Abstract: Fault isolation and propagation path identification can provide important information for steady operation of process.Traditional fault diagnosis methods based on contribution plots have problems that faulty variable cannot be isolated accurately,too many variables are identified as faulty variables,and faults cannot be detected in early stage.To solve these problems,a Causality Diagram Oriented Propagation Path Identification and Fault Isolation (CDPPIFI) method was introduced.Fault pre-detected was operated and suspected faulty variable was identified based on contribution plots.Then,starting from suspected faulty variable,causality among variables were analyzed and selected according to causality threshold.A causality diagram for the fault was constructed,and faulty variable was isolated based on this causality diagram.Statistics for faulty variable were calculated for fault detection.To verify the effectiveness of the proposed method,CDPPIFI was applied to a numerical simulation and continuous stirred tank reactor process.Results showed that this new method could accurately isolate faulty variable and identify propagation path.

Key words: causality diagram, fault isolation, propagation path identification, fault diagnosis

摘要： 故障定位和传播路径识别能够为维持系统稳定运行提供重要信息。传统基于贡献直方图的故障诊断方法存在着变量定位不准和早期故障不易被检测等问题。为解决上述问题,提出一种基于变量因果图的故障定位和传播路径识别方法(CDPPIFI)。具体地,首先进行故障预检测,在检测到故障后利用贡献直方图找出潜在的故障变量;然后,以潜在故障变量为起点,分析变量之间的因果关系,构建故障变量的因果图模型,对因果图模型进行溯源,定位故障源头;最后,对源头变量数据构建监测指标,从而实现早期故障检测。为了验证方法的有效性,设计了控制系统数值仿真和连续反应搅拌釜基准实验,结果表明所提出方法能够准确地定位故障源并识别故障传播路径。

关键词: 因果图, 故障定位, 传播路径识别, 故障诊断