基于贝叶斯网络的复杂系统动态故障树定量分析方法

动态故障树的贝叶斯网络分析方法存在局部组合爆炸和备件门节点失效时间仅能是指数分布的不足.首先,给出动态故障树转换为离散时间贝叶斯网络的方法,该方法使用一个确定性函数来替代条件概率表,避免了局部组合爆炸.然后,根据备件门的失效机理和对应的贝叶斯网络结构特征,解决了备件节点失效时间仅能是指数分布的限制.最后,提出一种基于动态故障树的贝叶斯网络精确推理算法,基于该算法给出了系统失效分布、组件重要度等概率计算.实验结果表明,该方法能有效地分析和评估安全攸关系统的概率特性.

Quantitative Analysis Method of Dynamic Fault Tree of Complex System Using Bayesian Network

There exist limitations of local combinatorial explosion and only exponential distribution of spare nodes in Bayesian network(BN)-based dynamic fault tree(DFT)analysis method.First,an approach of mapping DFT into discrete-time BN is proposed in which a deterministic function instead of conditional probability tables is used to avoid local combinatorial ex-plosion.Second,according to the failure mechanism and BN structure of spare door,we remove the limitation that the failure time of spare nodes in BN is only exponential distribution.Finally,an exact inference algorithm of DFT-based BN is presented and based on which the failure distribution of system and the importance measurement of components is calculated.Experimental re-sults show that the proposed method can analyze and evaluate the probability characteristics of safety-critical systems effectively.