Performance Degradation Assessment of Rolling Bearings Based on OLPP and SVDD

An approach based on orthogonal locality preserving projection (OLPP) and support vector data description (SVDD) is proposed for the performance degradation assessment of rolling bearings. Firstly, the extracted features are selected according to the correlation with root mean square of vibration signals. Secondly, dimension reduction is performed on the selected features with OLPP and principal components are obtained. In the following, the performance degradation indicator (PDI) is constructed with the principal components fused by SVDD. Finally, the proposed approach is verified with the lifecycle data of two rolling bearings. The results show that the PDI has good monotonicity and is sensitive to the initial fault. Besides, the noise in the PDI could be efficiently suppressed.