Dynamic Process Monitoring Method Based on Recursive Generalized Eigenvalue Decomposition Using Temporal Covariance Matrix

摘要

A new dynamic process monitoring method for flow industry production was investigated. The proposed method utilized generalized eigenvalue decomposition (GED), which involved temporal covariance matrix pencil, to deal with the dynamic character of process. The recursive approach of GED was also used to make the algorithm wieldier to practical applications. Compared with Dynamic PCA, the new method performs better in sensitivity and robustness of the monitoring effect and can be hardly affected in computation cost when involving more temporal samples. The simulation using Tennessee-Eastman process shows the validity and superiority of the proposed method.