基于改进贝叶斯网络的加油站火灾爆炸分析

为了找出导致加油站发生火灾爆炸事故的基本事件及其可能性大小，以加油站火灾爆炸故障树为基础建立相应的贝叶斯网络风险模型。在FTA向BN转化算法的基础上对条件概率做出了修正。利用GeNIe软件计算加油站火灾爆炸事故基事件的后验概率，同时进行灵敏度和影响力分析。最后通过实例分析，找出了导致某加油站发生火灾爆炸事故危险性最大的因素集为:加油站接打手机、机械碰撞、给塑料容器加油、加油冒油、油枪渗漏等。结果表明，注重基事件的多态性和事件间逻辑关系合理性的新模型，能推算出更准确的基事件概率分布，同时可以找出导致事故发生的最有可能途径，为加油站事故预防，系统改进提供较为合理性建议。

Accident analysis of fire and explosion in gas station based on improved Bayesian Network

In order to find out the basic events and the corresponding possibilities of fire and explosion in gas station, a Bayesian network risk model was established based on the fault tree of fire and explosion in gas station fire and explosion. The conditional probabilities were corrected in the transformation process of fault tree to Bayesian Network. The posterior probabilities of basic events in fire and explosion accidents of gas station were calculated by using the software GeNIe. The strength of influence and the sensitivity analysis were carried out simultaneously. The purpose was to identify more sensitive events and more influential events. Finally, one example was analyzed about fire and explosion in gas station. The main factors of the accidents were identified: using the phone at the gas station, mechanical collision, plastic containers refueling, excessive refueling, oil gun leakage, etc. The results showed that the more accurate probability distribution of basic events can be deduced, which takes into account the rationality of logical relationships among events and the features of multiple states for basic events. The sensitive basic events and the most likely pathway trigging the accident can be identified. It can provide the reasonable proposals for accidents prevention and system improvement of the gas station.