基于动态不确定因果图的化工过程动态故障诊断及状态预测

发布时间：2018-02-04 10:45

本文关键词： 化工过程动态不确定因果图故障诊断故障状态预测概率推理　出处：《北京化工大学》2015年硕士论文　论文类型：学位论文

【摘要】：化工过程具有规模大、高复杂性、高风险等特点,为了避免经济及生命的损失,在故障发生时,有效及时地检测出故障是十分必要的。动态不确定因果图(Dynamic Uncertain Causality Graph, DUCG)是用有向图表达不同变量之间因果关系并进行推理的理论,可通过图形简洁地表达过程的因果关系并根据构建的图形知识库给出故障事件的概率推理。由于其能根据证据信息降低复杂度,并能处理不完备知识及不确定的证据信息,对于目前规模庞大的化工过程故障诊断有着自身的优势。基于上述背景,本文所做的研究工作如下：①运用DUCG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。②采用了TE (Tennessee Eastman)过程作为实验对象,建立了包含54个变量、114条因果关系的DUCG模型。该模型对TE过程中的故障都得到较高诊断排序概率,仿真结果达到了100%,说明了DUCG是一种行之有效的方法。③在故障诊断的基础上对故障的异常变量及状态进行预测,预测结果虽与实际结果不完全一致,但可预测出大部分实际异常的变量,预测止确率达到70%,说明了该功能具有一定的有效性。
[Abstract]:Chemical process has the characteristics of large scale, high complexity, high risk and so on. In order to avoid the loss of economy and life, when the fault occurs. It is necessary to detect faults effectively and in time. Dynamic Uncertain Causality Graph. DUCGG is a theory that uses digraph to express causality and reason between different variables. The causality of the process can be expressed succinctly by the graph and the probability inference of the fault event can be given according to the constructed graphic knowledge base, because it can reduce the complexity according to the evidence information. And can deal with incomplete knowledge and uncertain evidence information, for the current large-scale chemical process fault diagnosis has its own advantages. Based on the above background. The research work of this paper is as follows: 1. Using DUCG to realize the fault diagnosis of chemical process, and aiming at the oscillation signal of chemical process, the data sending module of the original DUCG system is improved. The te Tennessee Eastman process was used as the experimental object, and 54 variables were established. DUCG model of 114 causality. This model has a high probability of diagnosis and ranking for the faults in te process, and the simulation results reach 100%. It shows that DUCG is an effective method to predict the abnormal variables and states of faults on the basis of fault diagnosis. The prediction results are not completely consistent with the actual results. However, most of the actual abnormal variables can be predicted, and the accuracy rate of prediction reaches 70, which shows that this function is effective.
【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TQ02;TP277

【参考文献】