Smart Tunnel Fire Temperature Prediction Method with Fusion of Golden Eagle Optimization, Logistic Map, and Lévy Flight Mechanism

Abstract

Tunnel ceiling temperature prediction is an important issue in fire safety construction. However, due to the diversity of tunnel fires, there are few accurate and applicable temperature prediction methods. Therefore, an improved golden eagle optimization (GEO) algorithm is proposed in this paper, which is applied to the prediction of ceiling temperature distribution and peak temperature during tunnel fires. The algorithm can predict the ceiling temperature field in tunnel fires based on only a small number of sensors. In the initialization phase of the GEO algorithm, a logistic map is incorporated to enhance the diversity and thoroughness of the population distribution. Additionally, the Lévy flight mechanism is introduced during the position update process to strengthen both the global and local search capabilities. Unlike traditional methods, this approach does not rely on large data sets for training and can accurately predict the ceiling temperature field with a limited number of sensors. The proposed method’s performance was validated by full-scale tunnel fire experiments. Comparative analysis was conducted between the improved GEO algorithm, the standard GEO algorithm, and two traditional swarm optimization algorithms. Results show that the improved GEO algorithm offers higher prediction accuracy.