A Novel Fault Detection Method Based on the UKF and Its Application to a Fifth-Order Two-Phase Nonlinear Motor System

Abstract

Small faults are difficult to be detected under noisy conditions. And the small sensor faults will introduce modeling errors in the observation equations. Therefore, small faults need be detected successfully and quickly. A novel small fault detection (FD) method is proposed using the residuals generated by the unscented Kalman filter (UKF) for a fifth-order two-phase nonlinear motor. Firstly, the introduction of the UKF is given. Secondly, on the basis of the UKF, a fault detection scheme based on a local approach is proposed. The local approach is used to detect small faults from residuals obtained from the UKF. Besides, the comparison between local approach and generalized likelihood test approach is introduced to illustrate the effectiveness of the proposed method. Finally, the proposed fault detection method is applied to detect faults of a fifth-order two-phase nonlinear motor system.