Comprehensive Failure Analysis and Remaining Useful Life Prediction of Main Wind Turbine Bearings: A Multi-Parametric Approach

This paper presents a multi-parametric approach for the failure analysis and remaining useful life (RUL) prediction of main wind turbine bearings, with a particular focus on in-service grease condition as a diagnostic variable. Wind turbine bearings operate under severe and variable loads, making them highly vulnerable to wear, contamination, lubrication degradation, and premature failure. Given the high cost and logistical complexity of wind farm maintenance, especially in offshore environments, the study emphasizes the value of condition-based and predictive maintenance strategies to improve reliability and reduce unplanned downtime.

The proposed methodology combines historical bearing temperature, gearbox temperature, power generation data, and grease analysis results to identify abnormal wear patterns and estimate the time remaining before a critical threshold is reached. Exploratory data analysis is used to detect trends, anomalies, and correlations, while supervised multiple regression is selected as the most effective predictive model in the initial testing phase. In addition, failure modes of bearing grease are systematically examined through failure modes and effects analysis (FMEA) and fault tree analysis (FTA), highlighting mechanisms such as oxidation, viscosity loss, contamination, moisture ingress, additive depletion, and incorrect lubrication practices.

Results from a study of nearly 260 bearings over 18 months showed that 27 bearings exhibited abnormal behavior, of which 16 later confirmed worsening trends. Visual inspection of selected cases validated the usefulness of the model for early detection. The study concludes that combining grease analysis with operational variables can significantly enhance bearing health assessment and support more effective maintenance planning.