Moving window KPCA with reduced complexity for nonlinear dynamic process monitoring. (September 2016)

• Record Type: Journal Article
• Authors: Jaffel, Ines; Taouali, Okba; Harkat, Mohamed Faouzi; Messaoud, Hassani
• Is Part Of: ISA transactions. Volume 64(2016:Sep.)
• Page Start: 184
• Page End: 192
• Publication Date: 2016-09
• Subjects: KPCA -- RKPCA -- MW-RKPCA -- Nonlinear dynamic process -- Fault detection; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2016.06.002
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: This paper proposes an improved Reduced Kernel Principal Component Analysis (RKPCA) for handling nonlinear dynamic systems. The proposed method is entitled Moving Window Reduced Kernel Principal Component Analysis (MW-RKPCA). It consists firstly in approximating the principal components (PCs) of the KPCA model by a reduced data set that approaches "properly" the system behavior in the order to elaborate an RKPCA model. Secondly, the proposed MW-RKPCA consists on updating the RKPCA model using a moving window. The relevance of the proposed MW-RKPCA technique is illustrated on a Tennessee Eastman process. Highlights: This paper proposes a new Reduced Kernel Principal Component Analysis (RKPCA) technique entitled Moving Window RKPCA (MW-RKPCA). The proposed MW-RKPCA provides lower computation time and memory complexity. It consists on using RKPCA technique to select a reduced set of observations that 'sufficiently' approaches the system behavior. The updating procedure of the RKPCA model is achieved using a moving window technique. The proposed MW-RKPCA algorithm has been evaluated on Tennessee Eastman process.