Wall-function treatment
Maximum y+ 12.67
Area inside selected screen 0.0%
Near-wall screen FAILSignificant active-field residuals remain at the iteration limit.
VERIFICATION BENCH · REV 1.0
A finished solve can still be wrong.
Inspect the cases that pass, the controls that fail, and the evidence that tells them apart.
01 / LAMINAR PIPE · Re 200
Same target. Five real outcomes.
The analytical target is 0.64 Pa. Three grids pass the 2% allowance; two negative controls do not.
| Case | Cells | Δp / Pa | Error | Gate |
|---|---|---|---|---|
| coarse | 320 | 0.633629 | 0.996% | PASS |
| medium | 960 | 0.638447 | 0.243% | PASS |
| fine | 2,560 | 0.639663 | 0.053% | PASS |
| underresolved | 80 | 0.550638 | 13.963% | FAIL |
| wrong viscosity | 960 | 1.277473 | 99.605% | FAIL |
All five recorded outcomes are shown in the table above. Interactive gate selection is available on the full bench page when JavaScript is enabled.
What was measured? What changed after the first diagnostic?
D = 10 mm, L = 100 mm, bulk speed = 0.02 m/s, density = 1000 kg/m³, kinematic viscosity = 10⁻⁶ m²/s. Pressure is a fitted interior gradient × full length, not entrance loss. Darcy f = 64/Re = 0.32; Fanning f = 0.08.
The first gate rejected an ill-conditioned residual in an essentially zero transverse velocity. Criteria 1.1 introduced an explicit component-to-bulk scale check, then all seven CFD cases were rerun. This is a disclosed post-diagnostic correction, not a blind preregistration. Original diagnostic summaries remain in the archive. Pipe runs reached the fixed iteration limit; the explicit scale-aware acceptance checks are separate from the built-in stopping flag.
The underresolved case has two radial cells. The wrong-viscosity deck doubles viscosity without changing the original requirement. Their errors are measured, not assigned by a scripted FAIL label.
Conservation is a plot, not just a green cell.
02 / NEW LOW-MACH NOZZLE DIAGNOSTIC
This mesh is not ready.
Same 1,920-cell grid, same SST closure, different near-wall treatments. Both fail the selected y+ area screen. Their residual behavior differs; neither reaches the built-in stop. This does not revalidate the older wall-refinement-v6 case.
Direct low-Re walls
Maximum y+ 11.64
Area inside selected screen 6.2%
Near-wall screen FAILActive-field residuals are small; the negligible transverse component prevents the built-in stop.
Inspect residuals, thresholds and model assumptions
Bench screens: 95% of nominal wall area at 30 ≤ y+ ≤ 300 for the log-wall treatment; 95% at y+ ≤ 1 for direct low-Re walls. These are selected requirements, not universal validity laws. The two treatments may evaluate effective wall scaling differently; the comparison is not an independent wall-shear validation.
Direct low-Re pressure, in-plane velocity, k and omega residuals are below 10⁻¹⁰. Its normalized Uz residual remains about 0.003 despite max |Uz| / inlet speed being only 5.12 × 10⁻¹³. This is different from the wall-function case's significant active-field residuals. Both still fail near-wall suitability.
Inlet Re ≈ 5,333, no transition model, no independent nozzle reference. Direct-wall omega uses approximate first-cell height. The report explains what must be established before a performance prediction can be accepted.
03 / IMPORTED STEP · FILLET + INTERSECTING PORTS
Reject the defect. Keep the original.
Intact manifold
Topology PASSOne valid connected solid. Physics, fluid region and boundary assignments remain missing: overall readiness is not established.
Detached sliver
Intake FAILTwo connected components violate the single-solid contract. Quarantined for review before mesh or solve. Multiple solids can be intentional elsewhere; this gate does not silently delete them.
04 / VERSIONED, DOWNLOADABLE EVIDENCE
Keep the decks. Check the hashes.
Seven new OpenFOAM runs, two STEP imports, and five retained CalculiX solves. 295 hashed source and evidence files. These curated downloads have no 48-hour job expiration.
Versions, reproduction commands and retained CalculiX scope
OpenFOAM Foundation 10; CalculiX 2.20; Gmsh 4.15.2; CadQuery 2.7.0; OCP 7.8.1.1. The archive includes original decks, meshes, logs, final fields, criteria and an offline verifier. Local image IDs identify the executed environment; they are not public image download URLs.
python source/audit.py .
Run this after extracting the archive. Copy decks to a new scratch directory before rerunning. Complete reproduction instructions · Machine-readable results
The CalculiX subset contains the three-level pressure tube, a ported-tube fine solve and a thermal-gradient fine solve, recorded September 12. Original receipts preserve failed and incomplete gates. They are not five new solves.