Abstract: In this study, the original vibration signals of an Svenska Kullager-Fabriken (SKF) bearing's inner race, rolling element, and outer race with three fault sizes were extracted from Case Western Reserve University Bearing Data Center. The fault sizes were 0.007, 0.014, and 0.021 in(1 in=2.54 cm). By empirical mode decomposition (EMD), 17 intrinsic mode functions were found. Principal component analysis (PCA) was then conducted, and it was found that the relationship between the fault sizes of the inner and outer races and the first and second principal components could be accurately fitted by principal component fitting formulas. Therefore, the fault sizes can be obtained by real-time monitoring of vibration signals and signal analysis. Finally, the residual life of the bearing with a fault was predicted using the Paris-Erdogan formula and finite element simulation. This study has a significant meaning for prevention of equipment accidents caused by bearing cracks.

Key words: bearing, fault size, vibration signal, life prediction