基于混合贝叶斯网络的混合系统安全性分析方法

安全关键系统的安全性分析模型本质上是离散和连续失效分布共存的混合模型.传统的故障树和马尔科夫链分析方法仅能处理离散分布或指数分布的系统，难以对混合系统进行安全性分析.针对该问题，以DFT系统安全模型为基础，提出一种基于混合贝叶斯网络的混合系统安全性分析新方法.首先，利用狄拉克函数和单位阶跃函数分别表示动态故障树节点间的确定性关系和时序关系，将动态故障树转换为贝叶斯网络.然后，通过分段多项式来拟合网络节点的不同失效分布，提出一种k段n次多项式混合贝叶斯网络来表示动态故障树.最后，给出该混合贝叶斯网络的推理算法.实验分析表明本方法能有效地进行混合系统安全性分析.

A Novel Safety Analysis Method of Hybrid System on Hybrid Bayesian Network

The safety analysis model of critical system is essentially a mixed model of both discrete variables and continuous variables.The traditional analysis methods can only deal with the system based on discrete distribution or exponential distribution,so these methods are incapable of analyzing the safety of the hybrid system.To solve the problem,this paper presents a novel safety analysis method of hybrid system on hybrid Bayesian network(HBN).First,by using the Dirac function and unit step function to represent the deterministic relation and timing sequence of nodes in DFT respectively,we convert the DFT into a Bayesian network(BN).Second,The HBN with k-piece and n-degree polynomials is proposed to represent the DFT,in which the different failure distributions of nodes are fitted by piecewise polynomial functions.Finally,the inference algorithm of HBN is proposed.The experimental results show that the presented method can effectively solve the safety analysis of hybrid system.