Assessment of Wall Modelling for Large Eddy Simulations of Turbomachinery

The design process of turbomachinery components relies heavily on Reynods Averaged Navier Stokes (RANS) simulations. This approach is well suited for steady simulations and comes with a low computational cost. However, turbomachinery flows are complex and difficult to predict accurately with RANS computations. Large Eddy Simulations (LES), capable of resolving the larger scales of turbulence, are a promising way to improve the predictive capability of numerical simulations. The main drawback of LES for wall bounded flows is its high computational cost, scaling with Re1.86 [1]. Turbomachinery components are characterized by Re ≈ 105–6, implying simulations with several billions of cells, with most allocated to resolve the turbulent scales inside the boundary layers. A potential cost-reducing approach is to introduce wall modelling. However, several questions remain, notably the wall model interaction with the laminar-to-turbulent transition and the impact of grid resolution. To clarify these points we investigate the flow across a linear compressor cascade with Wall Resolved LES (WRLES) and Wall Modelled LES (WMLES) simulations. Various near-wall resolutions are tested at on and off-design conditions to characterize the impact of the wall model on the flow field and the aerodynamic losses. RANS simulations complement the analysis. The results indicate that the WRLES agree the closest with experimental measurements. WMLES with relatively high near-wall resolution capture most of the flow physics while allowing a significant speed-up. However, reducing the resolution further leads to unphysical flow separations, despite staying well in the range of wall model validity.