一种解决组合爆炸问题的三态故障贝叶斯网络

A Three-State Faulty Bayesian Network for Solving Combinational Explosion Problem

针对传统故障贝叶斯网络在可靠性分析以及故障诊断领域存在的组合爆炸问题,提出了一种三态故障贝叶斯网络建模方法。通过将传统的失效分为"物理故障"和"正常却不工作"2种模式,提出将串联系统中元件以及"或"门下的事件转换为包含正常状态以及上述2种失效模式的三态节点,取代仅有正常和失效2种状态的传统模型节点。在此基础上深入分析了构建三态故障贝叶斯网络模型的方法。最后建立了某型飞机左侧交流电源系统的三态故障贝叶斯网络并进行了深入分析。理论和实例分析均表明,建立的三态故障贝叶斯网络在减少模型参数的同时可有效避免"或"门引起的组合爆炸问题,为串联系统和"或"门占优的系统提供了一种新的故障贝叶斯网络建模手段。

Aiming at the combinational explosion problem of faulty Bayesian network in the field of reliability analy-sis and fault diagnosis, we propose a method of building a three-state faulty Bayesian network.By dividing tradition-al failure into the two modes of “physical failure” and “normal without working”, we propose three-state nodes converted from the events of OR gates and the components of the series system, thus replacing the two-state nodes of traditional models that just contain the modes of normal and failure.On the basis of that, we study in depth a method for building a three-state faulty Bayesian network.Finally a three-state Bayesian network representing the left AC power system of a certain aircraft is built and analyzed in depth.Both the analysis of the theory and the case show preliminarily that the three-state faulty Bayesian network proposed by us reduces the number of parameters while avoiding the combinational explosion problem caused by OR gates; this can be a new modeling means of building faulty Bayesian networks for OR gates-dominant systems and series systems.