Bearing early fault feature extraction with adaptive weighted Savitzky-Golay filtering

Early bearing fault characteristics are usually not obvious, and are easy to be interfered by external noise due to complex working conditions and external vibration. Savitzky-Golay filter has high computational efficiency and inherent feature retention ability, which is very suitable for fault feature extraction under low SNR conditions, but the current parameter selection standard is not robust enough to limit its engineering application. Aiming at the problem of early bearing fault feature extraction, an adaptive weighted Savitzky-Golay filter fault extraction method is proposed in this paper. Firstly, the parameter adaptive selection sequence of S-G filtering is given. Secondly, according to the structure characteristics of S-G filter, the frame of adaptive weight allocation based on window length is determined. Finally, the experiment verifies that the S-G filter after adaptive parameter optimization and weighted processing can accurately and efficiently obtain the bearing early fault characteristics under the condition of low SNR.