A method based LMD-SVD-CFS algorithm for roller bearings performance degradation assessment

This paper introduces an approach to evaluating the performance degradation of roller bearings using the LMD-SVD-CFS algorithm. Firstly, the original vibration signals are processed using the LMD technique to obtain some production function components (PFs). Following this, the top two PFs are selected based on their correlation coefficient and calculated by singular-value decomposition (SVD). Secondly, SV1 and SV2, the top two selected singular values (SVs), are utilized as the input parameters for the fast CFS. Finally, based on the differences between the sample features and the predefined clustering centers, a confidence value (CV) is calculated, which serves as a measure of the bearing performance state. The results of the experiment show that the proposed method outperforms the above time-domain indicators and clustering methods in detecting the early-stage degradation more precisely, without the need for presetting the number of clusters.