Investigation of wake loss generation in the wake of a linear low pressure turbine cascade using RANS and LES methods

The development of turbomachinery components for aircraft engines and power generation relies heavily on the usage of computational fluid dynamics (CFD). Due to the high number of simulations required in design optimizations, the Reynolds- averaged Navier-Stokes (RANS) approach with turbulence modeling will be the industry standard for many years to come as it offers fast turnaround times. However, the semi-empirical turbulence models were developed and calibrated for rather simple canonical flows, whereas the flow in turbomachines, particularly in off-design operation, is complex. In the past decades, several modifications of the linear eddy viscosity models have been developed to address the shortcomings. Thus, the predictive quality of RANS depends on the experience of the user to select an appropriate model combination. On the other hand, scale resolving methods like Large Eddy Simulation (LES) and Direct Numerical Simulation (DNS) are presently limited to small Reynolds numbers in an industrial context, as they require a high computational effort to (partially) resolve all turbulent spatial and temporal scales. However, in research context, LES and DNS simulations are feasible for selected configurations and operating points. The LES and DNS results of these cases can assess the validity of lower order turbulence models and contribute to the model development. Within the scope of this thesis, the mixing process of a turbulent wake flow of a low pressure turbine blade is investigated using at least two different Reynolds numbers. The simulations are performed with RANS turbulence models and compared to the results of LES simulations of the same geometry. This works aims to identify deficits of the turbulence model formulations and to highlight approaches to improve the investigated turbulence models. Starting with a literature review, the flow features of a turbulent wake flow are identified and described. Next, the numerical setup for RANS simulations is derived. The already existing LES results are post processed to derive suitable boundary conditions for the RANS simulations. A grid study is performed to ensure grid independent results. The results of the simulations are analyzed and compared to the available LES data. Finally, the results are documented.