基于PSR和SVM方法的化工过程单变量故障诊断

Single variable based fault diagnosis in chemical processes using phase space reconstruction (PSR) and support vector machines (SVM)

基于相空间重构(PSR)和支持向量机(SVM)算法本文提出了一种利用单一变量进行化工过程故障诊断的方法。首先进行变量筛选,然后对筛选出的关键变量进行相空间重构,再利用SVM对重构后的数据进行故障分类。通过对TE(Tennessee Eastman)过程几类故障进行仿真测试,结果表明在单一故障和多故障情况下,本方法均可实现化工过程的单变量故障诊断;与传统SVM方法相比,相空间重构可有效提高诊断正确率。此方法可为建立简单而有效的单变量故障诊断系统提供理论依据。

In chemical processes, usually only a few variables are responsible for a process fault. In order to construct a simple yet effective fault diagnosis system, a method for single variable based fault diagnosis in chemical processes using the phase space reconstruction （PSR） algorithm and support vector machines （SVM） has been proposed. The model was used to classify the faults of the Tennessee Eastman process （TEP）. The data transformed by PSR from the original selected single variable are used as the input of SVM. The results showed that for both single and multiple faults, PSR afforded a remarkable improvement in the performance of SVM for single variable based fault diagnosis in chemical processes.