文献类型：期刊文献

中文题名：PCA_SVM的多故障分类方法在TE过程中的应用

英文题名：Multi_faults classification method based on PCA_SVM and its application to the TE process

作者：何菲[1];杜文莉[1];钱锋[1]

机构：[1]华东理工大学,化工过程先进控制和优化技术教育部重点实验室,上海200237

年份：2010

卷号：27

期号：10

起止页码：1321

中文期刊名：计算机与应用化学

外文期刊名：Computers and Applied Chemistry

收录：CSTPCD;;北大核心:【北大核心2008】；CSCD:【CSCD2011_2012】；

基金：国家自然科学基金(208760442);国家863计划(2008AA042902);上海市科技项目(08DZ1123100);长江学者和创新团队发展计划资助;高等学校学科创新引智计划(B08021);上海市重点学科(B504).

语种：中文

中文关键词：故障检测;多分类;PCA;过程监控;SVM;KNN

外文关键词：fault detection; multiple classification; PCA; process monitoring; KNN; SVM

摘要：故障识别在化工过程监控中有着至关重要的作用。准确的故障识别能够帮助操作员及时发现并排除故障,避免生产事故的发生。本文运用传统PCA方法对TE过程中5种典型故障数据进行降维,并将所有故障数据投影到正常工况样本的PCA主元空间中,由正常数据样本计算出T^2统计量的闽值,根据Hotelling T^2统计原理,对所有故障数据进行检测,将检测到的故障样本通过SVM的多分类方法进行故障分类。通过TE过程仿真平台的实验表明,PCA SVM方法与PCA KNN、C SVM方法相比较,算法简单,容易实现,计算速度较快,并且突破了很多文献中只有2类或3类故障识别的局限,同时可以达到较高的多分类准确率。
Fault identification is significantly important in the chemical process monitoring.Exact fault identification can help operators to realize process fault and remove it timely in order to avoid production disaster.In this paper,traditional PCA method is used for reducing the dimension of five classical fault of TE process.All the fault sample data are projected into the normal data＇s principle space.Then all the faults are detected by the Hotelling T^2 statistical method.SVM classification method is applied to identify the detected fault data.The proposed approach also compares with PCA_KNN classification method and the C_SVM mentioned by E.Tafazzoli and M.Saif.Experimental results show that the proposed methods are with higher accuracy,simple algorithm and fast computation speed,the classification result is satisfactory.