基于动态贝叶斯网络的机载电子设备故障诊断

Fault Diagnosis of Airborne Electronic Equipment Based on Dynamic Bayesian Networks

针对目前机载电子设备故障诊断过程中诊断效率低以及采用传统动态故障树马尔科夫链分析方法存在系统状态空间爆炸的问题,提出了一种基于动态贝叶斯网络的故障树故障诊断方案;设计首先将基于零压缩二元决策图的动态故障树定性分析和贝叶斯网络的定量推理相结合获得系统最小割集,然后以集成传感器信息更新系统的部件诊断重要度和最小割集,最后综合考虑系统部件诊断重要度和最小割集设计了系统的故障诊断决策算法,得到故障诊断决策树;以机载光电雷达设备的故障诊断为例,通过对比有无证据条件下系统故障诊断中最小割集以及其诊断重要度,证明了此方案能够准确、快速地诊断出系统具体的故障原因,节省了诊断成本。

In order to solve the problem of low diagnostic efficiency and explosion of system state space by using traditional dynamic fault tree analysis method of the Markov during the process of the fault diagnosis of current airborne electronic equipment, it is proposed that a fault diagnosis design of fault tree based on Dynamic bayesian networks. Firstly, the design can combine dynamic fault tree qualitative analy- sis based on a zero compressed binary decision diagram and Bayesian network quantitative reasoning to obtain the minimal cut sets. Secondly, the degree of diagnostic importance of minimal cut sets and components are updated by integrated sensor information. Finally, considering the diagnosis importance degree of components and cut sets, a decision algorithm of system fault diagnosis is designed and the fault diagnosis decision tree is gained. Taking the fault diagnosis of photoelectric radar equipment as an example, by comparing minimum cut sets and diag- nostic importance of the fault diagnosis system under the conditions of the presence of evidence, it shows that the design can diagnose the specific failure reasons accurately and quickly and save the cost of diagnosis.