Gaspar García, CFD portfolio
Windsor squareback: an extrapolation a fourth grid did not support
CaveatedUpdated 2026-11-14
Objective. Measure the drag of the Windsor squareback (no wheels, 0° yaw) with steady RANS and establish how far it is from the wind-tunnel value once grid error is accounted for.
| Free-stream speed | 40 m/s |
|---|---|
| Reference length | 1.044 m |
| Reference area | 0.056 m² (half model) |
| Reynolds number | Not stated in the cited files |
| Fluid | Air, incompressible (Newtonian transport model) |
| Solver | OpenFOAM v2412, simpleFoam, steady |
| Turbulence model | k-omega SST |
| Wall treatment | Wall functions |
Comparison with experiment
| Quantity | Experiment | Simulation | Signed difference |
|---|---|---|---|
| Drag coefficient Cd, finest grid | 0.3070 | 0.2856 | −7.0 % |
| Lift coefficient Cl, finest grid | −0.030 | −0.142 | sign-magnitude failure |
| Drag, three-grid extrapolation (withdrawn) | 0.3070 | 0.2765 | −9.9 % |
Cd against Cells (millions)
- Simulation (simpleFoam, k-omega SST)
- Experiment
Verdict
On the finest grid steady k-omega SST under-predicts drag by −7.0 %. A fourth grid showed that the earlier three-grid extrapolation was not supported, so no grid-independent value is claimed.
Limits
- No positive observed order on four grids: no numerical uncertainty band.
- The tunnel's reference dynamic pressure convention is unresolved; the error's size depends on it.
- Lift is a sign-magnitude failure (−0.142 against −0.030).