基于动态贝叶斯网络的燃气管网燃爆风险分析

针对传统风险分析方法无法实现动态评估的局限性,提出了基于动态贝叶斯网络的燃气管网风险分析模型.基于管道失效原因与事故后果,构建了事故演变全过程网络结构,并运用马尔科夫理论将演变全过程与时间相关联,最终建立了动态贝叶斯风险分析模型.该模型基于贝叶斯理论进行泄漏、致灾模式及后果节点的概率计算,实现了对燃气管网事故的原因诊断及事故发展态势预测.以松原"7.4"燃气爆炸事故为例,应用本模型演示了燃气管网事故推演技术,结果表明:事故是由第三方施工破坏引发燃气泄漏,由于燃气在土壤中的扩散范围不断扩大,扩散范围内受限空间数量不断增加,人员活动情况也更加复杂,进而增大了燃气聚集与点火概率;结合事故现场条件,本模型对这起爆炸概率进行了推演,认为此次燃气爆炸概率将在泄漏60 min后达到42.3%,高于安全泄放概率;该结果与真实情况基本相符,进一步验证了模型的可行性与可靠性. 

Risk Analysis of Burning and Explosion of Gas Pipeline Network Based on Dynamic Bayesian Network

In view of the limitation that the traditional risk analysis method cannot realize dynamic assessment, a risk analysis model of the gas pipeline network based on dynamic Bayesian network was proposed. Based on the causes of pipeline failure and accident consequences, the network structure of the whole process of accident evolution was constructed, and the Markov theory was used to associate the whole process of evolution with time, and finally a risk analysis model based on the dynamic Bayesian network was established. The model calculates the probability of leakage, disaster mode and consequence node based on the Bayesian theory, and realizes the cause diagnosis and accident developing trend prediction of the gas pipeline network accident. Taking "7.4" gas explosion accident in Songyuan city as an example, this model is used to demonstrate the accident deduction technology of the gas pipeline network accident. The results show that the accident is resulted from gas leakage caused by the destruction of the third-party construction, and because the diffusion range of gas in the soil is constantly expanding, the number of restricted spaces in the diffusion range is increasing, and the situation of human activities is more complex, thus increasing the probability of gas accumulation and ignition. Combined with the conditions of the accident scene, the model deduces the probability of the explosion and concludes that the probability of the gas explosion will reach 42.3% after 60 minutes of leakage, which is higher than the safe release probability. The result is basically consistent with the actual situation, which further verifies the feasibility and reliability of the model.