Computational analysis of transitional and massively separated flows with application to wind turbines

Abstract

Laminar to turbulent transition and higher delity turbulence modeling have been implemented in the parallelized, unstructured mesh, compressible Navier-Stokes solver MaPFlow. In the case of laminar to turbulent transition, both boundary layer and transport equation approaches were assessed. These include the eN method and the transport equation models -Re , and AFT. All approaches were validated against a wide range of cases concerning airfoils, wings/blades and generic fuselage, stemming from both the wind energy and the aeronautics sectors. The focus was on integrated loads and transition locations. In the context of two-dimensional simulations, the boundary layer eN method and the AFT transport equation model exhibit better performance than the other alternatives. The -Re model is also a viable option if Reynolds numbers not higher than 6 x 106 are considered. For Reynolds numbers higher than this limit, the accuracy of the model was found to deteriorate. However, both the eN method and the AFT model cannot be used to predict cross ow transition in three-dimensional simulations. In such scenarios, the -Re model can give accurate results, provided that the Reynolds numbers fall within the aforementioned limit. In the case of higher delity turbulence modeling, both Large Eddy Simulation (LES) and Detached Eddy Simulation (DES) approaches were implemented. LES utilized the Smagorinsky subgrid model. Regarding DES, both Delayed DES (DDES) and Improved Delayed DES (IDDES) variants were considered. The focus was on ow cases with massive separation. Both LES and DES provided more accurate results than the baseline Unsteady Reynolds Averaged Navier Stokes (URANS) simulations when compared to experiments and reference results. Neither LES nor DES were pushed to their limits. DES is considered computationally less demanding, due to wall modeling inside the boundary layer region. Therefore, it is a more viable option than LES for industrial applications. However, due to wall modeling, DES is not expected to perform well in ows where the presence and development of small turbulent scales inside the boundary layer are important. In those cases, LES using ne meshes should be considered.