State-space residual based monitoring

Mercer, E; Martin, E; Morris, A

doi:10.1016/S1570-7946(02)80149-3

State-space residual based monitoring

Lookup NU author(s): Ewan Mercer, Professor Elaine Martin, Emeritus Professor Julian Morris

Downloads

Full text for this publication is not currently held within this repository. Alternative links are provided below where available.

Abstract

Although the process performance monitoring tools of dynamic Principal Component Analysis (PCA) and Canonical Variate Analysis (CVA) take into account process dynamics, the monitoring statistics still contain serial correlation. Consequently the traditional statistical basis for the calculation of control limits will be invalid resulting in either missed out-of-control signals or an excess of false alarms. A methodology is proposed whereby a CVA state-space model is first developed and then a PCA based monitoring scheme is formed using the model mismatch. In this case, the residuals will be independent and identically distributed and the standard control limits will be valid. The methodology is demonstrated on the benchmark Tennessee Eastman problem. © 2002 Elsevier B.V. All rights reserved.