Numerical studies towards adaptive large-eddy simulation

Large-eddy simulation developments and validations are presented for an improved simulation of turbulent internal flows. Numerical methods are proposed according to two competing criteria: numerical qualities (precision and spectral characteristics) and adaptivity to complex configurations. First, methods are tested on academic test-cases, in order to abridge with fundamental studies. Consistent results are obtained using adaptive finite volume method, with higher order advection fluxes, implicit grid filtering and ”low-cost” shear-improved Smagorinsky model. This analysis particularly focuses on mean flow, fluctuations, two-point correlations and spectra. Moreover, it is shown that exponential averaging is a promising tool for LES implementation in complex geometry with deterministic unsteadyness. Finally, adaptivity of the method is demonstrated by application to a configuration representative of blade-tip clearance flow in a turbomachine