Grid-convergence study of the front-wing CFD

Paper · Master's thesis, Ch. 5.1

Before any optimization result can be trusted, the underlying CFD has to be shown to be independent of the mesh. This paper documents the grid-convergence study for the front-wing simulations.

Method

Three systematically refined meshes were generated and run on the nominal geometry, with the lift and drag coefficients tracked against grid spacing. The analysis follows the standard Grid Convergence Index approach, which estimates the discretization error from the differences between successive refinement levels and the observed order of convergence.

Results

Both C_L and C_D converge as the mesh is refined, and plotting the absolute error against grid spacing on log-log axes confirms the expected order of convergence.

Mesh independence: coefficients versus grid spacing/assets/img/papers/gci-convergence.png

Fig. 1 — C_L and C_D against grid spacing across the three meshes.

Mesh quality metrics were recorded for each level alongside the computational cost, so that the chosen resolution is justified on both accuracy and affordability rather than accuracy alone.

Outcome

The study establishes the mesh resolution used throughout the rest of the campaign and bounds its discretization error, which is what allows differences between candidate designs to be attributed to geometry rather than to numerics.

This matters more in an optimization context than in a single simulation. If the discretization error were comparable to the performance differences between designs, the optimizer would be chasing numerical noise.

← See the project