Bearing performance degradation assessment by using LMD and CFS clustering without cluster number selection

A method based on local mean decomposition (LMD) with clustering by using a fast search (CFS) algorithm for roller bearings performance degradation assessment (PDA) is presented because CFS is rarely used in bearing PDA without clustering center point selection. Firstly, The LMD is used to decompose the original vibration signal into some production function components (PFs), then the effective first two PFs are selected according to the correlation coefficient and calculated by singular-value decomposition. Then the selected first two singular values are regarded as the input of the CFS for finding the clustering center point under different states of bearing health, namely “Normal”, “Slight”, and “Severe”. Finally, a health evaluation index (confidence value), which is obtained from the dissimilarity between the samples and the various clustering center points, is used to evaluate the health of the roller bearings. The results of the experiment show that the proposed method has a good ability to assess performance degradation well.