Fault Feature Enhancement of Rolling Bearings Based on Singular Spectrum Decomposition

Weak fault information features regularly in a defective rolling bearing. Consequently, fault diagnosis of rolling bearings is always challenging. Based on stationarity and linearity, traditional methods for data analysis are scarcely suitable for processing bearing fault data. Although applied to investigate nonstationary and nonlinear data, either of EMD and EEMD faces mode mixing. For overcoming the shortcoming, this paper introduced singular spectrum decomposition (SSD), a new method for analyzing nonstationary and nonlinear data, to examine bearing fault data and then proposed a novel method for fault feature enhancement of bearings based on SSD. Afterwards, the performance of the proposed method was benchmarked against each of envelope analysis, EMD and EEMD numerically and experimentally. Thus, the comparison indicates that SSD outperforms the other methods in retrieving physically interpretative components as a result of restraining mode mixing. Therefore, the proposed method demonstrates the potential for enhancing fault features of bearings.