A physical model-free ant colony optimization network algorithm and full scale experimental investigation on ceiling temperature distribution in the utility tunnel fire

Abstract

To advance understanding of the ceiling temperature characters in tunnel fires, a physical model-free ant colony optimization network algorithm is developed. Compared to the traditional physical model-based methods, the algorithm is not limited to the specific fire conditions and the structure of tunnels. The main advantage and contribution of the algorithm is that a novel ant colony optimization (ACO) network is constructed and firstly used to predict the ceiling temperature distribution in the tunnel fire as well as the maximal ceiling temperature based on only some sensors data. In order to verify the effectiveness of the algorithm, full scale burning tests were investigated in the largest fire experiment platform of the utility tunnel at the Tianjin Fire Research Institute, China. In addition, the developed ACO network algorithm has excellent performance by contrast with the commonly used back propagation (BP) neural network algorithm. By compared with the experimental results and the results obtained from the BP neural network algorithm, the ability and the effectiveness of the algorithm were supported. The algorithm can be used to predict the ceiling temperature in the tunnel fires for rapid and efficient fire disaster evaluation.