Browse by author
Lookup NU author(s): Daniel Castro-Rodriguez,
Professor Elaine Martin,
Professor Gary Montague
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
Multivariate statistics are being increasingly used for batch and continuous process monitoring and fault detection. Of the variety of methods available, Principal Component Analysis (PCA) has been one of the most widely employed. Different approaches have been formulated to take account of the dynamics of the process being monitored and the presence of outliers and/or missing data in the information gathered. However, outlier detection approaches have failed to address the situation when outliers exist in dynamic data. The present work proposes a robust dynamic PCA approach to address this problem. The technique is tested with data from a simulation of a polyvinyl acetate CSTR operating under normal and faulty conditions. Results are compared with those obtained using PCA, Dynamic PCA and Robust PCA. It is found that the proposed technique is more sensitive for fault detection than the other approaches considered when using data from the simulated process.
Author(s): Castro-Rodriguez DA, Martin EB, Montague GA
Publication type: Conference Proceedings (inc. Abstract)
Publication status: Published
Conference Name: AIChE Annual Meeting
Year of Conference: 2006
Library holdings: Search Newcastle University Library for this item