Novel Intelligent Data Processing Technology, based on Nonstationary Nonlinear Wavelet Bispectrum, for Vibration Fault Diagnosis

A novel integrated data processing algorithm for vibration fault testing for electromechanical devices, energy systems and engineering structures, based on the spectral kurtosis and the nonstationary nonlinear higher order wavelet bispectrum (WB), is proposed and investigated. A novel adaptive systematic approach for identification of frequency ranges for the WB is also proposed, investigated and successfully experimentally validated. Experimental validation of the proposed data processing technology is performed, using measured data, related to non-faulty rolling bearings and bearings at an early fault stage. The high effectiveness of early bearing diagnostics by the proposed nonlinear data processing technology has been experimentally demonstrated, using the Fisher criterion and probability of correct identification. Important advantage of the proposed technology is that it could be employed for data processing and identification of electromechanical devices and structures with unknown a priori frequency characteristics.