Aerodynamic force prediction of the laminar to turbulent flow transition around the front bumper of the vehicle using Dynamic-slip wall model LES

Abstract

The vehicle design process is an optimization that balances many factors including aerodynamics, exterior styling, cabin space and cost. Aerodynamic drag is of growing importance due to its impact on vehicle driving energy and ultimately fuel economy. Computational fluid dynamics (CFD) simulation can be an effective tool for predicting changes in drag due to changes in vehicle shape in the early stages of the design process. Traditional Reynolds Averaged Navier-Stokes (RANS) approaches struggle, however, to predict the impact on total drag due to design changes in certain vehicle parts such as front bumpers. The flow in these regions is characterized by stagnation points, thin transitioning boundary layers and separation. This paper explores the ability of Large-Eddy Simulation (LES) to predict changes in aerodynamic drag over a series of 6 bumper modifications to Honda's "N-ONE" compact car as a benchmark test. Two LES methodologies are compared with wind tunnel measurements and a RANS approach: (1) a wall resolved computation in ANSYS Fluent and (2) a dynamic slip wall-model (DSWM) approach in Cascade's CharLES solver. Despite fine resolution in the front bumper region, the Fluent LES was not able to capture the correct flow separation due to the cell distortion. Trade offs between cost and accuracy in the automated meshing process required for industrial design use produced non-hex cells off the surface that impacted solution quality. Automated meshing in CharLES was based on clipped Voronoi diagrams. Using the DSWM in the bumper region limited near wall resolution to y+ ~10. CharLES was successful in predicting trends in aerodynamic drag to within 2% of experiment and flow separation on the front bumper was also well predicted as indicated by pressure tap measurements. Simulation cost was approximately 60,000 core-hours per design iteration. This includes an acceleration in time to solution of approximately 20% using a geometric multigrid preconditioner to advance the simulation with a timestep double the convective CFL limit.