Circular Image

Daan ter Meulen

info

Please Note

1 records found

Benchmarking Hybrid OMA method against a high-fidelity numerical model

Master thesis (2026) - A. Tekin, Daan ter Meulen, A. Antonini, A. Cabboi
Operational modal analysis (OMA) identifies the modal properties of a structure from its response alone, mak- ing it attractive for structural health monitoring of offshore wind turbines (OWTs). However, its application to operating turbines is complicated due to closely-spaced modes and excitation that violates the classical OMA assumptions like narrow-band wave loading and rotor harmonics. A Hybrid OMA framework that combines peak detection with harmonic-discrimination tests have been proposed to address this, but have been bench- marked only on simplified models with limited structural physics and weak higher-order harmonics. Hence, this paper develops an enhanced numerical model of the NREL 5 MW offshore wind turbine on a monopile foundation. The model incorporates soil–structure interaction, hydrodynamic added mass and damping, flexi- ble blades, and physically derived rotor harmonics, while allowing each excitation mechanism to be activated independently. The resulting high-fidelity model is used to benchmark the Hybrid OMA framework under re- alistic operational conditions. In the updated identification framework, Bayesian Operational Modal Analysis (BAYOMA) is incorporated to resolve closely spaced modal pairs and quantify parameter uncertainty, whereas the Power Spectral Density Transmissibility (PSDT) step is omitted based on the findings of the benchmark presented in the original Hybrid OMA study. The results show that the combined hybrid OMA and BAYOMA framework identifies the structural modes accurately under operational excitation, including the second tower- bending and blade modes. Moreover, this research also overcomes the main limitation of the hybrid OMA method related to the wave-induced peak by discriminating it from the structural modes by its high apparent damping and its sea-state-dependent frequency. The kurtosis-based harmonic discrimination, however, is found to remain reliable only for the dominant 1P harmonic and to degrade for higher-order harmonics under realistic operational excitation ...