EWT WITH AUTOCORRELATION FEATURE ENHANCEMENT IN FEATURE EXTRACTION OF WEAK FAULT OF GEARBOX AND ITS APPLICATION

Using empirical wavelet transform (EWT) to decompose the weak gear fault signal, the gear meshing frequency and its sidebands may be divided into different frequency bands, which leads the separated AM-FM components to be unsatisfactory. Therefore, a new method based on the improved EWT and autocorrelation denoising feature enhancement is proposed. By the improved EWT method, the fault signal was decomposed into different AM-FM components. As a result, the included meshing frequency and its sidebands of an AM-FM component are located in the same frequency band. Then the appropriate AM-FM component was selected for autocorrelation denoising in order to enhance the analysis effect of improved-EWT method. Finally, the weak impact characteristic hidden in the original fault signal was clearly highlighted through envelope spectrum of the denoised signal. Simulation and experimental analysis results verify the effectiveness and advantage of this proposed method in the gear low-frequency weak fault feature extraction.