AI-based Evaluation Method of Mechanical Performance of Shield Tunnel Structures after Fire

An accurate and efficient evaluation of post-fire performance of tunnel structures is crucial for ensuring the reliability and cost-effectiveness of emergency disposal and repair work. In this study, an evaluation model is established to assess the post-fire bearing capacity of shield tunnels based on Artificial Intelligence technology. The key damage indices and their influence on post-fire behavior of shield tunnels are investigated based on Analytic Hierarchy Process and numerical simulations. The mechanical performance of tunnels after fire is classified using Q-type system clustering analysis method. The quantitative relationship between damage indices and performance grades is determined by BP neural network. It is found that the main factors influencing the performance of shield tunnel structures after fire include depth of spalling, depth of concrete deterioration, reduction of concrete strength, reduction of bolt strength, area of spalling, area of concrete deterioration, and number of bolt deterioration. A framework of five grades of fire damage is proposed according to post-fire relative residual bearing capacity of tunnels. The BP neural network can accurately and efficiently predict the performance grade of shield tunnels after fire, providing a useful tool for making wise and fast decision on post-disaster transportation and repair strategy.