事件树、故障树、决策树与贝叶斯网络  

Event tree,fault tree,decision-making tree and Bayesian network

作　　者：周建方[1] 唐椿炎[1] 许智勇[1] 

机构地区：[1]河海大学机电学院,江苏常州213022

出　　处：《河海大学学报（自然科学版）》2009年第3期351-355,共5页Journal of Hohai University(Natural Sciences)

基　　金：武汉大学水资源与水电工程科学国家重点实验室开放基金(2006B029);国家"十一五"科技支撑计划(2006BAC14B07)

摘　　要：在简要介绍贝叶斯网络技术的基础上,通过大坝失效事件树分析、导弹发动机故障树分析以及汽车销售决策树分析3个实例,分别将事件树、故障树及决策树3种分析方法与贝叶斯网络分析方法进行了比较,并给出了事件树、故障树和决策树向贝叶斯网络转化的一般规律:事件作为贝叶斯网络中的结点,根据事件之间的因果或影响关系将网络中的各结点用有向弧连接起来并由已知数据或专家经验确定各结点条件概率表.结果表明贝叶斯网络具有处理多状态复杂模型以及双向推理的优点.Based on an introduction to the Bayesian network and the analysis of three examples, an event tree of a dam failure, a fault tree for unexpected ignition of a missile engine, and a decision-making tree for sale of cars, the methods of the event tree, the fault tree and the decision-making tree were compared with the Bayesian network. A general law was developed to translate the event tree, the fault tree and the decision-making tree into the Bayesian network. Firstly, in the Bayesian network the nodes denoted the events. Secondly, the nodes were connected with directed arcs according to the causality and influence of the events. Finally, the conditional probability table for the nodes was determined in terms of the given data or experts＇ experience. The results show that the Bayesian network has advantages in dealing with complex multi-state models and bi-directional reasoning.

关 键 词：事件树 故障树 决策树 贝叶斯网络 

分 类 号：N945.1[自然科学总论—系统科学]