Research on Deep Learning of Tunnel Fire under the Effect of Cross Passage Exhaust

For a single-hole bidirectional highway tunnel with a cross passage for smoke exhaust, FDS numerical simulation software was utilized to analyze fire scenarios under different angles of cross passage and longitudinal position of the fire source, to determine the optimal smoke extraction conditions. Based on tunnel fire scenarios involving smoke exhaust with cross passage, a rapid prediction model for fire smoke parameters based on the Transformer was established. The results indicate that when the angle of cross passage is 60°, while the longitudinal position of the fire source is 0 m, the effect on controlling smoke spread is optimal. The Transformer-CNN and Transformer-GRU fire smoke parameter prediction models, constructed based on FDS numerical simulation results, can achieve rapid and accurate predictions of longitudinal temperature, visibility, and smoke height within the tunnel.