Recursive Partial Least Squares Algorithms For Monitoring C外语英文电子书.pdfVIP

Control Engineering Practice 11 (2003) 613–632 Recursive partial least squares algorithms for monitoring complex industrial processes a b, a Xun Wang , Uwe Kruger *, Barry Lennox a School of Engineering, University of Manchester, MI3 9PL, UK b Intelligent Systems and Control group, Queen’s University of Belfast, BT9 5AH, UK Receive d27 September 2001; accepte d21 May 2002 Abstract The monitoring of processes that exhibit non-stationary and/or time varying behaviour is discusse din this paper. It is shown that the application of recursive partial least squares (RPLS) algorithms together with adaptive confidence limits can lea dto a considerable reduction in the number of false alarms. The integration of these algorithms into the multivariate statistical process control (MSPC) framework is introduce dan dits extensions to multi-block approaches is discussed. Example studies are given with respect to a simulation of a flui dcatalytic cracking unit an dthe analysis of data obtaine dfrom an industrial distillation process. r 2003 Elsevier Science Ltd. All rights reserved. Keywords: Adaptive algorithms; Data reduction; Process models; Fault detection; Statistical process control 1. Introduction Applications of PCA or PLS for process monitoring are base don a predefine dPCA or PLS model that has Modern industrial processes, often present a large been pro