A comparison of deterministic and Bayesian model updating frameworks for identifying offshore wind turbine foundation parameters

• Comparison of model updating frameworks for foundation parameter identification • Numerical case studies demonstrate the identifiability of the foundation parameters • Deterministic results depend on the initial guess, motivating a Bayesian framework • Bayesian frameworks allow for parameter and model uncertainty quantification The rapid growth of the wind industry has resulted in larger wind turbines, whose modal properties lie in the lower frequency range. This has induced higher loads and stress cycles rendering accurate fatigue assessment increasingly important. Such assessment is highly affected by the precision in the estimation of turbine properties, including those related to the support conditions and foundation, which can be associated with high uncertainty. One approach to improve these estimates is to use structural monitoring data (e.g. from sensors mounted on the tower) to update the foundation parameters of offshore wind turbine models. However, the low identifiability of the parameters to be estimated can lead to divergent estimates across different fatigue estimation frameworks, which combined with the uncertainty inherent in the foundation properties, can compromise the reliable assessment of the remaining useful life. In this work, two Bayesian model updating frameworks are applied to update the foundation parameters of an offshore wind turbine and results are compared against a deterministic framework in a numerical example. The advantages and limitations of each framework are considered and the importance of accurately accounting for uncertainties as part of the model updating process is highlighted.