A Principal Component Approach in Diagnosing poor Control loop performance

• Main Authors: H., Zabiri, T.D.T. , Thao
• Format: Conference or Workshop Item
• Published: 2007
• Subjects: TP Chemical technology
• Online Access: http://eprints.utp.edu.my/3746/1/Microsoft_Word_-_ICCE_20.pdf; http://www.iaeng.org/publication/WCECS2007/WCECS2007_pp194-199.; http://eprints.utp.edu.my/3746/

Principal component analysis, both linear and nonlinear, are used to identify and remove correlations among process variables as an aid to dimensionality reduction, visualization, and exploratory data analysis. While PCA ascertains only linear correlations between variables, NLPCA reveals both linear and nonlinear correlations, without restriction on the character of the nonlinearities present in the data. In this paper, the use of PCA and NLPCA are investigated and compared for nonlinearity detection in regulated systems using routine operating data. Results from simulated and industrial data used in this study clearly show that NLPCA performance supersedes that of PCA in identifying and detecting nonlinearity in poor performing control loops.