Extraction and enhancement of unknown bearing fault feature in the strong noise under variable speed condition

Rolling bearings often run under variable speed condition, in addition to constant speed condition. How to achieve the bearing fault diagnosis under variable speed condition has been an important and hot issue. Nevertheless, there are few works on bearing fault diagnosis under variable speed condition especially for the feature extraction of unknown fault. Thus, this paper proposes a method based on fractional Fourier transform (FRFT) and stochastic resonance (SR) to extract bearing fault features. First, we use FRFT filtering algorithm to extract fault formation from the original signal. Next, we apply zero centering and high pass filtering to the signal which contains the fault information. Since the separated fault information is usually relatively weak and is not easy to identify, SR is used to enhance the weak fault feature information. Finally, bearing fault is diagnosed by observing the fault characteristic frequency in the time-frequency distribution plane. The method can achieve the extraction of the bearing fault characteristic frequency in the unknown situation and meanwhile remove a lot of noise interference. The method has been validated by numerical simulations and experimental analyses, where the scratches on both outer race and rolling element can be diagnosed successfully. By comparison with previous methods, fast kurtogram and variable mode decomposition, fault features extracted by the proposed method are much clearer and more accurate. The method may provide reference for the application of fault diagnosis in engineering occasions.