基于符号有向图与趋势分析的故障诊断框架  

Signed directed graph and trend analysis based framework for fault diagnosis

作　　者：高东[1] 许欣[1] 马昕[1] 张贝克[1] 

机构地区：[1]北京化工大学信息科学与技术学院自动化系,北京100029

出　　处：《计算机工程与应用》2014年第13期6-9,108,共5页Computer Engineering and Applications

基　　金：中央高校基本科研业务费项目(No.ZY1208)

摘　　要：针对传统基于符号有向图(SDG)的故障诊断方法可能遗漏真正故障以及分辨率不高的问题,提出了一种SDG与定性趋势分析相结合的故障诊断框架。故障发生后,通过提取变量的定性趋势来获取变量的状态,并根据基于定性趋势的相容规则进行反向推理,找到所有可能的故障。对诊断结果按照可信度系数(Confidence Index,C.I.)进行排序,提高诊断的分辨率。案例研究表明该框架能够保证诊断结果完备性并有效地提高诊断分辨率,可以应用到实际化工过程故障诊断中。For the problems of the real faults being missed and poor diagnosis resolution in Signed Directed Graph（SDG） based fault diagnosis, a SDG and qualitative trend analysis based framework has been proposed. When faults occur, the qualitative trends are extracted to represent the states of variables. And then inverse inference is carried out to find all the possible faults using consistence rules, which are basing on qualitative trends. The diagnosis resolution can be improved by ranking the results according to Confidence Index（C.I.）. The case study proves that the diagnosis resolution is improved and diagnosis completeness is satisfied using the framework. It can be applied in fault diagnosis in chemical process.

关 键 词：符号有向图 定性趋势分析 故障诊断 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]