Skip to content

0.2.6 G04 execution evidence

Decision boundary

This is the official execution of the controlled G04 READY selection. It does not close the global gate, change maturity, or replace missing family-wide and external evidence. The branch-local record originally left the global gate unchanged; multi-agent consolidation now records 026-G04 globally as PASS_WITH_LIMITATIONS with the bounded scope and limitations preserved.

The complete machine-readable per-case relation is in qualification/0_2_6/g04_execution_evidence.json. Every row records:

CASE_ID -> REQUIREMENTS -> ELEMENT FAMILY -> ORACLE STATUS -> POLICY RESULTS -> RESULT.

Execution provenance

Item Value
Source SHA 5e798e2fd052cb4fe8618d06495a2287f29e01b3
Branch agent-b/026-work
Runner scripts/run_vnv_026.py --profile FULL --case ...
Registry digest d2e5f587ac2d67c3c875782c8558aa7aac60cec09df9f43614319eff1b99ea59
Runner run ID vnv026-full-d2e5f587ac2d
READY selection 65 cases: 56 LIN + 9 SHL
PLANNED selection 4 not materialized; no requirement required them
NOT_APPLICABLE selection 3 mixed SHL records not executed

Results

Result Count
PASS 65
WARNING 0
EXPECTED_FAILURE 0
FAIL 0
SKIP (case execution) 0

Resolved element-family counts are BEAM2 2, MITC3 3, MITC4 3, TET4 14, TET10 8, HEX8 25, HEX20 9 and RBE2 1. No registered DISCRETE element case is READY in this selection. The RBE2 record is diagnostic-only and is not counted as DISCRETE G04 evidence. All 65 cases are linear-static dispatches and completed deterministically.

Quantitative policy checks

Policy Maximum observed Bound Status
Free relative residual 2.6741e-14 <=1e-8 PASS
Force balance 1.4873e-12 <=1e-10 PASS
Moment balance 3.2411e-12 <=1e-10 PASS

The four repeated HEX8 mesh series (nx=1,2,4,8) were executed and show sub-1% final adjacent changes for max_displacement. They are retained as diagnostic series only: the G04 registry did not predeclare the required q_ref, so the Owner-approved mesh policy is not marked PASS retroactively.

Analytical and external evidence

Twenty cases had an executable predeclared constrained-free-DOF analytical oracle. They all passed with maximum relative error 0.0 and tolerance 1e-10. Thirty cases declare ANALYTICAL in the registry but have no executable oracle configuration; they are reported as DECLARED_NOT_EXECUTED, never as analytical PASS. Fifteen cases have internal regression/invariant evidence only.

Code_Aster and CalculiX executables were unavailable. Both external correlations are therefore SKIP; no external PASS or non-comparable claim is made.

Requirement coverage and limitations

Requirement Status Limitation
G04-LIN-001 PASS All selected routes dispatched and completed.
G04-LIN-002 PARTIAL Foundation patch/invariant cases ran; family-wide patch coverage is not established.
G04-LIN-003 PASS Load, reaction, force/moment and residual audits ran for all cases.
G04-LIN-004 PASS_WITH_LIMITATIONS Isotropic routes pass; orthotropic/laminate records remain deferred from the base claim.
G04-LIN-005 PARTIAL Four-level HEX8 records ran; q_ref and quality/distortion evidence remain incomplete.
G04-LIN-006 PARTIAL No invalid-input case was in the READY G04 selection; rejection tests remain separate evidence.
G04-LIN-007 NOT_COVERED External executables unavailable.
G04-LIN-008 PASS Source, clean state, registry digest and per-case artifacts recorded.

The resulting execution status is PASS_WITH_LIMITATIONS: the numerical selection is clean, but the missing analytical configurations, external correlation, family-wide patch/failure coverage and predeclared mesh reference prevent an unconditional G04 decision.