Multivariate Frequency Transfer Bispectrum Estimator for Gearbox Drive System Fault Diagnosis Using Motor Current Signature Analysis

Bispectrum analysis is considered as a typical motor current signature analysis tool, has been widely concerned in fault feature extraction. However, current bispectrum utilizes the entire spectrum to demodulate fault feature components. Moreover, it applies vertical slicing to obtain fault information, which tends to cause ambiguity in the frequency components. This paper proposed a novel multivariate frequency transfer bispectrum estimator (MFTBE) for gearbox drive system fault diagnosis. Firstly, the current signals are processed using phase space reconstruction to construct the Hankel matrix and calculate its bispectrum signals. Subsequently, the resonance frequencies are selected as the frequency transfer value from bispectrum signals to construct the frequency transfer bispectrum, which is synchronously averaged to form the multivariate frequency transfer bispectrum (MFTB). Finally, the integral operation is utilized on the MFTB to obtain the MFTBE to extract gearbox drive system fault features. Numerical simulations and experiment data are utilized to verify the practicability and effectiveness of MFTBE. Analysis results prove that MFTBE is effective in extracting fault features and its performance is better than other well-advanced algorithms.