Recursive partial least squares algorithms for monitoring complex industrial processes

The monitoring of processes that exhibit non-stationary and/or time varying behaviour is discussed in this paper. It is shown that the application of recursive partial least squares (RPLS) algorithms together with adaptive confidence limits can lead to a considerable reduction in the number of false alarms. The integration of these algorithms into the multivariate statistical process control (MSPC) framework is introduced and its extensions to multi-block approaches is discussed. Example studies are given with respect to a simulation of a fluid catalytic cracking unit and the analysis of data obtained from an industrial distillation process. (C) 2003 Elsevier Science Ltd. All rights reserved.