Verification Hammer #02 · Mach 10 CUDA

The evidence is downloadable.

A V1–V10 Mach 10 verification chain culminating in an original 1,024,000-cell 3D CUDA result — published with fields, histories, provenance and cryptographic manifests so third parties can inspect the evidence themselves.

Download the Evidence Pack (.zip)Verify it yourself
Canonical Evidence Pack
ECHELEN_MACH10_VERIFICATION_HAMMER_V2_COMPLETE.zip
10,312,957 bytes · SHA-256
3743c3fbec799bb9cb2dff015e7c7d8f49f3613b90f2ba7c732108cd05117d99
Mach 10
declared freestream
1,024,000
V10 3D cells
13 / 13
V10 gates true
37 / 37
outer hashes verified
The published object

This page explains the result. The ZIP carries the evidence.

27,491,217 B

Original V10 evidence

Six V10 artifacts: NPZ fields, VTK, convergence CSV, dashboard, structured summary and nested manifest.

5 / 5

Nested V10 manifest

All five result artifacts protected by the original V10 manifest independently rehash correctly.

6 / 6

Historical continuity

The included V10 originals match the hashes and byte sizes recorded for the historical V10-1000 directory.

Physical condition + original 3D result

Mach 10, quantified.

Using the package's declared γ=1.4, R=287.05287 J/kg/K and T=220 K, Mach 10 corresponds to approximately 2,973 m/s — 2.973 km/s.

Evidence: EVIDENCE/V10_ORIGINAL/V10_3D_SUMMARY.json
160 × 80 × 80

The delivered pressure, density, temperature, Mach, total-enthalpy-ratio and solid-mask arrays have this actual shape: 1,024,000 values per volumetric field.

EVIDENCE/V10_ORIGINAL/MACH10_3D_FIELDS.npz
Original V10 3D dashboard
Original V10 dashboard. The visualization is accompanied by the underlying NPZ and VTK fields; the evidence is not limited to a rendered image.
V1 · analytic reference

Mach 10 benchmarked against an exact oblique-shock reference.

0.016019°

Finest-grid absolute shock-angle error.

3.877×10⁻¹²

Recorded finest-grid residual.

0.9997389

Recorded R² of the finest-grid shock fit.

V1 Mach 10 Euler fields. Public benchmark evidence.
Numerical refinement. The evidence chain records grid refinement toward the exact shock-angle reference rather than a single-grid image.
Evidence: EVIDENCE/V1_V6_PUBLIC/PUBLIC_DATA/PUBLIC_RESULTS_SUMMARY.csv and PUBLIC_VERIFICATION_MATRIX.csv.
V3–V6 · thermal / TPS / material screening

Paired uncertainty and adversarial screening are part of the evidence chain.

73.70%

V3 final objective reduction versus matched reference.

200 / 200

V4 candidate below matched reference.

260 / 260

V5 candidate below matched reference across Sobol plus corner cases.

V5 adversarial dashboard. Temperature reduction spans 5.53–90.15 K in the recorded paired set; hostile transition conditions remain visible in the evidence.
V6 finite-rate/reacting-wall screening. Recorded results include a 3930.36–3219.78 K thermochemical bracket, catalytic-wall increment, 30 MJ/kg recession estimate 1.019→0.966 mm, and constrained thermoelastic bound 0.204–0.755 GPa. These are screening results, not material qualification.
V9–V10 · repetition and cross-backend checks

The V10 PASS is not a lone record.

2 / 2

V10 executions PASS

350-iteration and 1,000-iteration V10 records both report V10_3D_PASS and all 13 declared gates true.

4 / 4

CPU–CUDA discrepancy 0.0

Both V9 and both V10 records report maximum CPU–CUDA discrepancy 0.0 and axis errors [0,0,0] for the declared tests.

Evidence: EVIDENCE/V7_V10_PUBLIC/EJECUCIONES_MACH10_PUBLICAS.json and original V10_3D_SUMMARY.json.
Checks directly recomputable from the delivered arrays

The stored 3D result can be interrogated.

EOS consistency

T=p/(ρR) reproduces the delivered temperature field with maximum stored difference 0.0.

Pressure symmetry

Direct Y and Z reflections of the delivered pressure array give maximum stored discrepancies 0.0 / 0.0.

Extrema

Published extrema for pressure, density, temperature, Mach and total enthalpy are recoverable directly from the NPZ.

Evidence: EVIDENCE/V10_ORIGINAL/MACH10_3D_FIELDS.npz cross-checked with V10_3D_SUMMARY.json.
Audited energy guard

The numerical limiter leaves quantitative accounting.

81,411,756

Cumulative corrected cell updates.

1,344.097 J

Cumulative removed energy.

2.611459%

Removed / (initial-domain + processed-inlet energy), reproducible from recorded values.

Evidence: EVIDENCE/V10_ORIGINAL/V10_3D_SUMMARY.json · energy guard and field extrema.
Why the package matters

Not “the first Mach 10 CFD.” A deliberately inspectable evidence architecture.

Analytic-reference verification + refinement + paired/adversarial thermal screening + finite-rate/TPS screening + repeated CPU–CUDA records + original 3D fields + audited energy guard + historical provenance + nested SHA-256 sealing.

The package does not claim experimental flight validation, independent third-party replication, reacting Navier–Stokes qualification or material certification. Those boundaries are explicitly preserved.

Verify it yourself

Start by downloading the Evidence Pack.

Download the Evidence Pack (.zip)
01Hash the downloaded ZIP. Expected SHA-256: 3743c3fbec799bb9cb2dff015e7c7d8f49f3613b90f2ba7c732108cd05117d99
02Unpack it and verify SHA256SUMS.txt. The audited package has 37/37 protected files matching.
03Open EVIDENCE/V10_ORIGINAL/: inspect the NPZ, VTK, convergence history, dashboard and JSON summary.
04Recompute dimensions, extrema, EOS consistency, Y/Z symmetry and energy arithmetic. Compare the original convergence history with the earlier public extract.
Evidence Pack SHA-256
3743c3fbec799bb9cb2dff015e7c7d8f49f3613b90f2ba7c732108cd05117d99