Development and Analysis of a Full Vehicle CFD Dataset for Rim Aerodynamics

Abstract

Rim geometry affects vehicle forces and flow around the wheels, but automotive computational fluid dynamics (CFD) datasets cover few rim designs. This thesis develops and analyses DrivAerRim to examine these effects and provide data for future geometric deep learning surrogates. The dataset contains full vehicle simulations for all 904 three dimensional Deep- Wheel rims on a common DrivAer estateback. Vehicle geometry and numerical settings remain fixed, with zero yaw and a speed of 140 km h−1. The simulations use steady Reynolds averaged Navier–Stokes equations with the realizable k−ϵ turbulence model. Moving reference frames represent rim rotation, while rotating wall conditions are applied to the tyres. Mesh sensitivity and wind tunnel comparisons of drag differences and surface pressure support this setup, although pressure errors vary by region. The drag coefficient range over the final 500 iterations is below 0.002 in all cases. Vehicle drag spans approximately CD = 0.2334 to 0.2549, while total and axle lift also vary with rim geometry. The combined contribution of wheel assemblies, wheel supports and wheelhouses is concentrated near 29% of conventional vehicle drag. Accumulated drag shows that the largest differences develop around the front and rear wheels. Associations between wheel mass flow and local drag differ by component and axle. Selected high drag cases generally show broader total pressure deficits near the ground and altered connections between wheel wakes and the downstream wake. The exported STL geometries, force coefficients, wheel region surface fields, global slices and local volumes are organised for open reuse. They support further aerodynamic analysis and surrogate training under the specified steady simulation conditions. Surrogate development and evaluation remain outside the scope of this thesis.