An Implicit Hybrid Turbulence Model for Wall-Bounded Turbulent Aerodynamics

Abstract

Resolving the small-scale streaks in viscous boundary layers require excessively refined, isotropic grids beyond the capability of all but the most powerful supercomputing facilities. This thesis contributes to the widespread effort in developing hybrid turbulence closures, which aim to reduce near-wall resolution requirements yet maintain turbulence-resolving capability in free-shear flow regions. This thesis evaluated the performance of an Implicit Large-Eddy Simulation (ILES) approach implemented within a very high-order accurate framework for a structured finite-volume, compressible solver. A new hybrid RANS-ILES model was proposed and systematically developed using hybrid length-scale modifications, a blending function, a boundary layer detection variable and a smoothing function. Different formulations for each of these mechanisms were numerically assessed, classified and eliminated to form the most optimal algorithm. This avoids some limitations of eddy-viscosity based subgrid-scale models used in conventional Large-Eddy Simulation and reduces case-specific calibration. The RANS-ILES hybrid model was used to simulate turbulent boundary layers highlighting its robust performance even on grids designed to induce Modelled-Stress Depletion. Under-resolved ILES results were degraded compared to URANS for such grids. Two Reynolds number regimes were used for flow around a cylinder and force coefficients, separation angles, wake profiles and shedding frequencies agreed well with past hybrid models, LES and experiments. The auxiliary transport equation successfully identified turbulent and non-turbulent regions in two automotive test cases. Surface pressures agreed well with experiments and a commercial LBM solver, despite some modelling simplifications. These simulations also leveraged high-order numerics and accurately predicted dominant spectral modes and coherent structures.