Skip to content

0.2.6 G10-LOT1 - Advanced Nonlinear / Research Audit

Status: PASS_WITH_LIMITATIONS for Lot 1. This is an intermediate checkpoint; the later Owner closeout records gate 026-G10 as PASS_WITH_LIMITATIONS.

This document records an audit of nonlinear and research routes already present in QF Solver. It is an evidence boundary, not a new numerical qualification and does not add physics, elements, solver behavior, or release scope.

Provenance

  • Execution source SHA: 51b3a7c8ace6731830109984a01ce31f79c44401
  • Execution worktree: clean at audit start
  • Evidence SHA: the commit containing this document and the companion JSON
  • Full regression: not run in Lot 1 by policy
  • Numerical source changes: none

The execution source SHA identifies the code audited by the targeted suite. The documentation evidence may be committed at a later SHA and must not be treated as a new numerical qualification.

Scope and ownership

G10 inventories and bounds existing nonlinear/research behavior. Existing qualification ownership is preserved:

Route Current classification Owning gate Boundary
Small-strain nonlinear static Qualified bounded, inherited G06 No scope expansion in G10
Total-Lagrangian elasticity Experimental only G07 G07 remains parked pending external evidence
Arc-length continuation Experimental only G07 Internal branch evidence is not a production claim
Finite-kinematic J2 Research / not qualified G10 No general finite-strain plasticity claim
J2 plus geometry Experimental only G10 Coupled external evidence is missing
Geometry plus frictionless contact Experimental only G10 G09 contact evidence does not qualify the coupled route
J2 plus geometry plus contact Experimental only G10 Triple coupling is not release-qualified
Modified Newton on finite routes Not qualified G10 Existing scope rejection is retained
Transaction/checkpoint/retry Pass with limitations G09/G11 Existing bounded evidence only
Structured failure diagnostics Pass with limitations G11 Existing bounded failure corpus only

G10 does not reopen or change G07, G08, G09, G11, or G12 decisions.

Code-path audit

The audited routes use the following existing paths:

  • NonlinearStaticSolver provides the common load-control Newton driver, material-state transactions, adaptive stepping, and the arc-length mixin.
  • GeometricNonlinearStaticSolver uses the existing Total-Lagrangian assembly, Green-Lagrange kinematics, and the common Full Newton driver.
  • CompositeNonlinearAssembly composes material, geometric, and contact contributions under the common residual/tangent infrastructure.
  • The current finite-kinematic material route uses Green-Lagrange strain and second Piola stress. Its own documentation does not claim general finite- strain plasticity.
  • Contact composition remains frictionless and bounded by the existing G09 contract. No new contact formulation is audited or proposed.
  • The checkpoint, rollback, convergence, finite-value, and typed failure paths are covered by the existing targeted contracts listed in the JSON matrix.

No real bug was identified in this lot. The targeted evidence did not show a new NaN/Inf path, silent convergence, determinism failure, or numerical source regression.

Targeted audit campaign

The Lot 1 suite comprised 171 targeted tests and completed with:

Result Count
Passed 171
Failed 0
Skipped 0

The command covered public and contract tests for geometric nonlinear analysis, Total-Lagrangian J2, load paths, checkpoint/restart, state transactions, failure modes, contact composition, constitutive V&V, multi-element behavior, benchmarks, the existing Code_Aster arc-length deck contract, and the geometric-nonlinear CLI. It was a targeted suite, not the full repository regression.

The machine-readable case-to-route and requirement mapping is in qualification/0_2_6/g10_research_audit_matrix.json. The adversarial Owner review and external-route selection are in 0_2_6_g10_owner_review_lot1.md.

Requirement disposition

The nine Lot 1 requirements are evaluated as follows:

  • Inventory, compatibility, ownership preservation, scope boundaries, and provenance: SATISFIED.
  • State/rollback/failure handling, determinism/finite values, and the compact targeted campaign: SATISFIED_BOUNDED.
  • Formulation-compatible external correlation for finite-kinematic and coupled routes: DEFERRED_LIMITATION.

No requirement is marked as a full G10 capability qualification. A passing targeted test means that the exercised contract behaved as expected; it does not promote a route from EXPERIMENTAL_ONLY or RESEARCH.

External evidence boundary

No new external run was performed in Lot 1. Existing G06 and G09 evidence is retained only where its documented scope is compatible. It does not establish:

  • external validation of finite-kinematic J2;
  • external validation of J2 plus geometric nonlinearity;
  • external validation of geometry plus contact as an equivalent formulation;
  • external validation of triple coupling; or
  • physical branch-selection validation for arc-length continuation.

Those gaps remain explicit closeout blockers if any of those routes are later proposed for a stronger maturity claim. An external tool or deck that is not formulation-compatible must remain skipped or bounded, never promoted to PASS.

Decision and next step

Lot 1 is a controlled audit checkpoint with PASS_WITH_LIMITATIONS status. The later Owner closeout records 026-G10 as PASS_WITH_LIMITATIONS. The only qualified-bounded capability carried into the classification is the existing small-strain nonlinear scope owned by G06; it is not broadened here. Transaction and failure diagnostics retain their existing bounded classifications under G09/G11.

The next G10 step is Owner review of this inventory and, only if requested, the smallest additional evidence needed for a specific route. A future step must not weaken thresholds, promote parked TL evidence, or treat internal coupling tests as external qualification.