EulFS : A Parallel CFD Code for the Simulation of Euler and Navier-Stokes Problems on Unstructured Grids

We present results with a parallel CFD code that computes steady-state solutions of the Reynolds-Favre averaged Navier-Stokes equations for the simulation of the turbulent motion of compressible and incompressible Newtonian fluids. We report on preliminary experiments on 2D and 3D problems, for both internal and external flow configurations. In this communication we present numerical results with an academic code developed by the first author [3] for simulating the turbulent motion of compressible and incompressible Newtonian fluids on 2D and 3D problems. The dynamic of the fluid is modeled using the Reynolds-Favre averaged Navier-Stokes (RANS) equations. Despite the non-negligible degree of empiricism introduced by turbulence modeling, it is recognized that the solution of the RANS equations still remains the only feasible approach to perform computationally affordable simulations of problems of engineering interest on a routine basis. The code is able to compute steady-state solutions of the RANS equations for both internal [1] and external [5] flow configurations. The computational domain is tesselated using unstructured grids made of triangles and tetrahedra, in 2 and 3 space dimensions, respectively. The integral, conservation-law form of the governing equations is discretized using Fluctuation Splitting (or Residual Distribution) schemes. This discretization technique was introduced in the early eighties by Roe [4] and shares common features with both Finite Element (FE) and Finite Volume (FV) methods. It features linear shape functions and compact stencils that result in more sparse matrices arising from the discretization. Turbulence is modeled using the Boussinesq approximation, and the eddy viscosity is computed by means of the one-equation Spalart and Allmaras turbulence model. Although the objective here is to calculate steady state solutions, the timederivative in the governing conservation equations is retained. As explained below, this is done because the integration strategy is partially based on pseudotime marching. This is achieved by discretizing the time-derivative with an