A Wind Turbine Bearing Performance Evaluation Method Based on Similarity Analysis of Fuzzy k-Principal Curves in Manifold Space

Condition monitoring (CM) is widely used in wind turbines (WTs) to reduce operation and maintenance (O&M) costs. Bearings are crucial components in WT and many bearing CM approaches have focused on vibration analysis. Statistical theory and artificial intelligence-based WT bearings evaluation methods require mass data for training, which makes the detection of incipient failures barely possible. In this paper, a WT bearing performance evaluation method is proposed based on the similarity analysis of fuzzy k-principal curves (FKPCs) in manifold space. For a start, 38 features are extracted from bearing vibration signals to constitute high-dimensional feature matrices. The feature matrices for the healthy samples and the samples to be evaluated are then transformed into 3-D space. Afterward, the FKPCs are extracted and the similarities among the curves of samples are calculated based on the Hausdorff distance to evaluate the performance of the bearing. Bearing degradation experiments are investigated to verify the efficiency of the proposed method. The results indicate that the proposed FKPC method can portray the degradation trend of bearings accurately with the capability of detecting incipient failures. The proposed method can be applied in the case of small-size training samples with stable performance.