摘要
传统的事件树、故障树和静态贝叶斯网络在事故风险分析中存在一定的局限性,动态贝叶斯网络对于描述多态性、相关性、时序性、交互行为的复杂系统在节点的安全风险以及预测未来一段时间内事故发生概率方面较为突出。提出将动态贝叶斯网络模型与模糊数学方法相结合的思路,分析氨制冷系统预防氨泄漏爆炸事故风险的方法。先根据氨泄漏爆炸事故故障树,利用GENIE软件建立了氨泄漏爆炸事故风险评估的模糊动态贝叶斯网络(Fuzzy-DBN)模型;然后利用模糊数学方法确定动态贝叶斯网络模型中的条件概率参数,利用动态贝叶斯网络算法推理计算各根节点发生的后验概率,并进行前向推理,预测事故发生概率;最后将该风险评估模型应用于某氨泄漏爆炸事故风险分析,通过案例分析验证了该模型的可行性。得出导致氨泄漏爆炸事故发生的关键火源因素及泄漏因素,并基于动态贝叶斯网络,根据氨制冷系统发生氨泄漏报警情况,对氨泄漏爆炸事故发生概率进行预测。
Due to the limitations of traditional event trees,fault trees,and static Bayesian networks in accident risk analysis,dynamic Bayesian networks are useful for describing complex systems with polymorphism,correlation,timeliness,and interactive behavior at the nodes.The safety risk assessment and prediction of the probability of accidents in the future will be more prominent.This paper proposes the idea of combining a dynamic Bayesian network model with fuzzy mathematical methods.Based on the fault tree for the leakage and explosion of ammonia,the paper applies GENIE software to establish a fuzzy dynamic Bayesian network model,and applies fuzzy mathematical methods to determine the conditional probability parameters in the dynamic Bayesian network model,and applies dynamic Bayesian network algorithm to calculate the posterior probability of each root node,and perform forward reasoning to predict the accident probability.At the end,the paper applies this risk assessment model to the risk analysis of ammonia leakage explosion accidents,and obtain the key ignition source factors and leakage factors that lead to the occurrence of ammonia leakage explosion accidents,and obtain the accident probability of source factors and leakage factors.Based on the dynamic Bayesian network,the ammonia refrigeration alarm system is based on ammonia leakage alarm conditions,the probability of leakage and explosion accidents is predicted,and the case analysis shows that the mathematical model has good feasibility.
作者
程洁
许开立
陈守坤
徐晓虎
CHENG Jie;XU Kaili;CHEN Shoukun;XU Xiaohu(School of Resources and Civil Engineering,Northeastern University,Shenyang110819,China)
出处
《安全与环境工程》
CAS
北大核心
2020年第5期147-152,164,共7页
Safety and Environmental Engineering
基金
“十三五”国家重点研发计划项目(2018YFC0808406、2017YFC0805100)。
关键词
氨泄漏爆炸事故
风险分析
动态贝叶斯网络(DBN)
模糊数学
氨制冷系统
事故发生概率
leakage and explosion of liquid ammonia
risk analysis
Dynamic Bayesian Network(DBN)
fuzzy mathematics
ammonia refrigeration system
accident probability
