东南大学学报（自然科学版）2013，Vol.43Issue(z1)：93-97,5.DOI:10.3969/j.issn.1001-0505.2013.S1.020
可变窗自适应核主元分析的化工过程故障诊断算法
Fault diagnosis algorithm of chemical industry process based on variable window adaptive KPCA
摘要
Abstract
As the kernel principal component analysis (KPCA) method is easy to generate false alarm for nonlinear and time-varying chemical process,a variable window fast adaptive kernel principal component analysis algorithm is proposed.When large amounts of data blocks are provided,the proposed algorithm first updates and downdates the KPCA model separately,then adjusts the size of the moving window by calculating SPE and T2 statistics to further update the KPCA model.This algorithm overcomes the weakness of the traditonal adaptive KPCA that it can only handle the data of one observation at one moment,and it can effectively eliminate the impact of abnormal samples,thuS improving the accuracy of fault detection.The algorithm is applied to monitor the ketone-benzol dewaxing process.Compared with the KPCA and the moving window KPCA (MWKPCA) methods,simulation results show that the proposed method can decrease the false alarm rate and has more reliable performance for nonlinear and time-varying chemical process.关键词
故障诊断/核主元分析/可变窗/酮苯脱蜡Key words
fault diagnosis/ kernel principal component analysis/ variable window/ ketone-benzol dewaxing分类
信息技术与安全科学引用本文复制引用
赵小强,杨武,薛永飞..可变窗自适应核主元分析的化工过程故障诊断算法[J].东南大学学报（自然科学版）,2013,43(z1):93-97,5.基金项目
甘肃省自然科学基金资助项目(1112RJZA028)、甘肃省教育厅硕士生导师资助项目(1003ZTC085). （1112RJZA028）
