WINDTWIN: Condition monitoring of wind turbine gearbox toward digital twin ecosystem

点击翻译按钮获取中文摘要

The wind turbine gearbox (WTG) unusually operates in harsh working environments, making the WTG prone to failures and resulting in unexpected shutdowns and enormous economic loss. Therefore, it is critical to conduct condition monitoring and predictive maintenance for the safe and efficient operation of wind turbines. This project aims to develop a digital twin ecosystem for the health management of wind turbine gearboxes. More specifically, an intelligent modeling calibration algorithm is developed for the high-fidelity model establishment of WTG. Also, novel fatigue models are developed for simulating the degradation characteristics of WTG. To achieve the seamless convergence of the physical structures of WTG with its virtual model, first, novel health indicators are developed for estimating the WTG degradation progression; second, the Bayesian inference is improved to include uncertainties existing in measurements and models for bridging the physical structures and virtual models. The novel health indicators and improved Bayesian inference can help update the virtual model in real-manner, ensuring the degradation behaviors of WTG can be well revealed and reflected by the virtual models. Moreover, novel transfer learning algorithms are developed to minimize the discrepancy between the virtual models and individual wind turbines and extend the capability of the developed WTG digital twin ecosystem. With the developed digital twin WTG health management ecosystem, safe and reliable operations of wind turbines can be realized, and the maintenance cost and downtime of the wind turbine are expected to be reduced significantly (around 35% and 75%, respectively); also, the productivity of wind turbines could increase by 30%. The unique research approach of this fellowship will be hosted by Prof. Asoke Nandi from Brunel University London and co-supervised by secondment supervisor Prof. Daniele Dini from Imperial College London.