PSO–BPNN Fusion Algorithm for Tunnel Fire Ceiling Temperature Prediction under Longitudinal Ventilation

Abstract

The ceiling temperature in tunnel fires is crucial to determining safety parameters for both occupants and structures. Numerous studies have neglected the effect of longitudinal ventilation velocity in tunnels. This study introduces a novel hybrid methodology that combines model- and data-driven approaches to effectively predict temperature distribution along ceilings under longitudinal ventilation. This study introduces a ceiling temperature prediction model based on theoretical analysis, optimized using a fusion algorithm of particle swarm optimization and backpropagation neural network (PSO-BPNN). The feasibility of the methodology is verified through full-scale experiments and numerical simulations. The prediction outcomes are contrasted with those from the conventional BPNN algorithm to demonstrate the proposed methodology’s effectiveness and superiority. The methodology presented in this study focuses on the case of a single ignition source in tunnel fires, which can be extended to the case of multiple ignition sources in tunnels in future studies.