Abstract Summary
This paper presents design and installation of a continuous monitoring system for an offshore wind turbine (OWT) in Rhode Island, USA. The monitoring system has been collecting continuous data from accelerometers, strain gauges, and inclinometers since April 2021 at 50 Hz sampling rate. An automated output-only operational modal analysis is developed to extract modal parameters (natural frequencies, damping ratios, and mode shapes) of the OWT every 10 minutes. The 1st bending modes in fore-aft (FA) and side-to-side (SS) directions are reliably identified under various operational and environmental conditions. The evolution of modal parameters is reported over a full year of monitoring and under different conditions. The variabilities of the first few modal parameters are found to be correlated with several ambient and operational variables such as wind speed, yaw angle, rotor speed, and power production. The identified frequencies of the SS mode are relatively constant over time and have very small variation, but those of FA mode have larger variability. Furthermore, the damping ratios of the SS mode are low and show small variability over time, but the FA mode has substantially larger damping with larger variability due to the effects of aerodynamics. A recursive Bayesian inference framework is implemented for joint input-parameter estimation and digital twinning of the turbine using output-only data. The digital twin is then used for virtual sensing and fatigue life estimation through prediction of strain time history at different hotspot locations. Strain predictions are compared with actual strain measurements under different OWT operation conditions, including idling, ramp-up, and full power generation and the predictions show to agreement with actual measurements.
