An intelligent tunnel firefighting system and small-scale demonstration

Abstract

Disastrous fire event in the confined tunnel is a fatal hazard, threatening the lives of trapped people and firefighters. Considering the rapid development of fire and the complex environment of tunnels, an accurate and timely fire identification system is in urgent need for guiding the evacuation, rescue, and firefighting actions. This study proposes an intelligent system and digital twin composed of four main components to collect, manage, process and visualize the tunnel fire information. As demonstrated in a laboratory-scale tunnel model, the AI model is trained with a large numerical database to successfully identify the fire size and location. The whole system is assessed in terms of accuracy, timeliness and robustness. The AI model attained an overall accuracy of 98% in predicting the tunnel fire scenarios. The total time delay is around 1 s from the on-site measurement of temperature to the final display of the tunnel fire scenario on a remote user interface. Moreover, the system is robust enough to predict fire, even if part of the temperature sensors is failed or destroyed by fire. The proposed intelligent system will be a valuable step for smart firefighting from the concept to practice.