A gear fault feature extraction method based on adaptive AR model and MOMEDA

Gear fault features are submerged in strong background noise with the influence of background noise and vibration signal transmission paths. The adaptive auto-regressive (AR) model with the largest spectral kurtosis (SK) can effectively eliminate the linear stationary part of the signal, which is selected for signal preprocessing. Multipoint Optimal Minimum Entropy Deconvolution Adjusted (MOMEDA) is suitable for the extraction of periodic fault signal fault features in rotating machinery. Therefore, the adaptive AR and MOMEDA are combined to extract gear fault features, which is verified by experimental results.