A turbulent CFD simulation based on the Large Eddy Simulation approach (LES) is under development at Autodesk. The code is being designed to provide designers with rapid estimates of drag coefficient early in the product development phase. The efficacy of this approach depends on the solver's ability to predict drag with reasonable accuracy. To this end a series of wind tunnel experiments were performed to measure drag on a model automotive body and validate the computational results.
An automobile CAD model was transformed into a physical object by 3D printing at one fifth scale. Wind tunnel testing included drag coefficient variation with wind speed, yaw angle, wake maps of velocity and turbulence intensity, and high speed photography of smoke and helium bubble flow over the automobile surface and rear.
A CFD simulation incorporating an integrated rapid voxel mesher was used to simulate each experiment. Simulation uncertainty arising from voxel size and simulated tunnel dimensions was established in detail. In conjunction with experimental uncertainty an estimate of the quality of validation was obtained.
Simulated drag coefficient was found to be within the combined uncertainty of experiment and simulation over all but the lowest velocities when the automotive body was aligned directly into the wind. Yaw angle measurements showed a clear asymmetry in measured drag coefficient which the solver could not replicate since the CAD model was completely symmetric. The source of experimental asymmetry is a topic of ongoing investigation.