Sensor fault detection and isolation of an industrial gas turbine using partial adaptive KPCA. (April 2018)

• Record Type: Journal Article
• Authors: Navi, Mania; Meskin, Nader; Davoodi, Mohammadreza
• Is Part Of: Journal of process control. Volume 64(2018)
• Page Start: 37
• Page End: 48
• Publication Date: 2018-04
• Subjects: Adaptive kernel PCA -- Aeroderivative gas turbine -- Dynamic systems -- Fault detection and isolation (FDI); Process control -- Periodicals; Fabrication -- Contrôle -- Périodiques; Process control; Periodicals; Electronic journals; 660.281
• DOI: 10.1016/j.jprocont.2018.02.002
• Languages: English
• ISSNs: 0959-1524

Abstract

Highlights: The problem of sensor fault detection and isolation for nonlinear dynamical systems is studied. The adaptive KPCA approach is developed to detect the occurred faults. Isolating of the detected faults is done using the partial structured residual concept. The proposed partial AKPCA methodology is applied to an industrial gas turbine. Abstract: In this paper, sensor fault detection and isolation of time-varying nonlinear dynamical systems is studied by utilizing an adaptive kernel principal component analysis (KPCA) solution as a useful method to overcome the weaknesses of conventional KPCA approach in dealing with time-varying dynamical processes. Toward this goal, adaptive Hotelling's T 2 is used with KPCA to tackle the time-varying behavior of nonlinear systems. Moreover, for fault isolation, partial adaptive KPCA (AKPCA) is proposed where a set of residual signals is generated based on the structured residual set framework. The simulation studies demonstrate that using the proposed methodology, the occurrence of sensor faults in the nonlinear dynamic model of an aeroderivative gas turbine can be effectively detected and isolated in the presence of component degradation.