How close is the virtual tunnel?

Sphere drag vs. Reynolds number

Direct LES regime, compared with the Schiller–Naumann correlation.

What the numbers mean

• Direct regime (Reynolds number resolved by the grid): within a few percent of reference data, converging with resolution.
• Bluff and sharp-edged bodies at flight Reynolds numbers (drones, pods, rockets with fins, blunt bases): within about 10–15%.
• Cars (the Ahmed body on a rolling road, 25° and 35° rear slant): with Refine near the body (the nested zoom grid) within 6–14% of the wind-tunnel measurement; on the standard grid alone about 2× too high. Use the zoom grid for car drag.
• Highly streamlined bodies at high Reynolds number: still over-predicted, about 2.5× (it was 3.5× before the wall model). The grid carries a capped Reynolds number, so the boundary layer above the first cell stays thicker than in flight and the pressure does not recover at the tail. Trends and relative improvements remain meaningful; absolute drag is conservative.
• Lift on thin wings in the CFD is not validated yet: the wing case comes out 50–66% low at every fidelity, for the same boundary-layer reason. Lift, L/D and pitching moment from the CFD are shown as indicative only, and lift-based CFD goals are offered as indicative only.
• Wing model (what the optimizer uses for wings, tails and NACA sections): the vortex lattice gives exactly the elliptic-wing span efficiency, the circular-wing lift slope within 2% of the exact solution and NACA camber within 0.1°; NeuralFoil section lift and drag are within 7% of NASA's NACA 0012 wind-tunnel data. No low-Reynolds-number (RC-size) section data is in the suite yet: there, section drag rests on NeuralFoil's published agreement with XFoil.
• RC propellers: blade-element thrust and power within 8% of UIUC wind-tunnel data for two blade families (each fitted on one propeller, checked on others), including the same blade with 2, 3 and 4 blades; the actuator-disk slipstream matches momentum theory within 1%.
• Compressibility (General tunnel): above Mach 0.3 the surface pressures are corrected with the standard Kármán–Tsien rule up to each body's critical Mach number; above it, to Mach 1.2, the drag is a transonic estimate. Neither band has its own validation case yet.
• Supersonic drag, lift and heating: sphere drag within 5%, cone wave drag within 0.2% of exact theory, ogive nose wave drag within 7% of Bonney's data (second-order shock-expansion along each meridian), thin-plate lift within 2%, stagnation heating within 7% of the Sutton–Graves correlation. Normal force on slender bodies is over-predicted at low supersonic Mach (+38% at Mach 2, +10% at Mach 4).
• Each optimization compares designs on the same grid with the same settings, so systematic error largely cancels in the reported improvement. The random part is measured on every job (the original re-run shifted by part of a grid cell), and a change is only accepted when it beats that error.