采用动态故障树分析诊断系统故障的信息融合法  

Information Fusion Method for System Fault Diagnosis Based on Dynamic Fault Tree Analysis

作　　者：段荣行[1] 董德存[1] 赵时旻 

机构地区：[1]同济大学交通运输工程学院,上海201804 [2]上海申通轨道研究咨询有限公司,上海201103

出　　处：《同济大学学报（自然科学版）》2011年第11期1699-1704,共6页Journal of Tongji University:Natural Science

基　　金：国家"863"高技术研究发展计划资助项目(2007AA11Z247);国家自然科学基金资助项目(61074139)

摘　　要：为提高系统故障诊断效率,提出了一种利用动态故障树分析诊断系统故障的信息融合方法,该方法充分发挥动态故障树建模和贝叶斯网络推理各自优势,通过集成系统结构信息和传感器信息来诊断系统故障.采用高效的零压缩二元决策图生成系统所有最小割集,并采用贝叶斯网络方法计算部件和最小割集的诊断重要度;根据传感器证据信息对系统特征函数化简,同时对部件和证据条件下割集的诊断重要度进行更新;综合考虑部件和割集诊断重要度设计了系统诊断决策算法,生成诊断决策树以指导维修人员恢复系统故障;最后通过实例验证了该故障诊断方法的有效性.An information fusion method was proposed to diagnose system faults with dynamic fault tree（DFT） analysis to improve the efficiency of system diagnosis,which made full use of the advantages of both DFT for modeling and Bayesian networks（BN） for the inference ability and incorporated system structure information as well as sensors data into fault diagnosis.All minimal cut sets were generated via an efficient zero-suppressed binary decision diagram,while the diagnostic importance factor of components and minimal cut sets were calculated using BN.Furthermore,these reliability analysis results together with the characteristic function of the system were updated after receiving the evidence data from sensors and used to develop diagnostic decision algorithm to optimize system diagnosis.Then,a diagnostic decision tree was generated to guide the maintenance crew to recover a system.Finally,an example was given to illustrate the efficiency of this method.

关 键 词：动态故障树 离散时间贝叶斯网络 诊断重要度 期望诊断代价 

分 类 号：TB114.3[理学—概率论与数理统计]