A High Resolution Method for On Line Diagnosis of Induction Motors Faults

The wide use of the induction motors for industrial and transport applications leads to increase the interest in their monitoring for on-line detection of incipient faults. The traditional diagnostic techniques are based on the stator current or torque spectral analysis by FFT-based methods. However, the well known spectral leakage phenomena of FFT can cause not accurate estimation of the spectral components; moreover, its frequency resolution problems limits a fast detection of the fault condition. In order to solve these problems, this paper presents an high resolution method, the ESPRIT method, to make the fault identification faster and accurate. A proper dynamic model of the induction machine is used to simulate the occurring of a fault condition. Simulation results allow to compare the ESPRIT and FFT-based methods, considering the case of some rotor bars broken.