Home
Scholarly Works
Multivariate Forecasting of Batch Evolution for...
Conference

Multivariate Forecasting of Batch Evolution for Monitoring and Fault Detection

Abstract

To monitor a batch process, which is dynamic in nature, it is necessary to consider the time varying relationship of its variables throughout the entire run. MPCA models built with batch wise unfolded data have been used extensively for batch process monitoring, these methods will not only consider the known samples to asses the ongoing batch run, but will also consider a dynamic forecast of the future unknown samples. Such forecast, implicit in the methodology, is uncovered and analyzed in this work; and proven to be a powerful feature of a batch-monitoring scheme built with MPCA and the batch-wise unfolded matrix of batch data.

Authors

Muñoz SG; Kourti T; MacGregor JF

Volume

37

Pagination

pp. 71-76

Publisher

Elsevier

Publication Date

January 1, 2004

DOI

10.1016/s1474-6670(17)31796-2

Conference proceedings

IFAC-PapersOnLine

Issue

9

ISSN

2405-8963
View published work (Non-McMaster Users)

Contact the Experts team