<![CDATA[Aerodynamic prediction for full scale passenger vehicles relies on the use of mesh resolutions which accurately represent boundary layer evolution and wake dynamics while maintaining reasonable computational expense. To verify the drag prediction for two production-derived vehicle geometries (Notchback and Hatchback) simulated at 15° steady state crosswind using incompressible RANS with SST k−ω turbulence models, the verification process consisted of a systematic set of five progressively refined polyhedral meshes (1.5 million cells - 7.2 million cells) created using a controlled refinement template to maintain consistent near-wall treatment within all five meshes. The drag results showed significant improvement from the coarsest mesh to the finest mesh (≈ 14% improvement for Notchback ≈ 12% improvement for Hatchback) and then clearly exhibited asymptotic results as evidenced by the difference between M4 and M5 decreasing to less than approximately 1.5%, indicating that M4 provides mesh-independent accuracy with over 20% less computational cost than M5. Furthermore, a Reynolds number sweep across the range of representative full-scale Reynolds number values demonstrated that drag is effectively insensitive to Reynolds number once the fully turbulent regime is reached and wake structures between the Notchback and Hatchback. Through this analysis it has been determined that targeted refinement strategies around A-pillar and rear-end separation zones and the near wake will provide the greatest accuracy and cost-effective use of computational resources as compared to uniform global densification, thus providing a validated mesh resolution strategy for using RANS simulations to predict drag for full scale passenger cars under steady state conditions.]]>
