A new Burg method based approach to MCSA for broken rotor bar detection

This paper presents a new approach to Motor Current Spectral Analysis (MCSA) for broken rotor bar detection. Said approach consists of applying the Discrete Fast Fourier Transform on stator current data of a three-phase induction motor, and analyzing the amplitude spectrum obtained for characteristic broken rotor bar sidebands. The difference from traditional techniques is borne out by the very short time window over which the stator current is sampled, which enables early detection of broken bars. This short-time current data is extrapolated to a longer time window using signal modeling and prediction via the Burg method, thus mitigating the shortcomings of the DFFT over short data windows. The spectrum of the extrapolated signal shows adequate frequency resolution for effective fault detection.