Skip to content

Verification, validation, evidence and maturity

Current release documentation: QF Solver 0.2.11
Current version DOI: 10.5281/zenodo.23214487
Published: 2026-10-07
Previous version DOI (0.2.10 only): 10.5281/zenodo.23106744

The cumulative solver retains the 0.2.10 development made since the published 0.2.8 baseline, including the development-only 0.2.9 cycle. The 0.2.10 V&V summary describes the bounded records; the capability index is the public navigation to current scopes. The 46-combination 0.2.8 registry is historical authority for those published element-analysis decisions and is not silently rewritten here.

QF Solver 0.2.11 adds experimental gyroscopic modal and Campbell evidence. Those results, their availability and their maturity decisions are summarized separately in the 0.2.11 V&V page; they do not rewrite the published 0.2.10 record. The selected 0.2.11 wheel and sdist are published on PyPI, with matching selected assets on GitHub Releases and Zenodo. The version DOI above identifies the Zenodo record; this page summarizes evidence but is not a substitute for its controlling records.

Four distinct claims

  1. Implementation: a code path exists.
  2. Verification: a declared test, invariant, or numerical comparison was executed and met its stated checks.
  3. Bounded acceptance / qualification: a decision explicitly accepts a named scope and limitations.
  4. Physical validation: a model is compared with physical observations.

Passing software tests or comparing against another FEM solver is not, by itself, physical validation. An Owner-accepted experimental record remains experimental if that is what the decision says. Maturity labels must not be upgraded merely because an implementation or evidence pack exists.

Evidence chain

A bounded campaign records a prospective scope and gates, a frozen execution, results and source/environment provenance, independent checks and their limitations, replay where applicable, fail-closed behavior, and a final decision record. Historical failures and superseded dispositions remain preserved; they are not rewritten to match a later desired outcome.

Maturity vocabulary

Label Meaning
QUALIFIED_BOUNDED The declared combination passed its frozen qualification gates within the recorded scope; it is not universal.
EXPERIMENTAL_BOUNDED A usable route with limited evidence and explicit restrictions, not a general qualified capability.
EXPERIMENTAL Exploratory or partially verified evidence is insufficient for bounded qualification.
RESEARCH_ONLY Discovery or feasibility work; no public supported-use claim.
NOT_VALIDATED The implementation or architecture may exist, but required validation gates did not pass or were not run.
INTERNAL Retained for development and traceability, not a public supported route.
GO_WITH_LIMITATIONS An audit disposition for its declared campaign; it is not automatically a maturity promotion.

The 0.2.8 registry counts remain 32 QUALIFIED_BOUNDED, 14 EXPERIMENTAL, and 0 NOT_QUALIFIED across 46 specified combinations. These counts are not a current-version quality score and exclude separate mixed workflows and capability records.

Inherited structural scope boundaries

  • Small-strain J2 and the 0.2.8 linear registry retain their original bounded scopes.
  • Corotational J2 has an Owner-accepted bounded HEX8 scope with large rotations and small local strain; no general finite-strain claim follows.
  • The Total-Lagrangian StVK audit covers selected serial static TET4/HEX8 cases and states GO_WITH_LIMITATIONS without maturity promotion.
  • Coupled nonlinear Owner acceptance is limited to selected static TET4, TET10, HEX8, and HEX20 evidence. It excludes friction, dynamics, MPI/PETSc, external solver correlation, and independent global FEM/Newton solution.
  • Frictionless contact remains experimental. Frictional contact evidence is narrow, mesh-sensitive, and not formally requalified against the current source.
  • PETSc/MPI acceptance is limited to two-rank linear-static cases with replicated input/root-side assembly. No general nonlinear distributed or scaling claim follows.
  • Code_Aster evidence is bounded same-mesh linear-static numerical correlation for recorded observables, not physical validation.
  • WP14 remains HOLD while the whole-repository G03 archive scan is failed. The selected wheel/sdist audit does not clear that gate or automatic GitHub source archives.

The detailed development-cycle records remain in qualification/0_2_9. They are engineering evidence, not user-level replacements for the concise scope statements above.

Experimental rotating scope

Gyroscopic modal analysis is limited to linear straight circular BEAM2 shafts with centered rigid axisymmetric disks and a dense serial QEP. Campbell diagrams orchestrate this route and use explicit complex-mode tracking. The high-frequency pair at 100 rad/s remains ambiguous; the tracker does not force continuity. GYRO-06 supplies internal mesh-convergence evidence, not independent physical validation. Both capabilities remain EXPERIMENTAL.