基于贝叶斯网的复杂系统故障诊断方法  

Methods on Complex System Fault Diagnosis Based on Bayesian Network

作　　者：杨昌昊[1] 竺长安[1] 胡小建[2] 

机构地区：[1]中国科学技术大学,合肥230009 [2]合肥工业大学,合肥230009

出　　处：《中国机械工程》2009年第22期2726-2732,共7页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(70631003;70971035);安徽省自然科学基金资助项目(070416241)

摘　　要：针对复杂系统故障诊断时表现出的事件多态性、故障逻辑关系不确定性以及信息的不确定性等特点,在故障贝叶斯网基础上建立了基于故障贝叶斯网的复杂系统故障诊断方法。基于故障贝叶斯网的复杂系统故障诊断是通过故障推理实现的,包括推理结构转变、信念初始化、信念传播与故障概率计算等过程,建立了推理结构转变、信念初始化、信念传播与故障概率计算方法,并以转子故障诊断为例验证了诊断方法的高效可行性。In allusion to characteristics which exhibits in complex systems, such as multi--states of event, uncertainty of fault logic relations and information etc. when the faults were diagnosed. Fault diagnosis methods based on Bayesian network were built on complex systems on the basis of the fault Bayesian network, and the methods pointed out that fault diagnosis of complex systems based on Bayesian network was carried out through fault reasoning, which included processes of transforming inference structure, initializing beliefs, propagating beliefs and computing probability distributions, and computation methods on them were set up. Methods proposed were accounted for with an example of rotor fault diagnosis.

关 键 词：贝叶斯网 故障诊断 推理 故障树 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置] N941.4[自动化与计算机技术—控制科学与工程]