Posters

Abstract

We present a method to model and predict failures in both the wind turbine gearbox and its individual components throughout the turbine's entire lifespan. Our model considers different interconnected load-bearing parts of the gearbox like gears, bearings and axles by using industry-standard probability distributions which are dynamically modified depending on external parameters. We are able to consider various maintenance strategies and sensorics. This allows us to compare different scenarios and assess their monetary benefits as well as specific failure modes over the lifetime of a wind turbine or even a whole wind park. To ensure reliable data, the model is run a high number of times, and the resulting data is aggregated using state-of-the-art statistical methods. Our approach is highly parallelizable and can efficiently utilize cloud computing with a high number of processors and cores. The model was mainly created to have reliable data to back up our newly developed sensor detecting wear particles in gearboxes and as such this sensor can be fully integrated into the model. However, we believe that the model also can be very useful beyond its original intention by producing reliable statistical data of possible gearbox failures over the lifetime of a wind turbine.