Fleet-based model updating for design optimization and structural health monitoring (SHM)
To date, we see a discrepancy between modal properties as calculated during the design stage, and the modal properties which are identified from built offshore wind turbines. With this research project we want to bridge the gap between characteristics obtained from existing integrated finite element models and measured characteristics from operational turbines, by developing a fleet-based model updating strategy.
Being able to calibrate finite element models such they effectively represent built turbines is essential for e.g. model based response extrapolation techniques which allow to derive virtual measurements at arbitrary locations on the structure, based on a limited set of response measurements from easily accessible locations of the structure.
In the initial stage of this project the goal was to clearly delineate and explore the feasibility as well as technical challenges which come with the goal of performing a fleet-based / fleet-wide model updating. Therefore an extensive sensitivity study has been performed on 50 turbine specific integrated FE models which resemble all turbines located within an offshore wind farm in the Belgian north sea. The results obtained from this study have been compared to identified natural frequencies on the as-built counterparts, resulting in a inventory of discrepancies between the first two fore-aft and side-side frequencies for a complete wind farm.
Based on this study, a research strategy will be formulated with the aim to investigate the feasibility to use (long-term) resonance frequency measurements to (continuously) update a site-specific dynamic model of an individual wind turbine. The second step is to exploit the availability of response measurements at different sites as well as the similarities between structural and environmental aspects at these sites in order to improve the conditioning of the inverse updating problem. The final objective is to simultaneously update dynamic models of an entire fleet.
The results of this research serve to:
1. Identifying modelling uncertainties related to e.g. the soil stiffness and/or soil-structure-interaction on a farm wide level. Using the optimized soil profiles a comparison can be made with design, allowing to create valuable insights on assumptions and current methods.
2. Accurate finite element models will allow to improve existing fatigue monitoring techniques. To this end, model based response extrapolation techniques can be employed, which allow to translate a limited set of response measurements into response estimates at arbitrary locations on the structure.