Monday  |  Salon 9  |  02:50 PM–03:10 PM

#18526–A Hybrid Physics-based and Data-Driven Approach for Digital Twinning of an Offshore Wind Turbine

This study presents the development of a hybrid digital twin of an offshore wind turbine, which combines both physics-based and data-driven models. The turbine is instrumented with accelerometers and strain gauges that collect its vibration response continuously. The main objective is to model the complex dynamics resulting from varying environmental and operational conditions. Specifically, the model aims to account for the aerodynamic stiffening caused by gyroscopic effects, as well as the aerodynamic damping at high wind and rotor speed values. These effects will be accounted for considering an additional stiffness and damper at the hub height in the digital twin of the structure. A long-short term memory (LSTM) model will be used to predict the stiffness and damper properties according to the operational information (wind speed, rotor speed, power generation, yaw and pitch angle) collected in the monitoring Supervisory Control and Data Acquisition (SCADA) system data. The predicted values will be compared from the prior estimation of the stiffness and damper obtained via a finite element model updating approach, and backpropagate through the LSTM network to adjust its weights. The final optimal prediction of the stiffness and damper at the hub will be used to correct the digital twin of the wind turbine and reliably predict the dynamic response of the turbine.

Burak Bagirgan   Tufts University
Eleonora Tronci   Northeastern University
Babak Moaveni   Tufts University
Eric Hines   Tufts University