Numerical investigation of a quarter scale DrivAer model based on Detached-Eddy Simulation

In this work , the validation of a quarter scale DrivAer model based on Detached-Eddy Simulation (DES) is proposed. The study has been conducted through the open-source software OpenFOAM, adopting a transient solver based on the Navier-Stokes equations. The Delayed Detached-Eddy Simulation (DDES), variation of pure DES, has been implemented. The turbulence model adopted is the Spalart-Allmaras, based on a single transport equation. First, half computational domain has been generated. The longitudinal symmetry plane has been used in order to split the domain into two symmetric halves. In so doing, the simulation time has been reduced. The flow has been considered as incompressible, because of low Mach numbers involved. All simulations have been performed at a velocity of 45 m/s and at Re 3.2x10^6. The High Reynolds approach has been adopted. A comprehensive analysis of the main mesh parameters is presented. Their influence on computational cost and mesh quality has been investigated. The aim is to discretize the boundary layer by means of prism layers. In the following, a study about the boundary layer on the ground related to the wind tunnel empty section is presented. The evolution of the ratio between the average and the free stream velocity and of the total pressure coefficient is investigated. The CFD results are compared with the experimental data obtained in the model scale wind tunnel of the FKFS Institute of Stuttgart. Different computational domains and boundary conditions have been implemented. The solution with a progressive decrease of the distance between the inlet of the computational domain and the DrivAer is reported. The results were considered acceptable to start the validation process of the DrivAer. To prove the robustness of the model in consequence of an important change of geometry, two different configurations have been studied: the notchback and estateback with detailed underbody. According to the convergence of the residuals and the stability of the main aerodynamic coefficients over time, the simulations have been run for 2 seconds and results have been averaged for the last 0.5 seconds. For the investigation, three mesh approaches have been defined: coarse, medium, and fine. Being the mesh a known source of error, its convergence is pursued by increasing the level of refinement. The fine approach proves to be closer to the experimental results, being able to better discretize the surface of the DrivAer. In order to correctly model the boundary layer, a study adopting a number of prism layers between 3 and 6 has been performed. In conclusion, in order to estimate how the symmetry approximation adopted influences the results, the half model baseline and the corresponding full model have been investigated.