Skip to content

0.2.6a0 Master Plan

Objective

0.2.6a0 is a maturity, reproducibility and architecture cycle. It makes the qualification system executable as controlled work packages before any claim is expanded. This foundation run adds no FEM physics and does not certify a new release.

Ordered Execution

  1. G00 baseline and provenance.
  2. G01 architecture audit.
  3. G02 registry, runner and manifest contracts.
  4. G03 corpus design.
  5. G04-G13 controlled capability batches, each followed by evidence and gate review.
  6. G14 full regression and architecture freeze.
  7. G15 Owner review.

Every batch follows: audit, implement only the approved narrow change, verify, benchmark where applicable, correlate, gate, then move to the next package. An OPEN gate is recorded as OPEN; it is never converted to PASS by a new label.

G07 Step 1 Contract Boundary

The official gate is 026-G07, titled Geometric nonlinear and arc-length review. Step 1 is governance and contract work only; it does not close the gate and does not create numerical evidence.

The machine-readable contract is qualification/0_2_6/g07_requirements.json; the capability-by-family matrix is qualification/0_2_6/g07_capability_matrix.json. The candidate bounded qualification scope is Total-Lagrangian elasticity on TET4 and HEX8. TET10 and HEX20 remain RESEARCH for this gate. Existing finite-kinematic J2 and coupled nonlinear workflows are excluded from G07.

The existing arc-length route is limited to EXPERIMENTAL / PASS_INTERNAL_RESEARCH. A successful internal run cannot promote it to a production-qualified capability. The contract requires branch, turning-point, restart/rollback and sensitivity evidence, but unresolved acceptance bands are marked PROPOSED_OWNER_REVIEW rather than assigned a new numerical threshold.

Step 2 Execution Plan

Step 2 may start only after the Owner approves the policies marked PROPOSED_OWNER_REVIEW:

  1. Run bounded TL elasticity cases for TET4 and HEX8: small-strain limit, objectivity/rigid-body invariance, large-rotation domain, residual and equilibrium, tangent finite differences, mesh refinement at coarse/medium/fine/refined levels, increment refinement, and reproducibility.
  2. Run high-order TET10 and HEX20 only as explicitly labelled research smoke cases; do not aggregate them into the qualified G07 claim.
  3. Run the existing arc-length shallow-arch and common FEM snap-through cases with nominal, smaller and larger recorded arc-step settings, plus branch tracking, limit point, restart/rollback and controlled failure cases.
  4. Use the existing analytical shallow-arch oracle and compatible Code_Aster or CalculiX comparisons when available. An unavailable or non-comparable external tool is SKIPPED, never PASS.
  5. Emit per-case JSON, complete curves/diagnostics, environment and tool versions, source_sha, dirty=false, threshold source and artifact digests. Stop on unexplained divergence, a required new formulation, or a threshold change.

G07 remains NOT_STARTED until a separate evidence pack and Owner decision meet the contract.