High fidelity aeroelastic stability analysis of complex blades in 3D flow

This work uses time marching simulations of different fidelity levels, combined with system identification, to evaluate the aeroelastic stability of operating wind turbines with modern, complex blades. It aims at incorporating higher aerodynamic fidelity by allowing the use of fluid-structure interaction (FSI) simulations to capture the aerodynamic response from complex blade geometries more correctly than with classical engineering aerodynamic models.

In the light of computationally expensive FSI simulations, an efficient setup for the simulations is investigated and found in the use of forcing sequences. They are applied as point forces to the blades, which induces the desired dynamic response of the turbine. Subspace identification, the MOESP algorithm in particular, is used to identify the system and obtain the modal parameters.

The approach is tested on the IEA 10MW reference turbine with a traditional BEM-based multibody simulator, an advanced model of the same simulator using the co-called vortex cylinder and near wake models, and finally, the FSI simulations using the Ellipsys3D CFD solver using Reynolds-Averaged Navier-Stokes (RANS) modeling with a k-ω SST turbulence model. The comparison shows similar trends in most cases, even though absolute values vary slightly. In some conditions the results are less aligned, and especially for high wind speeds and high yaw misalignments, damping values can differ significantly. It is also shown how high yaw misalignment reduces the damping in most cases. Frequencies and mode shapes tend to agree rather well across all codes.

Knowing the uncertainty bounds of modal parameters in stochastic conditions, e.g. in turbulent inflow, adds value to the results. To extract this information from only one simulation, an uncertainty quantification method for MOESP-type algorithms is derived. The approach uses the first-order Delta method to propagate covariances through the steps of the PO-MOESP algorithm. The method is validated by a Monte Carlo series of turbulent simulations and applied to a measured dataset from an Aventa AV-7 turbine. While of high accuracy and consistency, the required data length renders the approach more suitable for lower fidelity simulation tools with faster compute times than the FSI case.