LAUSR

This work proposes a novel real-time detection scheme for incipient stator inter-turn short circuit fault in voltage sources inverter-fed induction machines. Both non-sinusoidal input voltage and the short circuit fault causes harmonics in the motor stator current and these combined harmonic components complicate the spectral analysis-based diagnosis in inverter-fed motors. Aim of the analysis is to identify the effect of inverter fundamental/switching frequency on early detection and classification of the inter-turn fault. Discrete wavelet transform based analysis is performed on stator current using daubechies1 wavelet and statistical parameter L2 norm has been computed for the detailed and approximate coefficients at different decomposition levels to obtain the most precise feature of fault. Support vector machine-based learning algorithm is used for the accurate classification of the incipient fault. The proposed method is independent of switching and fundamental frequency, the modulation index and mechanical load. Real-time detection is possible even with infinitesimal fault current of 350 mA by the proposed method. The competency of the proposed algorithm is validated using simulation and verified by hardware with VSI-fed induction motor drive.