Nonlinear Fault Detection Based on An Improved Kernel Approach

• Published in: IEEE access Vol. 6; pp. 11017 - 11023
• Main Authors: Wang, Guang, Jiao, Jianfang
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Computational modeling; Eigenvalues; Eigenvalues and eigenfunctions; Fault detection; Iterative methods; Kernel; kernel-based; Kernels; Monitoring; Nonlinear process; Principal component analysis; Product design; Quality assessment; quality-related; Singular value decomposition; Subspaces
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2018.2802939

Quality-related issue is a recently raised subject that attracts a lot of attention in process monitoring community. Since most industrial processes present more or less nonlinear characteristics, the study of nonlinear quality-related methods is thus very necessary. Most of the existing methods are based on a kernel partial least square (KPLS) model; however, they usually have a very large amount of computation due to the iterative computation of KPLS. To make matters worse, the logic of these methods is complex, since they use four subspaces to detect a fault. In this paper, we will propose a new kernel-based method whose computation only involves eigenvalue solution and singular value decomposition. Besides, it has a simple logic using only two subspaces. What is more, it has a stable performance with high computational efficiency. All these advantages of the new method are demonstrated by simulation results.