Abstract: Fault Tree Analysis (FTA) and Dynamic Bayes Network (DBN) were proposed to diagnose the urban gas pipeline accident. Making full use of FTA's good systematicness and logicality, the leakage accident scene model was built. Giving full play to the advantages of DBN's reverse reasoning ability and posterior probability expression accuracy, and establishing the corresponding FTA model by analyzing the possible causes of urban gas pipeline leakage accident, and converting it into DBN model and optimizing it, the multi -source observation information is finally combined to quickly identify and diagnose the accident scene. The reasoning conclusion is consistent with the results of the late accident investigation report, which indicates that the method has reasonable feasibility.

Key words: gas pipeline, leakage, risk analysis, fault tree analysis, dynamic Bayesian network, emergency management