Browse by author
Lookup NU author(s): Professor Elaine Martin,
Emeritus Professor Julian Morris
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
Multivariate statistical representations have been widely used in the process manufacturing industries for process performance monitoring, in particular for the detection of changes in current operation and the onset of process disturbances or faults. Applications of the technology have focused to a lesser extent on manufacturing processes where drift occurs over time as part of normal process operation, e.g., due to reactor fouling, machine wear, ramping of temperatures during process operation, and changes due to set-point adjustments. In this paper, an extension to the methodology based on the statistical projection technique of principal component analysis (PCA) is proposed for the monitoring of processes where drift and set-points changes are common place, i.e., exponentially weighted PCA. The technique is illustrated through its application to a polymer film manufacturing process where the representation is required to adapt quickly to changes in the process that are part of normal operating procedures, but remain sensitive to the detection of deviations from normal operation.
Author(s): Lane S, Martin EB, Morris AJ, Gower P
Publication type: Article
Publication status: Published
Journal: Transactions of the Institute of Measurement and Control
Print publication date: 01/01/2003
ISSN (print): 0142-3312
ISSN (electronic): 1477-0369
Publisher: Sage Publications Ltd.
Altmetrics provided by Altmetric