基于主元分析残差空间的自适应统计方法及其应用

Adaptive Statistical Method in Residual Subspace of PCA-Based and Its Application

提出一种基于主元分析(PCA)残差空间的自适应统计方法。将原始空间分解为主元空间和残差空间,然后根据残差空间得分变量方差变化趋势,给出由Hotelling’s T^2和欧式距离相结合的自适应统计量T_e^2。将PCA-T_e^2、PCA-Q以及核主元分析(KPCA)-Q等方法应用于田纳西-伊斯曼(TE)过程中,仿真结果表明PCA-T_e^2具有较高的检测性能。

An adaptive statistical method in the residual space of principal component analysis（ PCA）-based is proposed in this paper. First,the original space is separated into principal component space and residual space. Next,in residual space,an adaptive statistic variable,Te^2,which is the combination of Hotelling＇s T^2 and Euclidean distance,is given. The proposed method,PCA-Q and kernal pricipal component analysis（ KPCA）-Q are applied to the Tennessee Eastman（ TE） process. The comparison of monitoring results shows that the proposed method is superior to PCA-Q and KPCA-Q.