Skip to content

0.2.6 G07 B2 — TL external completion

Evidence ID: 026-G07-B2-TL-EXTERNAL-001 Gate: 026-G07 Step: B2_TL_EXTERNAL_COMPLETION

Decision

G07_B2 = PARTIAL. The run did not change the G07 gate status, TL formulation, Newton/physics code, thresholds, or default path.

The TET4 bounded external path is complete and usable as PASS_WITH_LIMITATIONS. The HEX8 Code_Aster path is complete, but the QF path does not reach the end of the same case; therefore no HEX8 matched-path claim is made.

Controlled provenance

  • B1 start / B2 baseline: 1fd3cd41d1dd21e26d90851d89aa60d9429dabd9
  • runner source: scripts/run_g07_b2_tl_external.py
  • runner source SHA: 52e45e44ab3b38031dd5062b7ed85dcb30f53d4d
  • external solver: Code_Aster 18.1.0
  • pinned image: simvia/code_aster@sha256:4629a21a109309bb97fbdc27d750445cc869e151e2e2ed6290f69539614e4435
  • machine-readable evidence: qualification/0_2_6/g07_b2_tl_external/g07_b2_tl_external_evidence.json

The runner records case-definition hashes, input hashes, runtime status, and UTC capture provenance in the JSON artifact.

TET4 — complete bounded path

The same geometry, mesh, material, nodal loading, and bounded endpoint were used for both paths. The imperfect column was evaluated over 16 increments, from 0.2 through 0.8 of the documented critical-load estimate. All 16 Code_Aster and 16 QF points are finite, load-monotone, and path-continuous.

Observable Result
displacement maximum relative difference 1.4115368881294746e-14
displacement maximum absolute difference 6.245221350186542e-16
signed reaction maximum relative difference 6.204270246045218e-14
signed reaction maximum absolute difference 3.694859482287339e-12
QF maximum relative residual 1.3368719921494182e-09
QF minimum det(F) 0.9945552598356373
matched points 16/16

The QF physical support reaction is defined as the negative of the QF external-minus-internal residual at fixed DOFs and is compared with the Code_Aster REAC_NODA resultant. Column stress/strain fields are explicitly not compared because integration measures and sampling are not mapped one-to-one.

Result: TL_TET4_EXTERNAL_COMPLETE = YES, TL_TET4_RESULT = PASS_WITH_LIMITATIONS.

Limitations are the tested bounded load domain, the incompatible stress/strain field mapping, and the fact that this is not a general TL or large-deformation qualification.

HEX8 — complete external half, incomplete matched QF path

The existing bounded HEX8 compression case was used unchanged: HEX8_m4_a10_compression_l0.2_d0.12, 128 increments, distortion 0.12, aspect 10.0. Code_Aster completed all 128/128 external points with the same case geometry, mesh, material, and nodal load definition.

The QF production path did not complete. It failed deterministically at increment 48/128 with:

NumericalConvergenceError: Full Newton did not converge at increment 48; relative residual=4.392803e-04.

Because both full runtime paths are required for a matched comparison, the HEX8 result is FAIL for this B2 proof. No external stress, energy, or external det(F) comparison is claimed; those fields are not transformed to directly comparable QF measures. A separate diagnostic probe with 512 increments also failed later in the same physical path region (increment 192), so it was not treated as a proof or used to alter the route.

Result: TL_HEX8_EXTERNAL_COMPLETE = NO for the matched proof, TL_HEX8_RESULT = FAIL.

Scope and remaining gaps

  • No real solver bug was identified: the observed HEX8 outcome is an explicit production-path non-convergence, not a silent pass, NaN/Inf, state corruption, or changed classification.
  • REAL_BUG_FOUND = NO.
  • FUNCTIONAL_SOURCE_CHANGED = NO.
  • NUMERICAL_REGRESSION = NO.
  • Full regression is skipped by the B2 policy.
  • G07-TL-008-HEX8-COMPLETE-HISTORY remains blocking for the requested complete HEX8 matched-path claim.
  • The G07-B1 Arc-Length G07-ARC-002 gap remains unchanged and is outside this B2 step.
  • G07 Owner closeout is therefore not ready; no TL promotion is performed.