An Adaptive Cross-Validation Thresholding De-Noising Algorithm for Fault Diagnosis of Rolling Element Bearings Under Variable and Transients Conditions

Developing a proper approach for extracting the fault characteristics of rolling element bearings is a significant behavior in variable operation conditions. Generally, without a tachometer, the tacholess order tracking method is well established by instantaneous rotation frequency curve estimation using time-frequency analysis. However, the fault features are often masked in strong background noises. An adaptive cross-validation thresholding de-noising algorithm is employed to improve the performance of the traditional envelope order spectrum method. First, the general linear chirplet transform is applied to estimate the instantaneous rotation frequency curve. Second, with the help of the instantaneous rotation frequency curve, the raw non-stationary signal is transferred to an angular domain by the re-sampling technique. Third, the adaptive cross-validation thresholding de-noising algorithm is employed to purify the angular domain signal to improve the fault feature extraction performance of the envelope order spectrum method. Comparisons using numerical simulations and experimental investigations of bearing faults under variable speed conditions are given to show the superiority of the present method.