Skip to content

QF Solver 0.2.5a0 — Owner Review 025-G03

Decision identity

Field Value
Gate 025-G03
Owner decision APPROVED
Mesh decision APPROVED_BOUNDED_REFINEMENT
Qualified source SHA 85c75d06955976251dd54ad782f57f1eb5a7f8f4
Qualified source worktree CLEAN
Owner-evidence manifest qualification/reviews/qf_solver_0_2_5_g03_owner_evidence_manifest.json
Contract lowered NO

This is an explicit Owner closure for the bounded G03 scope. It is not a whole-release decision and does not authorize a push, tag or PyPI publication.

Decision scope

This review covers the bounded first tangent-instability path only. It does not qualify post-buckling, imperfection-sensitive collapse, nonlinear stability beyond the first tangent loss, or external buckling for every solid family.

Evidence

Evidence Result Location
Sparse initial-stress geometric tangent PASS src/solveur/core/buckling.py and solid assembly implementations
Exact pure-geometric tangent unit check PASS tests/unit/test_linear_buckling.py
Targeted buckling/regression tests 26 passed local targeted run on qualified source SHA
Euler refinement PASS results/vnv_0_2_5/g03_euler_final/summary.json
Code_Aster factor/mode probe PASS_EXTERNAL_CORRELATION_BOUNDED results/vnv_0_2_5/g03_final/summary.json
Adversarial failure contract PASS results/vnv_0_2_5/g03_final/adversarial.json

The qualified source SHA is:

85c75d06955976251dd54ad782f57f1eb5a7f8f4

Both final evidence manifests record dirty=false at generation time. The Code_Aster image is pinned to simvia/code_aster@sha256:4629a21a109309bb97fbdc27d750445cc869e151e2e2ed6290f69539614e4435. The current local replay could not start because the image launcher lacks mpi4py; this is recorded as an environment limitation and does not replace the archived Code_Aster execution.

Measured correlation

Quantity QF Solver Code_Aster Difference
First critical factor 221.54828247814925 221.774 1.018e-3 relative
Best modal MAC 0.9999999989229131 1.0 reference bounded agreement
QF critical-mode residual 1.72e-15 archived residuals <= 5.07e-15 converged

The Euler four-level campaign reports critical-load changes of 15.63 %, 6.57 % and 3.31 % between successive levels. The finest level is within 5.89 % of the analytical Euler reference. These results support a bounded refinement trend; they do not establish universal convergence.

Qualified claim

linear_buckling is qualified for the sparse first tangent-instability path in the recorded total-Lagrangian solid scope, with TET4 as the externally correlated family. TET10, HEX8 and HEX20 remain internal/research for this gate. CalculiX remains a SHOULD comparison and its negative high-order probe is retained as diagnostic evidence; it is not a G03 blocker.

CONTRACT_LOWERED = NO

Owner decision

Decision: APPROVED for the bounded scope above.

The release-level Owner decision remains separate from this gate closure. No push, tag or PyPI publication is authorized by this document.