Risk prediction and diagnosis of urban gas pipeline accidents based on polymorphic fuzzy Bayesian network

Abstract

In order to evaluate the risk level of the urban gas pipeline system, and provide the reference for follow-up prevention efforts, a quantitative analysis method of gas pipeline accident risk was proposed based on polymorphic fuzzy Bayesian network. Firstly, risk factors were sorted out from 86 accident investigation reports, so that the city gas pipeline risk element system was established. Subsequently, the fault tree model was built to seek the match between risk hazards and accidents, which can convert into the Bayesian network structure. After that, fuzzy set theory and probability distribution method were introduced to calculate the prior probability of the root node and the conditional probability of the intermediate nodes, evidence-based inference of Bayesian network was used to predict the probability of accidents, analyze the importance of risk elements, and reverse diagnose key causal factors. Finally, this method was applied to the risk analysis of the “10·21” large pipeline gas leakage accident in Shenyang. The results of the case validation show the a priori probability of the accident is 688%, which verifies the effectiveness of the risk system. Besides, important risk elements derived from prediction and backward diagnosis are consistent with the direct causes analyzed in the accident investigation report. The polymorphic fuzzy Bayesian network approach for gas pipeline system risk can evaluate gas pipeline accident risk accurately and identify key risk-causing factors, which provides some reference for decision making in the safety management of city gas pipelines.

Keywords

• risk evaluation and failure analysis; gas piping system; fuzzy theory; polymorphic fuzzy bayesian networks; probability distribution