A DNS/URANS approach for simulating rough-wall turbulent flows

Abstract

A novel hybrid method combining direct numerical simulation (DNS) and the Reynolds-averaged Navier Stokes (RANS), denoted as a stress-blended method (SBM), has been developed. The SBM is targeted at simulating turbulent flows over arbitrary rough surfaces in which computational savings can be achieved by making the DNS domain as small as possible. Within the SBM framework, a RANS model is enforced above the roughness layer to prevent the momentum build-up which arises in simulations where the computational domain is too small to represent the largest eddies. The SBM is validated for turbulent channel flow, both for smooth wall turbulence and using a parametric forcing approach to mimic roughness effects, with a computational cost that scales linearly with Reτ. The method is then applied to selected subsets of a scanned grit-blasted surface. For the same subset, the roughness function is found to be within 1% of available DNS. Comparisons of small and large subsets showed differences of over a factor of two in equivalent sand grain roughness, indicating the importance of choosing representative surface samples. Simulations in the fully rough regime are carried out using one to two orders of magnitude fewer points than in a typical DNS. Since no assumptions on the roughness properties or the flow structure (such as outer layer similarity) are made, we expect the SBM to be applicable to non-equilibrium turbulent boundary layer flows.