Rédiger un PREreview

With the large-scale construction and operation of urban underground utility tunnels in China, the associated safety issues cannot be ignored. This paper proposes a risk assessment method for cable duct fires by optimizing the Bayesian model and combining bidirectional inference and decision support to numerically analyze the impact of different ventilation and fire protection measures on fires, and evaluate risk levels using the risk matrix method. During forward and diagnostic reasoning analysis, the predicted probability of a cable duct fire occurring is 4.967‰, and the probabilities of the cable duct fire developing to various stages are 3.533‰, 1.378‰ and 1.232‰, respectively. Bayesian network-assisted decision analysis yields the maximum causation chain for cable duct fires as: insulation damage (X6) → poor connection (D1) → leakage (C2) → cable core overheating (B1) → cable self-ignition (A1) → cable ignition (T). Then, through Bayesian network probability self-adaptive analysis, the probability of cable duct fire occurrence and the probabilities of various consequence scenarios increase over time, with the upward trend reflecting the deteriorating operating conditions of the cable duct. Finally, based on the dynamic Bayesian model and inference, the predicted probability of cable ignition increases from 4.967‰ at T0 to 14.69‰ at T29.