Data Driven Assessment of Offshore Wind Turbine Foundation Stiffness-Paper

Paper

Pdf Summary

The paper presents a data-driven method for estimating the stiffness of installed offshore wind turbine monopile foundations. The framework combines vibration monitoring, modal identification and finite-element model updating, enabling foundation behaviour to be assessed under real field conditions.<br /><br />The method was applied to four days of monitoring data from a 3.6 MW turbine at the Nobelwind offshore wind farm. Acceleration measurements from three tower locations were analysed during turbine standstill. Stochastic subspace identification and a modified eigensystem realization algorithm identified natural frequencies and mode shapes. These modal properties were then used to tune foundation stiffness parameters in a one-dimensional finite-element model.<br /><br />For the first two fore-aft modes, the updated model closely matched the measured modal properties: predicted frequencies matched the identified values, and mode-shape correlation was very high. The researchers compared the inferred foundation behaviour with predictions from the newer PISA design method and the established API/DNVGL approach. PISA provided a close fit, with a modest underprediction of stiffness. API/DNVGL substantially underpredicted lateral and rotational stiffness and showed larger discrepancies in predicted frequencies and ground-level deformation.<br /><br />The results demonstrate that monitored vibration data can help validate foundation design methods, calibrate structural models for lifetime reassessment, and detect changes in foundation performance, such as those associated with scour or cyclic soil behaviour. The study is based on a single turbine and standstill data; future work will address coupled higher modes and explore site-specific refinement of PISA predictions.

Keywords

offshore wind turbine

monopile foundation

stiffness estimation

vibration monitoring

modal identification

finite-element model updating

natural frequencies

mode shapes

PISA design method

API/DNVGL