Research On Accident Consequence Assessment Of The City Gas Pipeline Failure Based On Bayesian Networks

It exits the dangers of explosion and fire and poisoning when conveying gas.The most dangerous of gas accident was pipeline leaks that became the main hidden trouble to affect the city gas pipeline.It was a frequent phenomenon that the casualties and facilities destroyed accident happen accompanied by the conveying or use of gas.Setting up a model is to assessment for the city gas pipeline failure and to combine it with the probability of pipeline failures to improve the measure of prevent accident occur and to provide a scientific referenced for risk manager.So it is necessary for assessing the serious degree of pipeline failures consequence.The objective of apply BN(Bayesian Network)to the evaluation of gas pipeline accident consequence that is to establish a way of forecast the serious degree of pipeline failures consequence.Firstly,Based on the former study,the paper analyzed gas pipeline accident consequence and carried out a preliminary discussion of evolvement rule of gas pipeline accident consequence.According to researched on the result of fires and explosion,including casualties and accident loss and environmental impact and social impact.The four facts were a measure of the serious degree of pipeline failures consequence.Starting from the whole process of gas pipeline failure accident,and put forward the main factors influencing the serious degree of pipeline failure consequence.To provide theoretical basis for established the BN model of accident consequence assessment.Secondly,The Bayesian networks was constructed,including node selected and determine logistics nodes.The most important is that determine logistics nodes by using machine learning and expert knowledge to construct the BN.Thirdly,parameter determine of BN,according to the 25 gas pipeline failure accident investigation report.Corresponding to each node variables and accessing sample data of BN node according to the scope of each node variables.And then the sample statistics was used to calculate the BN nodes of conditional probability network nodes.It was concluded that each node variables of conditional probability tables,so as to construct a Bayesian Network assessment model for gas pipeline accident consequences.Finally,combining an application example,the validity of model was verified.Through the concrete analysis,the application of Bayesian Network inference function to evaluate building accident consequences.Verifying the validity of the model by contrast the model assessment of accident consequences and the actual accident consequences.It was concluded that Bayesian networks can be applied to the evaluation of accident consequence.