Vulnerability Evaluation Of Urban Gas Pipelines System Based On BP Neural Network

Currently our study of system safety was conducted from the perspective of risk,rather than to consider how to improve the system from the nature of security. But thelong-term safety management proved that, evaluating system security purely from theperspective of risk to and making decisions is not enough. In order to study the intrinsicsafety of urban gas pipeline network, the article referencing the index system andevaluation method of risk evaluation for urban gas pipeline network, proposed conceptand study ideas of vulnerability which is suitable for urban gas pipeline network system.According to the industry characteristics of the urban gas pipeline network andrelated areas of research, the paper established the vulnerability index system of urbangas pipeline network with4layers from the angles of physic, structure and socialfunctions, and the secondary layer index included “Sensitivity Degree” and “CopingCapacity”. The paper divided indexes to5levels, and identified the scope of each levelthrough field investigation and expert discussions. Then, with the samples, the papertrained the network model of “Sensitivity Degree” and “Coping Capacity” applying BPneural network by MATLAB R2010a. The paper established the vulnerability functionmodel. Finally, the vulnerability of urban gas pipeline network is divided into fivegrades, used to evaluate the vulnerability of urban gas pipeline network, and eachgrades had its own output scope.The urban gas pipeline network system of a town in Chongqing was applied,and the output numerical value was0.81, meaning “no vulnerability”. Comparing withthe result by fuzzy comprehensive evaluation method and the operation condition of thepipe network in recent10years, their results are consistent. It indicates that the urbangas pipeline network system vulnerability index system and the model established wereappropriate and reliable to evaluate the urban gas pipeline network system.The model built is able to scientifically deal and calculate sample data, and resultsare automatically output, with strong generalization ability, which is better be used toevaluate the unknown samples. The paper provides a new idea for vulnerabilityassessment of urban gas pipeline network.