Health assessment of wind turbine bearings progressive degradation based on unsupervised machine learning

Abstract

High-speed shaft bearing (HSSB) failures are exorbitant since they lead electrical energy generation to halt suddenly. In order to identify the health condition of the wind turbine and preserve the sustainability of energy production, a nonlinear vibration-based monitoring technique based on kernel principal component analysis (KPCA) has been developed. After extracting degradation characteristics from the time, frequency, and time-frequency domains. The most sensitive features are then fused using KPCA to capture the monitored bearing’s operating conditions; this method demonstrated its efficiency in dealing with the nonlinearity of the system. To detect flaws in HSSB and assess whether it is healthy, degraded, or broken, [Formula: see text], and SPE charts have been used. Real run-to-failure data from a wind turbine HSSB is used to validate the proposed technique. The suggested strategy caught the nonlinear relationship in the process variables more successfully than existing techniques, including linear PCA, and demonstrated enhanced process monitoring performance.