去主元相关性DKPCA故障检测与诊断方法  

Fault Detection and Diagnosis Method Based on Removing Principal Component Correlation DKPCA

作　　者：张成[1] 韩宏宇 李元 ZHANG Cheng;HAN Hong-yu;LI Yuan(School of Science,Shenyang University of Chemical Technology,Shenyang 110142,China;School of Computer Science and Technology,Shenyang University of Chemical Technology,Shenyang 110142,China;School of Information Engineering,Shenyang University of Chemical Technology,Shenyang 110142,China)

机构地区：[1]沈阳化工大学理学院,辽宁沈阳110142 [2]沈阳化工大学计算机科学与技术学院,辽宁沈阳110142 [3]沈阳化工大学信息工程学院,辽宁沈阳110142

出　　处：《计算机技术与发展》2023年第4期161-167,共7页Computer Technology and Development

基　　金：国家自然基金科学项目(61673279);辽宁省自然科学基金项目(2019-MS-262);辽宁省教育厅基金项目(LJ2019013)。

摘　　要：针对动态核主元分析(Dynamic Kernel Principal Component Analysis,DKPCA)在动态非线性过程监控中没有降低数据动态性的影响,导致统计量T^(2)具有显著自相关性的问题,提出一种基于去主元相关性的DKPCA(Dynamic Kernel Principal Component Analysis based on Removing Principal Component Correlation,DKPCA-RPCC)故障检测与诊断方法。首先,对原始数据X进行时滞扩展生成增广矩阵Y并使用KPCA计算主成分M;其次,利用已知数据重构增广矩阵Y,再使用KPCA计算主成分M;然后,通过主成分之间的差异来构建统计量进行故障检测;最后,利用基于变量贡献图的方法进行故障诊断。通过数值例子和田纳西-伊斯曼(Tennessee-Eastman,TE)过程进行仿真验证,并将仿真结果与KPCA、DPCA和DKPCA的结果进行对比。仿真结果说明,该方法在动态非线性过程监控中构建的统计量故障检测性能更高且具有较低的自相关性。Aiming at the problem that Dynamic Kernel Principal Component Analysis(DKPCA)does not reduce the influence of data dynamics in dynamic nonlinear process monitoring,resulting in significant autocorrelation of statistic T^(2),a fault detection and diagnosis method for DKPCA based on removing principal component correlation(DKPCA-RPCC)was proposed.Firstly,the original data X was time-delayed to generate the augmented matrix Y,and the principal component M was calculated by KPCA.Secondly,the known data was used to reconstruct the augmented matrix Y and then calculate the main component M by KPCA.Then,the difference between principal components was used to construct statistics for fault detection.Finally,the method based on variable contribution graph was used for fault diagnosis.Numerical examples and Tennessee-Eastman(TE)process were used for simulation verification,and the simulation results were compared with those of KPCA,DPCA and DKPCA.The simulation results show that the proposed method has higher fault detection performance and lower autocorrelation in dynamic nonlinear process monitoring.

关 键 词：动态核主元分析 过程监控 自相关性 去主元相关性 故障诊断 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]