A time-frequency method for multiple fault detection in three-phase induction machines

It is well known that induction machine stator current is a nonstationary signal and its properties change with respect to operating conditions. The computed spectrum using the fast Fourier transform (FFT) does not provide accurate time-domain information about the operating conditions. As a result, the FFT spectrum analysis makes difficult to recognize fault conditions from the normal operation of the induction machine. A time-frequency approach is proposed in order to track some frequency components associated to electrical and load faults. This method is applied to a 18.5 kW three-phase induction machine with three broken rotor bars and under the effect of load torque variation which can result from a load fault.