Analysis of extended partial least squares for monitoring large-scale processes

This brief analyzes the recently proposed extended partial least squares (EPLS) algorithm and reveals that it does not: 1) allow the generalized score variables to be geometrically interpreted, 2) reconstruct the recorded process variables, and 3) produce statistically independent variables for process monitoring. To overcome these deficiencies, an improved EPLS algorithm is introduced, which utilizes generalized scores to identify statistical monitoring models. The brief finally presents an industrial application study of a chemical reaction process to show that improved EPLS offers enhanced diagnosis of abnormal process behavior.