Wavelet-based Bispectra for Motor Rotor Fault Detection

Wavelet-based bispectral analysis is addressed for condition monitoring of induction machines. This advanced signal processing technique combining wavelet analysis and bispectral techniques allows the detection and characterization of non-Gaussian and non-stationary signals with time resolution and the discrimination linear processes from nonlinear ones. In the present investigation, application of this new technique to detect and identify an induction machine's rotor faults by measuring vibration data and analyzing the nonlinearity of the machine, due to the fact that damaged or faulty machines often generate highly nonlinear signals. The usefulness and statistical robustness of this approach are confirmed in the experiment. The results and analysis indicate that this novel signal processing technique can be effectively applied to motor rotor fault detection.