| Author: | Zhang, Xiaoning |
| Title: | Smart tunnel fire forecast and safety management driven by artificial intelligence of things |
| Advisors: | Huang, Xinyan (BEEE) Fu, Xiao (BEEE) Li, Heng (BRE) Wu, Xiqiang (BEEE) |
| Degree: | Ph.D. |
| Year: | 2024 |
| Subject: | Tunnels -- Fires and fire prevention Tunnels -- Safety measures Fire extinction -- Data processing Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Building Environment and Energy Engineering |
| Pages: | xi, 144 pages : color illustrations |
| Language: | English |
| Abstract: | Tunnel fire is one of the most severe global fire hazards and causes a significant number of economic losses and casualties every year. Over the last 50 years, numerous full-scale and reduced-scale tunnel fire tests, as well as numerical simulations have been conducted to quantify the critical fire events and key parameters to guide the fire safety design of the tunnel. During the firefighting operation, the firefighters can only judge the fire scenarios by their experience and perception. Insufficient and inaccurate information may lead to injuries and fatalities. To ensure the safety of firefighters and the fire rescue efficiency, it is necessary to build a systematical and powerful tunnel firefighting system that can used to guide the firefighting, rescue, and evacuation. This thesis is presented in a manuscript-style format: the first chapter gives an introduction of the research background and motivation, and an independent paper that has been published or submitted to a journal publication is contained in the subsequent chapter. The final chapter summarizes the overall outcomes and potential research directions for future work. Chapter 1 introduces the research background and motivation. Tunnel fires could result in serious consequences duo to its confined structures, which make it difficult to efficiently exhaust smoke and heat. During tunnel fire accidents, accurate information inside the tunnel can facilitate decision-making, firefighting operations, and rescue activities. However, the understanding of tunnel fires and techniques for tunnel firefighting are still insufficient. It is important to build a data-driven real-time fire forecast and management system to ensure the safety of firefighters and trapped people by promoting decision-making and safety management. Chapter 2 aims to establish a database that contains all existing experimental data of tunnel fire, which is critical to building the data-driven fire monitoring or forecast model, and the application of artificial intelligence in tunnel fire was also explored. This tunnel-fire database summarizes seven key parameters of flame, ventilation, and smoke in that is open access. The test conditions, experimental phenomena, and data of each literature work were organized and categorized in a standard format that could be conveniently accessed and continuously updated. Based on this database, machine learning is applied to predict the critical ventilation velocity of a tunnel fire as a demonstration. The review of the current database not only reveals more valuable information and hidden problems in the conventional collection of test data but also provides new directions in future tunnel fire research. The established database and methodology help promote the application of artificial intelligence and smart firefighting in tunnel fire safety. Chapter 3 explores the real-time prediction of fire information, temperature distribution, and critical events in real-time based on artificial intelligence algorithms. The numerical model was first validated by the full-scale tunnel fire test, and then a numerical database of 300 transient tunnel-fire scenarios is established under various initial fire locations, fire sizes, fire growth and spread rates, and ventilation conditions. The proposed dual-agent deep-learning model combining the Long Short-term Memory (LSTM) model and Transpose Convolution Neural Network (TCNN) is trained with the database. With the input data of on-site temperature sensors, the dual-agent model can forecast transient fire scenarios with changing location and size 30 s in advance. This study demonstrates the feasibility of the AI model in identifying and forecasting the rapid-changing fire scenarios inside a tunnel in smart firefighting practices. Chapter 4 proposes a framework based on computer vision to recognize vehicle types, track, and predict vehicle locations, and estimate fuel load distribution, fire risk, and evacuation risk inside a tunnel. All road vehicles are classified into 10 categories based on their usage and size, and the fuel load and fire HRR of different types of vehicles are obtained from the literature. Video cameras installed at the road tunnel inlet and outlet are used to monitor the traffic flow, and a deep learning model (i.e., YOLO v7) trained from public vehicle datasets is used to identify the vehicle types and estimate the average speed at the entrance. The average speed of the vehicles is assumed to be constant when passing through the tunnel, and then vehicle locations, fuel load distribution, fire risk, and evacuation risk were predicted in advance. It can also be used to monitor the traffic flows inside the tunnel by comparing the number of vehicles entering and exiting the tunnel. The proposed method was then verified by real-world traffic scenarios. This work investigated the feasibility of using entrance and exit traffic cameras to forecast tunnel fuel load distribution and assess tunnel fire risk. Chapter 5 proposed a digital twin framework for real-time tunnel fire safety management, driven by dynamic sensor data and Artificial Intelligence of Things (AIoT) technologies. A deep learning model trained by the Transformer network and simulation dataset is used to predict real-time fire location and size. Then, the AI model is integrated into a 3D digital twin platform developed by the game engine Unity 3D. The performance of the proposed digital twin framework is demonstrated using numerical experiments and full-scale tunnel fire tests. Results show that the established AI model achieved promising accuracy in predicting fire location and power for both numerical and experimental data. The digital twin platform can also visualize the 3D fire scene that supports evacuation, firefighting, and emergency rescue. This research demonstrates the feasibility of using a 3D environment and digital twin in real-time fire safety management. Chapter 6 summarizes the overall outcomes of the AIoT-based tunnel firefighting system in its current stage. The tunnel fire forecast and tunnel fire safety management problems are systematically explored. According to the present findings, the challenges the researcher needs to overcome will be discussed. Then, new technologies and methods will be used to overcome these challenges and achieve more practical questions. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/12970