Application of statistics filter method and clustering analysis in fault diagnosis of roller bearings

Condition diagnosis of roller bearings depends largely on the feature analysis of vibration signals. Spectrum statistics filter (SSF) method could adaptively reduce the noise. This method is based on hypothesis testing in the frequency domain to eliminate the identical component between the reference signal and the primary signal. This paper presents a statistical parameter namely similarity factor to evaluate the filtering performance. The performance of the method is compared with the classical method, band pass filter (BPF). Results show that statistics filter is preferable to BPF in vibration signal processing. Moreover, the significance level alpha would be optimized by genetic algorithms. However, it is very difficult to identify fault states only from time domain waveform or frequency spectrum when the effect of the noise is so strong or fault feature is not obvious. Pattern recognition is then applied to fault diagnosis in this study through system clustering method. This paper processes experiment rig data that after statistics filter, and the accuracy of clustering analysis increases substantially.