可增加即時火災危險區域預測的結果作為輸入，得以規劃全程安全且正確的逃生路徑。

Nowadays, fire simulation is usually used to evaluate or predict the occurrence of fire. Fire simulation allows us to predict the development of fire in a specific environment and various data. However, in order to simulate as close to the real situation as possible, many complicated calculation processes are involved, which requires a very large amount of computing resources. The large amount of calculations brings a lot of time and it is therefore difficult to simulate the development of a fire when every second counts.
Planning an escape route when a fire occurs requires an understanding of the current situation and future development of the fire dangerous area in order to plan a safe escape route and avoid unexpected dangers during escape. Therefore, if there is a fast fire prediction that can be completed in the early stage of the fire, it will make the escape planning more complete and safer.
This paper proposes two methods for real-time fire dangerous area prediction. The output of this research can be used as a basis for rapid decision-making when a fire occurs. For example, when planning a fire escape route, the current and future hazardous area distribution can be considered to find a correct and safe escape route.

[1] 中華民國內政部消防署，「106年全國火災統計分析」，2018。
[2] 林金宏，「活著離開3」，雙葉書廊，台灣，2018。
[3] Bjorn Karlsson, James G. Quintiere, “Enclosure Fire Dynamics,” 2000.
[4] Forney, Glenn P, “Smoke View(Version 5)-A Tool For Visualizing Fire Dynamice Simulation Data, Volume I ： User’s Guide.” No. Special Publication (NIST SP)-1017-1,2017.
[5] P. Smardz, V. Novozhilov, “Validation of Fire Dynamics Simulator (FDS) for forced and natural convection flows,” University of Ulster, pp. 16-20, 2006.
[6] 黃育祥（2005）。應用火災工學與火災模擬軟體FDS於火場之重建。中央警察大學消防科學研究所碩士論文，桃園縣。 取自https://hdl.handle.net/11296/7b7wxc
[7] Jen-Hao Chi, “Reconstruction of an Inn Fire Scene Using the Fire Dynamics Simulator (FDS) Program,” Journal of Forensic Sciences, vol. 58, 2012, pp. S227-S234.
[8] J. Wahlqvist and P. van Hees, “Evaluating methods for preventing smoke spread through ventilation systems using fire dynamics simulator,” Fire and Materials, vol. 41, no. 6, 2016, pp. 625-645.
[9] T. Liang, J. Liu, and P. Tao, “The Implementation of the Numerical Simulation by Utilization of Pyrosim on the Rectification of the Hazard of Fires,” Proceedings of the 11th International Conference on Computer Modeling and Simulation - ICCMS 2019, 2019, pp. 31-35.
[10] Z. Chaczko and F. Ahmad, “Wireless Sensor Network Based System for Fire Endangered Areas,” Third International Conference on Information Technology and Applications (ICITA'05), Sydney, NSW, 2005, pp. 203-207.
[11] K. Sha, W. Shi and O. Watkins, “Using Wireless Sensor Networks for Fire Rescue Applications： Requirements and Challenges,” 2006 IEEE International Conference on Electro/Information Technology, East Lansing, MI, 2006, pp. 239-244.
[12] J. Lloret, M. Garcia, D. Bri, and S. Sendra, “A wireless sensor network deployment for rural and forest fire detection and verification,” Sensors (Basel, Switzerland), vol. 9, no. 11, Jan. 2009, pp. 8722-8747.
[13] A. Filippoupolitis and E. Gelenbe, “A Distributed Decision Support System for Building Evacuation,” 2009 2nd Conference on Human System Interactions, Catania, 2009, pp. 323-330.
[14] G. Gorbil, A. Filippoupolitis and E. Gelenbe, “Intelligent Navigation Systems for Building Evacuation,” Computer and Information Sciences II, 2011, pp. 339-345.
[15] A. Filippoupolitis and E. Gelenbe, “An Emergency Response System for Intelligent Buildings,” Sustainability in Energy and Buildings, vol. 12, 2012, pp. 265-274.
[16] J. Liu, R. Rojas-Cessa and Z. Dong, “Sensing, calculating, and disseminating evacuating routes during an indoor fire using a sensor and diffusion network,” 2016 IEEE 13th International Conference on Networking, Sensing, and Control (ICNSC), Mexico City, 2016, pp. 1-6.
[17] 蔡依芸（2020）。基於數位孿生架構之建築消防系統火災模擬。國立成功大學製造資訊與系統研究所碩士論文，台南市。 取自https://hdl.handle.net/11296/6hx7nx
[18] M. Birky, B. Halpin, Y. Caplan, R. Fisher, J. McAllister and A. Dixon, “Fire fatality study,” Fire and Materials, vol. 3, no. 4, 1979, pp. 211-217.
[19] 國家地震工程研究中心，「建築的構成」，檢自http：//streethouse.ncree.narl.org.tw/03.html(Apr.21,2020)。
[20] Lianfa Li, Ying Fang, Jun Wu, Jinfeng Wang, “Autoencoder Based Residual Deep Networks for Robust Regression Prediction and Spatiotemporal Estimation,” arXiv：1812.11262, 2018.
[21] Zenghao Chai, Chun Yuan, Zhihui Lin, Yunpeng Bai, “CMS-LSTM： Context-Embedding and Multi-Scale Spatiotemporal-Expression LSTM for Video Prediction,” arXiv：2102.03586, 2021.
[22] ThunderHead Eng, “ImportCAD-4,” ThunderHead Eng, 2019.
[23] Myung Bae Kim, Yong Shik Han, Myoung O. Yoon, “Laser-assisted visualization and measurement of corridor smoke spread,” Fire Safety Journal Volume 31, Issue 3,1998, pp. 239-251.
[24] 陳弘毅，「火災學」，大碩教育，2019。
[25] 葉冠廷（2019）。用於數位孿生建築消防系統之逃生演算法。國立成功大學製造資訊與系統研究所碩士論文，台南市。 取自https://hdl.handle.net/11296/8ujkjx
[26] S. Ahmed, M. Liwicki, M. Weber and A. Dengel, “Automatic Room Detection and Room Labeling from Architectural Floor Plans,” 2012 10th IAPR International Workshop on Document Analysis Systems, 2012, pp. 339-343.
[27] Jiachen Zhao, Fang Deng, Yeyun Cai, Jie Chen, “Long short-term memory - Fully connected (LSTM-FC) neural network for PM2.5 concentration prediction,” Chemosphere, Volume 220, pp. 486-492, 2019.
[28] Xingjian Shi, Zhourong Chen, Hao Wang, Dit-Yan Yeung, Wai-kin Wong, Wang-chun Woo, “Convolutional LSTM Network： A Machine Learning Approach for Precipitation Nowcasting,” arXiv：1506.04214, 2015.
[29] Li Yang, Xu Huahu, Bian Minjie, Xiao Junsheng, "Attention Based CNN-ConvLSTM for Pedestrian Attribute Recognition" Sensors, Volume 20, no. 3, 2020.
[30] L. Zhang, G. Zhu, L. Mei, P. Shen, S. A. A. Shah, and M. Bennamoun, “Attention in convolutional LSTM for gesture recognition,” in Proc. NIPS, 2018, pp. 1953-1962.
[31] John Burge, Matthew Bonanni, Matthias Ihme, Lily Hu, “Convolutional LSTM Neural Networks for Modeling Wildland Fire Dynamics,” arXiv：2012.06679, 2020.
[32] V. Le, T. Bui and S. Cha, “Spatiotemporal Deep Learning Model for Citywide Air Pollution Interpolation and Prediction,” 2020 IEEE International Conference on Big Data and Smart Computing (BigComp), 2020, pp. 55-62.
[33] Seongchan Kim, Seungkyun Hong, Minsu Joh, Sa-kwang Song, “DeepRain： ConvLSTM Network for Precipitation Prediction using Multichannel Radar Data,” arXiv：1711.02316, 2017.
[34] D. Wang, Y. Yang and S. Ning, “DeepSTCL： A Deep Spatio-temporal ConvLSTM for Travel Demand Prediction,” 2018 International Joint Conference on Neural Networks (IJCNN), 2018, pp. 1-8.
[35] Y. Zhou, H. Dong and A. El Saddik, “Deep Learning in Next-Frame Prediction： A Benchmark Review,” in IEEE Access, vol. 8, pp. 69273-69283, 2020.
[36] Casper Kaae Sønderby, Lasse Espeholt, Jonathan Heek, Mostafa Dehghani, Avital Oliver, Tim Salimans, Shreya Agrawal, Jason Hickey, Nal Kalchbrenner, “MetNet： A Neural Weather Model for Precipitation Forecasting,” arXiv：2003.12140, 2020.
[37] Yunbo Wang, Zhifeng Gao, Mingsheng Long, Jianmin Wang, Philip S. Yu, “PredRNN++： Towards A Resolution of the Deep-in-Time Dilemma in Spatiotemporal Predictive Learning,” arXiv：1804.06300, 2018.
[38] Bernard Lange, Masha Itkina, Mykel J. Kochenderfer, “Attention Augmented ConvLSTM for Environment Prediction,” arXiv：2010.09662, 2020.
[39] Zhihui Lin, Maomao Li, Zhuobin Zheng, et al. “Self-attention convlstm for spatiotemporal prediction,” AAAI 2020, pp. 11531-11538, 2020.