G05 Final Qualification Evidence¶
Decision Boundary¶
This pack records the Owner closeout for 026-G05 as
PASS_WITH_LIMITATIONS. It does not promote a capability merely because its
runner is executable.
| Item | Result |
|---|---|
| Qualified numerical source | 6cd971756af0f1b8a114fd98b188714da39ff081 |
| Final evidence replay SHA | 77f147ee261d97fd5e6b88281daafb4deba9ce53 |
| Internal family campaign | PASS_INTERNAL_PREQUAL, dirty=false |
| Modal cases | 14 / 14 |
| Newmark cases | 32 / 16 |
| Harmonic cases | 12 / 12 |
| Covered family rows | TET4, TET10, HEX8, HEX20, BEAM2, MITC3/MITC3+, MITC4, discrete |
| Unexpected failures | 0 |
| Official gate | PASS_WITH_LIMITATIONS; Owner decision recorded |
Owner-Approved Bounded Policies¶
The Owner approved the following policies on 2026-08-29. They retain the existing residual, energy and response-quality requirements; no tolerance was weakened.
| Route | Promotion metric | Bounded policy |
|---|---|---|
| Modal | final adjacent tracked-mode frequency change | <= 1%, at least three compatible levels; preserve residual and MAC when available |
| Newmark | final adjacent displacement-history error | <= 1%, same physical interval and comparison times; residual and energy remain separate criteria |
| Harmonic | final-grid amplitude change | <= 1%; preserve phase, peak location and complex residual; exact singular resonance is excluded |
Internal Oracles And Refinement¶
The campaign records generalized modal residuals, mass/stiffness orthogonality, zero-frequency harmonic behavior, complex residuals, energy drift and an assembled first-eigenmode oscillator reference. The maintained SDOF Newmark and harmonic closed-form tests also pass.
Newmark was executed at 30, 60, 120 and 240 steps per period for each baseline
family. The worst final adjacent error recorded by the campaign is
4.720960909695235e-3, below the approved one-percent band. Modal mesh
refinement remains explicitly family-specific bounded evidence. Harmonic
baseline and refined grids were executed; peak and phase remain review data,
not an amplitude-only claim.
External Correlations¶
| Family | Route | Oracle | Result | Maximum recorded error / limit |
|---|---|---|---|---|
| Discrete | Modal, Newmark, harmonic | Code_Aster DIS_T |
PASS | 3.05e-11 / 1e-7 |
| BEAM2 | Modal | Code_Aster POU_D_E |
PASS | 2.65e-4 / 1e-2 |
| BEAM2 | Newmark, harmonic | Code_Aster POU_D_E |
PASS | 4.75e-13 / 1e-7 |
| TET4 | Modal, Newmark, harmonic | Code_Aster TETRA4 |
PASS | 4.64e-2 / 1e-1 |
The pinned Code_Aster image is
simvia/code_aster@sha256:4629a21a109309bb97fbdc27d750445cc869e151e2e2ed6290f69539614e4435.
The raw results and their digests are listed in
qualification/0_2_6/g05_final_evidence.json.
These four correlations were replayed successfully on the final evidence replay
SHA with dirty=false. TET10, HEX8, HEX20, MITC3 and MITC4 have no same-SHA,
formulation-compatible external deck in this run and are not claimed as
externally correlated.
Bounded Limitation¶
TET10, HEX8, HEX20, MITC3 and MITC4 rely on internal quantitative V&V and analytical evidence pending comparable independent external decks. This limitation is part of the approved G05 scope and is not an unresolved gate blocker.
No numerical formulation, solver API, physics model or gate requirement was modified by this pack.