Long-term validation of a model-based virtual sensing method for fatigue monitoring of offshore wind turbine support structures: Comparing as-designed with state-of-the-art foundation models

Abstract

Model-based virtual sensing offers a viable approach for monitoring fatigue loads on operational offshore wind turbines. These methods combine response measurements with first-principle, or data-informed models, to estimate load time series at hard-to-access locations. However, their accuracy depends on the fidelity of the underlying model, which is largely influenced by uncertainties in the soil–structure interaction (SSI) models.

This study evaluates the impact of different SSI modelling approaches in terms of a virtual sensing validation study targeting strain estimation above and below the mudline of a bottom-founded offshore wind turbine. To this end, different numerical models derived from, and validated against, design documentation serve as input to a dual-band modal decomposition and expansion (MDE) method. The considered SSI models range from an API/DNV-based foundation model to a PISA-based model including scour protection. Virtual sensing results are generated for two-year equivalent datasets, obtained for three operational offshore wind turbines, each equipped with extensive load monitoring systems. One turbine is used to assess the effect of the model updates, while two additional turbines are used to assess the across-site consistency. The estimated strains are directly compared against available strain validation data, in terms of damage-equivalent stress, and are accumulated to give a single comparative metric representative for the two-year periods.

Results show that PISA-based soil reaction curves significantly improve agreement with measured strains while adding a scour protection model has a relatively smaller impact. These findings highlight the importance of accurate foundation modelling in virtual sensing and demonstrate the feasibility of fatigue monitoring at hard-to-access locations.