Presentations

Abstract

Accurate remaining useful life (RUL) assessment of a wind turbine driveline is critical at two distinct lifecycle stages: during active operations and at the end of life for lifetime extension. For active assets, precise RUL metrics enable operators to shift from reactive to predictive maintenance, optimizing logistics, and minimizing downtime. As assets approach their design life, these estimates become the financial backbone for decisions regarding lifetime extension or residual value assessment for decommissioning and resale. Precise measurements of the load are crucial in this decision making since 20% more driveline load can result in the 50% less running life. High-frequency robust fiber-optical sensors provide precise measurements and give new operation loading insights.  In our previous work, we proposed a method to determine various RUL metrics of a wind turbine driveline and its components at any point during its operational life based on the torque and rotational speed that the main shaft was exposed. In this work, we extend the method in two ways by incorporating high-frequency fiber-optic measurement data. Firstly, we account for the load distribution between the planets of the main shaft of the gearbox, which is expressed by the load sharing factor. An uneven load distribution causes higher loads on specific components in the gearbox and results in a lower overall RUL. Secondly, we quantify the discrepancy between RUL estimates derived from standard SCADA data versus high-frequency fiber-optic measurements.  To illustrate the impact of the load distribution, we use data from a wind turbine from the Wageningen Universiteit. The torque and speed are available through historical SCADA data of the gearbox. The load sharing factor is determined using fiber-optical sensors measuring the strain on the ring gear of the main shaft. We compare the results of the RUL estimates using three different load distribution factors: the measured factor using fiber optical sensors, equal load distribution, and a conventional safety value. The results show that the RUL of the wind turbine driveline components is sensitive to the load distribution of the planetary stage. Furthermore, using the safety value leads to unrealistic degradation that would trigger premature and unnecessary intervention. Conversely, using measured data provided a more granular and realistic view of component health. Finally, our comparison shows an accuracy loss in using SCADA data compared to high-frequency data.  We conclude that integrating load sharing factor and high-frequency load measurements into remaining useful life calculations is necessary for accurate failure risk assessment. By moving away from conservative safety values toward measured, data-driven strategies, operators can implement predictive maintenance strategies and maximize the operational life of the drivetrain. This work should help fleet operators to accurately determine the failure risk of their assets, plan maintenance accordingly and reduce O&M costs and extend turbine lifetime by 10-20%.