基于贝叶斯网络的概率安全评估方法研究

针对传统的事件树/故障树分析方法的局限性,提出了基于贝叶斯网络的系统概率安全评估方法.该方法采用了一体化的建模思想,将故障树和事件树转化为贝叶斯网络进行分析,不仅可以得到常用的安全性分析结果,还可以得到其他有用的信息,非常便于推理和诊断.而且该方法还可以处理多态、相依、非单调和非确定性逻辑关系,有很强的适用性.通过1个油井平台瓦斯泄漏实例说明了该方法的有效性,并与传统的故障树/事件树分析方法和基于二元决策图的分析方法进行了比较.

Probabilisfic safety assessment research based on Bayesian networks

According to the deficiency of event tree/fault tree analysis,a new probablilistic safety assessmet method based on Bayesian networks is proposed.The event trees and fault trees are mapped into Bayesian networks and the general safety analysis results and other useful information can be obtained for reasoning and diagnosis.The new method can deal with multi-state,dependency,non-monotonic and uncertainty correlations in safety analysis.The comparison with event tree/fault tree and BDD(binary decision digram)-based methods are carried out by means of an example of gas leak accident of oil well platform taken from literature,which demonstrates the effectivness of the method.