Suppression of Fundamental Frequency Component Through a Novel Kalman Filter for On-Line Diagnosis of Induction Motor Faults

For diagnosis of induction motor faults through motor current signature analysis, spectral leakage of the fundamental frequency component (FFC) is one of the key factors deteriorating its reliability, because in the spectrum of the stator current (SC) signal, the widened peak of FFC can easily cover up some fault characteristic harmonics, especially the sideband harmonics (SBHs). Herein, a novel model containing both linear and nonlinear parts is proposed for developing the Kalman filter (KF) with enhanced FFC suppression performance. The developed KF is a combination of the extended KF (EKF) and linear KF (LKF). When implementing the KF, the EKF based on the nonlinear part of the model is first exerted on the supply voltage signal to estimate the fundamental frequency (FF) of the power grid. Secondly, based on the linear part of the model, the LKF utilizing the estimated FF as its inherent parameter is exerted on the SC signal to estimate the FFC. After subtracting the estimate of FFC from the SC signal, the peak of FFC in the spectrum is suppressed significantly, leading to effective identification of SBHs, even at the slip as low as 0.5%. The KF based on the novel model proposed in our research exhibits much more excellent FFC suppression effect than KFs based on purely (non)linear models.