Towards Improved Scale-Resolving Modeling and Simulations of Turbulent Flows

Doctoral thesis, 2022

Scale-resolving simulations are viewed as powerful means for predicting complex turbulentflows, as often encountered in aeronautical applications. However, since turbulent scalesspan over a considerable range from the smallest Kolmogorov scales to the largest ofequivalence to configuration size, scale-resolving computations are often demandingon computational resources and, furthermore, on the underlying numerical methodsused in the simulations. Nonetheless, hybrid RANS (Reynolds-Averaged Navier-Stokes)-LES (Large-Eddy Simulation) techniques are considered computationally accurate andaffordable for aeronautical industry applications. This thesis explores and develops numerical methods suitable for hybrid RANS-LES. These methods are implemented in the Computational Fluid Dynamics (CFD) solverM-Edge. A low-dissipative, low-dispersive numerical scheme was analyzed and verified in subsonicLES of turbulent channel flow and Decaying Isotropic Turbulent (DIT). It was shown thatnumerical dissipation and dispersion needs to be carefully tuned, in order to accuratelypredict resolved turbulent stresses and the correct decay of turbulent kinetic energy. The reported results are in good agreement with reference DNS and experimental data. The numerical scheme was further adapted and analyzed for compressible flow, where good agreement with reference DNS and experimental data is achieved for hybrid RANS-LESof supersonic turbulent channel flow and supersonic baseflow. The optimized numerical scheme was then examined in hybrid RANS-LES computations of developing turbulent channel flow. In order to mitigate the grey area the LES zone, synthetic turbulence was applied at the RANS-LES interface using the Synthetic Eddy Method (SEM) and the Synthetic Turbulence Generator (STG). It was shown that using upstream turbulent statistics from a precursor LES or RANS, the recovery length of the skin friction coefficient can be reduced with improved mitigation of the grey area. A new implicit gradient reconstruction scheme was developed, which is suitable for node-centered solvers. It was shown that the reconstruction scheme achieves fourth-order scaling on highly irregular anisotropic grids for an analytical academic case. The Navier-Stokes Characteristic Boundary Condition (NSCBC) was implementedand verified for transport of an analytical vortex. It was shown that special boundarytreatment is needed for transporting turbulent structures through the boundary with minimal reflections.