Fault Diagnosis for Dynamic Nonlinear System Based on Variable Moving Window KPCA

Abstract

Kernel Principal Component Analysis (KPCA) is a noteworthy nonlinear extension of the most popular dimensionality reduction methods, Principal Component Analysis (PCA). It has been extensively used for process monitoring. The time varying property of industrial processs require the adaptive ability of KPCA. The Variable Moving Window KPCA (VMWKPCA) is developed to monitor the dynamic processes. This new method is based on the variation of the size of the moving window depending on the normal change of the system. For fault diagnosis a set of structured partial VMWKPCA were used. The fault detection and diagnosis with the proposed VMWKPCA are tested using the Continuous Stirred Tank Reactor (CSTR] process. The simulation results proved that the new method is effective for fault detection and diagnosis,