Smart prediction for tunnel fire state evolution based on an improved fire simulation curve through particle swarm optimization algorithm

Tunnel fires have received significant attention because they can cause large economic losses every year. To estimate the fire development and realize a prompt response to the fire disaster are thus crucial to tackle the above issue. Here, a smart prediction method for fire state evolution is developed based on an improved fire simulation curve through the particle swarm optimization algorithm. The improved fire simulation curve is established by supplementing the stable stage during the tunnel fire based on a fire curve proposed in 2002. Experimental results from a prototypical tunnel fire test indicate that the proposed method can effectively portray the behavior of a typical tunnel fire development. Moreover, it is accurate over a wide range of conditions, including different heat release rates, different measurement noises, different sampling intervals, and different prior durations. Particularly, the fluctuation range of estimation results is becoming larger with the increasing noise level. The influence of the sampling interval is mainly reflected in the calculation time. Prior duration is the most important factor of the prediction accuracy, the longer prior duration corresponds to the more precise results. 200s and 300s can be regarded as the critical duration by considering the estimation error of the whole-time domain and the preliminary stage, respectively. The proposed method has a broad application prospect, it can avoid complicated temperature spreading mechanisms, the additional workload for prior information, and the limitation of the specific fire scene.