摘要：This paper presents advances in the development of pseudo-DNS, a new turbulence model basedon pre- computed inertial stresses and a concurrent multiscale approach. The model relies on solvinga “coarse” subproblem by computing, on a relatively coarse grid, the flow of a pseudofluid, using adatabase of sub-grid inertial stresses precomputed in separate “fine” subproblem simulations. Theseinertial stresses are assumed to depend only on a dimensionless tensor, namely Id, related to the kine-matic viscosity of the real fluid and the local velocity gradient and local coarse mesh size. Databasesare assembled using DNS simulations of a set of fine problems in hexahedral domains with structuredmeshes, using “jump” boundary conditions to impose the desired velocity gradients. Each fine subprob-lem is simulated until a statistically steady state is achieved. The results are post-processed to obtaintime-mean inertial stresses averaged in a subdomain, called Representative Volume Element or RVE.The RVE simulation results are stored in the database. A distinction is made between a “boundary layerRVE” (or “wall RVE”, characterized by having one of its faces in contact with a solid wall), and an“internal RVE” otherwise. Pseudo-DNS requires a database to be built for each one of these two types.For the wall RVE, fortunately, the tensorial Id may be reduced to a scalar one. The present work focuseson the construction of the wall RVE database, using DNS simulations of incompressible turbulent flowbetween parallel moving plates with null imposed pressure gradient. We present a mesh convergenceanalysis and the resulting wall RVE database for a range of Id numbers from 0 to 30000.
