基于层次模块消元法的贝叶斯网络算法研究  

Research on bayesian network algorithm based on the hierarchical module elimination method

作　　者：汤规成[1,2] 许开立[1] 李德顺[3] 

机构地区：[1]东北大学资源与土木工程学院,辽宁沈阳1108191 [2]中国石油辽河石化公司,辽宁盘锦124002 [3]沈阳理工大学环境与化学工程学院,辽宁沈阳110159

出　　处：《中国安全生产科学技术》2012年第12期58-62,共5页Journal of Safety Science and Technology

基　　金：辽宁省教育厅科学研究一般项目

摘　　要：贝叶斯网络对于解决复杂设备不确定性引起的故障具有优势。考虑到多数情况下贝叶斯网节点的联接形式均为汇合联接,分析了常见的汇合联接形式,提出一种新的算法,即层次模块消元法。该算法首先根据装置故障形式及特点,构建贝叶斯网;然后将构建的贝叶斯网进行分层,各层组成一个新的模块;接着分别计算新模块在各种状态下的概率;最后将各层次模块结合在一起,求得叶节点各种状态下的概率。实例应用表明,层次模块消元法与桶消元法计算结果一致,适合装置各种状态概率计算。The advantages and disadvantages of the fault tree and bayesian network were discussed.Considering that node connection forms of the bayesian network are confluent connections in most cases,the confluent connections were analyzed,and a new algorithm whose name is hierarchy module elimination method was proposed.According to fault forms and characters of the equipment,the bsyesian network was constructed firstly,then a new module was formed by constructed layers;secondly,the probabilities of the new module in various state were calculated respectively.At last,the probabilities of the leaf node in various states were obtained.The practical case showed that the probability of leaf node from the hierarchical module elimination is the same as the probability of leaf node from bucket elimination method,so the hierarchical module is suitable for calculating probabilities of device fault in various states.

关 键 词：故障树 层次模块 贝叶斯网 故障概率 

分 类 号：X913.4[环境科学与工程—安全科学]