Transonic wing-body validation — external aerodynamics at cruise

Abstract

We characterise Aether's performance on the standard transonic wing-body benchmark ^{[1, 2]} at cruise conditions, comparing predicted lift and drag polars against measured wind-tunnel data. The model achieves lift-polar mean deviation under 0.6% across the angle-of-attack range, drag-polar mean deviation under 1.0%, and shock-location prediction within 1.2% of measured chord position. We give the case setup, the per-α breakdown, the decomposition of the drag residual into pressure, viscous and wave-drag components, and a discussion of what 1% drag accuracy means in cruise-fuel-burn economics.

1. Why this benchmark

The standard transonic wing-body benchmark ^{[1, 2]} is one of the workhorses of industrial CFD validation. The geometry is publicly available, the reference measurements are from a major transonic wind-tunnel facility at realistic Reynolds numbers, and the case has been run by enough groups over enough years that the published literature establishes a clear envelope of expected results. New CFD methods are routinely held to this benchmark; old methods are calibrated against it.

What makes the benchmark useful is the combination of three features. First, the geometry is realistic — a wing-body configuration at sizes and Reynolds numbers relevant to real transport aircraft. Second, the measurements are high-quality and the protocols are documented; the reference data is not contaminated by tunnel-wall corrections that are hard to reproduce. Third, the cases span enough of the cruise envelope to reveal whether a method is accurate across the range or just at a single point.

We picked this benchmark because (a) accurate drag prediction on transonic wing-bodies is the canonical industrial-CFD problem, (b) the reference data is public, and (c) every commercial-CFD method has been validated against it, so we can publish a head-to-head where the reader can verify the methodology against the open literature.

2. Setup

We used the standard wing-body geometry ^[2], standard cruise conditions (Mach = 0.75, Reynolds = 3 × 10⁶ based on the mean aerodynamic chord, angle-of-attack sweep from 0° to 2°), and a structured mesh refinement study at three densities (4 million, 8 million, and 16 million cells).

Turbulence model was the SST k-ω family, matched between Aether and a leading commercial RANS baseline. Convergence criterion was 10⁻⁵ on the residuals. Wall y⁺ at the design cruise condition was approximately 1 across the wing upper and lower surfaces; near-wall mesh refinement was the same for both tools.

Result tabulation was blinded. The aerodynamicist comparing predicted polars to measured did not know which tool had produced which polar until pass/fail had been recorded per α point — the same blinding protocol we use across our benchmark publications ^{[3, 4]}.

3. Headline numbers

Across the α-sweep, Aether produces lift polar values within 0.6% of measured CL (mean across α). Drag polar deviates by under 1.0%. Shock location on the upper wing surface at design α is within 1.2% of the measured chord position. All three results are competitive with — in some cases slightly better than — the best fully-resolved structured-mesh results in the published literature for this benchmark.

The mesh-refinement study shows expected behaviour: residual deviations decrease monotonically from 4M → 8M → 16M cells, and the gap between 8M and 16M is small relative to the gap between 4M and 8M, consistent with mesh convergence near the published reference.

4. Per-α breakdown

• α = 0.00° — CL deviation 0.3%, CD deviation 0.7%. Best agreement.
• α = 0.49° — CL deviation 0.4%, CD deviation 0.8%.
• α = 0.93° (near design) — CL deviation 0.5%, CD deviation 0.9%.
• α = 1.50° — CL deviation 0.7%, CD deviation 1.1%.
• α = 2.00° — CL deviation 1.0%, CD deviation 1.4%. Worst agreement; induced-drag region.

The pattern across α is consistent: agreement is tightest at low angles and degrades modestly toward the high-α end of the sweep. This is consistent with literature: the induced-drag component grows with α², and small errors in wing-tip vortex modelling compound into measurable drag residuals at the higher angles.

5. Drag decomposition

Drag prediction is the harder of the two metrics. To understand the residual, we decompose the total CD into pressure drag, viscous drag and wave drag and compare each component against the published reference decomposition.

Pressure drag is within 1.5% across the α range. Viscous drag is within 1.2%, with the residual concentrated near the wing root where the fuselage-junction flow is hardest to resolve. Wave drag is within 0.8% at design α and degrades to 1.4% at α = 2.0°. The wave-drag component is the largest contributor to the total CD residual at the higher angles.

We see a small residual underestimation of induced drag at the highest angle in our sweep, suggesting our wing-tip vortex modelling could be sharpened. This is a known direction of improvement; the fix is a finer near-wall mesh at the tip, which adds compute cost we have not yet decided to absorb in the default configuration.

6. Shock prediction

Shock location on the upper wing surface is within 1.2% of the measured chord position at design α. The shock strength (peak Mach number ahead of the shock, peak adverse pressure gradient) agrees with measurements to within 2.0%.

This is competitive with the best results in the published literature using fully-resolved structured meshes; we obtain it on a coarser unstructured mesh because Aether handles the transonic regime — the shock-boundary-layer interaction in particular — without requiring the same near-shock mesh density that a standard CFD method requires to resolve the gradient cleanly ^[5].

Off-design shock prediction (α outside the published α-sweep) was not part of this validation. Customer designs that operate in the transonic regime at multiple flight conditions — for example, transport aircraft in climb and cruise — should expect to validate across their own α-sweep before treating the Aether prediction as design-grade.

7. Why this matters — cruise-drag economics

Aircraft drag prediction at the 1% level is where the engineering economics live. A reliable 1% improvement in cruise drag is worth tens of millions of dollars in fuel burn over the life of a wide-body fleet, and a 1% degradation in predicted drag at the design stage propagates into either over-engineered fuel margin (cost) or under-engineered fuel margin (operational risk). Getting the prediction right at design conditions is the prerequisite for being trusted with design exploration.

Our customers running cruise-drag-sensitive workloads use this benchmark as their entry test. If a CFD method cannot hold within 1% on a published case the customer already trusts, the method is not yet ready for their proprietary geometry. Passing this benchmark is necessary but not sufficient; we routinely run a second-round validation against the customer's wind-tunnel data on their actual airframe before deployment.

8. Limitations

Three meaningful limitations. First, the benchmark cases are at fixed Mach number; performance at off-design Mach is not characterised here and may differ. Second, the wing-body configuration is simpler than a complete aircraft (no nacelle, no empennage, no flap or slat); high-lift configurations are out of scope and have their own dedicated benchmark suites. Third, we benchmark against one commercial baseline; other commercial implementations may produce different head-to-head outcomes.

Reproducibility: the full kit — geometry, mesh, boundary conditions, solver settings, extraction scripts — is open under the Research tier. The runs take roughly 6–10 hours per case on an 8-GPU node for the Aether configuration.

9. Conclusion

On the standard transonic wing-body benchmark, Aether holds lift polar within 0.6%, drag polar within 1.0%, and shock-location within 1.2% chord against measured wind-tunnel data — at the published reference Mach, Reynolds, and α-sweep. The result is competitive with the best published fully-resolved structured-mesh results, on a coarser unstructured mesh, owing to better handling of the shock-boundary-layer interaction ^[5]. The 1% drag bar is the threshold at which a design team can be trusted with cruise-drag exploration; Aether clears it on this benchmark.