Gearbox Fault Feature Extraction Method Based on SWT and Fast Kurtogram

Abstract

Using continuous wavelet transform (CWT) to decompose the fault signal of gearbox with multi-stage transmission, it is found that the resolution of CWT time-frequency distribution is inferior. In addition, the traditional spectral kurtosis parameters need to be estimated manually. In order to solve the above two problems, a new method based on synchrosqueezing wavelet transform (SWT) and fast kurtogram is proposed. The complex gear signal was decomposed on multiple layers with SWT to obtain several intrinsic mode components (IMCs). Then, appropriate components were selected for signal reconstruction and fast band-pass filtering with fast kurtogram with appropriate parameters. Finally, the filtered signal is subjected to envelope demodulation analysis to extract the gear low-frequency fault characteristics, and the fault diagnosis for the multi-stage transmission gearbox was realized. Through experimental verification, this proposed method can improve the concentration of the signal frequency curve and enhance the time-frequency distribution resolution, and it is very suitable for the gearbox fault diagnosis with impact fault.