基于KPCA与LS-SVM的化工过程故障诊断算法研究  

A fault diagnosis algorithm of chemical industry process based on KPCA and LS-SVM

作　　者：解庆[1] 杨武[2] 赵小强[2] 

机构地区：[1]甘肃蓝科石化高新装备股份有限公司,兰州730070 [2]兰州理工大学电气工程与信息工程学院,兰州730050

出　　处：《工业仪表与自动化装置》2012年第5期3-7,共5页Industrial Instrumentation & Automation

基　　金：甘肃省自然科学基金项目(1112RJZA028);甘肃省教育厅硕士生导师项目(1003ZTC085)

摘　　要：针对核主元分析方法(KPCA)在复杂化工在线监控过程中初始故障源难以辨识的问题,该文提出了一种基于核主元分析和最小二乘支持向量机的集成故障诊断方法。该方法首先运用KPCA对数据进行预处理,在特征空间构建T2和SPE来检测故障的发生,然后计算样本的非线性主元得分向量,将其作为最小二乘支持向量机的输入值,通过最小二乘支持向量机的分类进行故障类型的识别。将上述故障诊断方法应用到Tennessee Eastman(TE)化工过程,多种故障模式下的仿真结果表明,该方法不但能有效地辨识故障,而且提高了故障检测和故障诊断的速度。When kernel principal component analysis（KPCA） method is applied to the on-line moni- toring of complex chemical industrial process, primary fault sources are usually difficult to identify. So an integrated fault identification method based on KPCA and least squares support vector machine （ LS - SVM） is proposed. First the data are analyzed by using KPCA, T2 and SPE are constructed in the feature space for detecting faults. If T2 and SPE exceed the predefined control limits, a fault may have occurred. Then nonlinear principal component score vectors of the samples are calculated and inputted into least squares support vector machine to identify the faults through least squares support vector machine classifi- cation. The proposed method is applied to the Tennessee Eastman （TE） chemical industry process. Simu- lation results of multiple fault modes demonstrate that the proposed method can not only effectively identi- fy various types of fault sources , but also improve speed of fault detection and diagnosis.

关 键 词：化工过程 故障诊断 核主元分析 最小二乘支持向量机 

分 类 号：O212.4[理学—概率论与数理统计]