Skip to content

026-G06 J2 Prequalification Evidence

Decision boundary

026-G06 remains NOT_STARTED as an official gate. This pack records a controlled prequalification replay and proposes PASS_WITH_LIMITATIONS; it does not promote the J2 maturity claim or close the gate automatically.

The official scope is the J2 maturity extension. Small-strain J2 is the in-scope capability. Finite-kinematic J2 remains RESEARCH_NOT_QUALIFIED, and coupled nonlinear workflows remain EXPERIMENTAL_NOT_QUALIFIED.

Provenance

Item Value
Execution source SHA ad99f699a3b07368401f7607060bc06bb1860205
Branch chore/0.2.6-foundation
Source worktree dirty=false
Solver qf_solver 0.2.6a0
Registry digest d2e5f587ac2d67c3c875782c8558aa7aac60cec09df9f43614319eff1b99ea59
Campaign manifest SHA-256 9bb8f3ab3642b9a9bf3f1e13f4053d25aa8ebdca7bcd3145eb5deed6601851c8
Quantitative manifest SHA-256 ee9c3d7b894f034e87384c2a6251100b9416291af35c5d7ed82c1a9bb543e5b5
Environment Windows AMD64, CPython 3.13.1, NumPy 2.2.6, SciPy 1.15.2

Controlled G06 corpus

Command:

python scripts/run_vnv_026.py --profile G06 --output results/vnv_026_g06_prequalification
Result Count
PASS 79
EXPECTED_FAILURE 1
FAIL 0
BLOCKED 0

The 80 cases contain 19 TET4, 6 TET10, 33 HEX8, 17 HEX20, 2 MITC3, 2 MITC4 and 1 BEAM2 case. The J2 subset is 7 PASS: TET4 (3), HEX8 (2) and HEX20 (2). There is no dedicated TET10 J2 case in the current G06 registry. TET10 J2 V&V tests outside this corpus pass, but they are not silently counted as a G06 corpus case.

The single expected failure is VNV026-RBT-G06-024, an inverted TET4 rejected as INVALID_ELEMENT. It is a controlled negative case, not an unexpected failure.

Quantitative evidence

The independent analytical study passes 20/20 cases across TET4, TET10, HEX8 and HEX20 with maximum relative error 0.0 against the constrained free-DOF stiffness oracle and tolerance 1e-10. This is linear-elastic evidence and is not a J2 constitutive oracle.

The separate HEX8 uniform axial bar study uses four levels (nx=1,2,4,8). Relative tip-displacement errors are 0.09, 0.0577737, 0.0441261 and 0.0399926; the trend is non-increasing. Residuals remain between 9.04e-17 and 2.46e-15. This is bounded mesh evidence for that bar only, not a general J2 mesh-convergence claim.

Targeted verification

The J2, nonlinear-state, tangent, solid-element and robustness selection returned 102 passed, 3 skipped, 0 failed in 186.66s. Ruff, compileall, registry validation, anti-forgetting validation and git diff --check passed. Full regression was not rerun because the G05 closeout and this pack contain no functional FEM change.

External evidence and limits

No compatible Code_Aster or CalculiX G06 deck was executed in this local prequalification. Both external entries are therefore SKIPPED, never PASS. The pack makes no external-correlation claim.

Remaining evidence required before an official G06 closeout is a dedicated TET10 small-strain J2 corpus case if the family-wide scope is retained, a compatible external correlation, and an explicit Owner promotion decision. Finite-kinematic J2 and coupled nonlinear workflows remain outside the qualified claim.

Status

Item Status
Prequalification pack PASS_WITH_LIMITATIONS
Official 026-G06 NOT_STARTED
Proposed decision PASS_WITH_LIMITATIONS
Unexpected failures 0
Numerical source change NO

SUPPORTED != TESTED != VERIFIED != QUALIFIED remains enforced.