Broken rotor bar detection in inverter-fed induction motors by time-corrected instantaneous frequency spectrogram

Abstract

Fault detection in induction motors in non-stationary regimes managed by inverters is an existing industrialized demand. Inverter-fed machines have being used due to their adjustable velocity and quick dynamic response, although the injection of objectionable frequency components. The condition monitoring of these motors can significantly reduce the cost of maintenance in the early detection of faults. Under this special performing situation, many of the identification methods researched have problems in distinguishing induction motor failures. This paper introduces an identification methodology founded on the mixed handling of two techniques: Time-corrected instantaneous frequency and the Spectrogram, also known as reassigned spectrogram, where a thresholding is used so that any element with values smaller than the signal-noise-rate are set to zero, and the noise is reduced in the final spectrogram. The suggested method is tested in an inverter-fed induction motor at the time of startup continue by a steady-state condition. It was confirmed its effectiveness to find one broken rotor bar. From the obtained results, the suggested method confirms to be precise enough to identify the failure progression in the time-frequency domain under various operating conditions (startup and steady-state conditions) in inverter-fed induction motors.