动态不确定因果图在化工系统动态故障诊断中的应用

摘要

In chemical processes, it is necessary to effectively diagnose the fault on time in order to avoid losses of economy and lives. Dynamic uncertain causality graph ( DUCG) is a method, which represents and infers the dy⁃namic, uncertain causalities of the process system according to directed graph. Based on the characteristics of pro⁃cessing information, DUCG has its own advantages for fault diagnosis in chemical processes on a large scale. There⁃fore, this article applies DUCG to realize fault diagnosis of chemical processes by constructing the object system knowledge base and probabilistic reasoning on fault data. The data transmission module of the former DUCG system is improved to deal with the vibrational signals in the chemical process, and to widen the scope of application. The Tennessee Eastman ( TE) simulator is taken as the experimental subject to test the effectiveness of DUCG methodol⁃ogy and software. 54 variables and 114 causalities are included in the constructed DUCG knowledge model. Accord⁃ing to this model, all the failures simulated by TE are diagnosed in a high probability of ranking. The correct diag⁃nosis rate is 100%. In comparison of Bayesian Network ( BN) , the mean correct diagnosis rate is 79.71% reported⁃ly, showing that DUCG is an effective method.%为了避免化工工程中经济及生命的损失，有效及时检测出故障是十分必要的。动态不确定因果图（ DUCG）是一种根据有向图实现动态不确定因果关系表达与推理的方法。其处理信息的特性，对于目前规模庞大的化工过程故障诊断有着自身的优势。因此运用DUCG，通过构建对象系统知识库、对故障数据进行概率推理，实现化工过程的故障诊断，并针对化工过程的震荡信号，对原DUCG系统的数据发送模块做出改进，使之适用范围更全面。为了验证DUCG理论的有效性，采用TE过程作为实验对象，建立包含54个变量、114条因果关系的DUCG模型。该模型对TE过程中的故障得到较高诊断排序概率，诊断正确概率达到了100％，与贝叶斯网络的平均诊断正确概率79．71％相比，说明了DUCG是一种行之有效的方法。