QF Solver 0.2.8 WP01 maturity baseline — historical/intermediate record

Historical / intermediate record. This page preserves the chronological WP01 and WP13 working history for auditability. Interim dispositions below do not define the current public status. For current 0.2.8 maturity, use the central capability index and the consolidated registry.

Current authoritative 0.2.8 state

This file is a chronological WP13 record, so interim counts and failed campaigns below are retained as historical observations. The authoritative current element-analysis source is qualification/0_2_8/consolidated_registry.json: 32 QUALIFIED_BOUNDED, 14 EXPERIMENTAL, 0 NOT_QUALIFIED, total 46.

Connected conforming mixed static, modal, translational-MPC and multi-material workflows are separate QUALIFIED_BOUNDED records. Mixed Newmark and harmonic, the bounded .inp subset, family-aware mixed HDF5 storage, bounded frictionless contact and HEX8-SRI are separate EXPERIMENTAL_BOUNDED workflows/capabilities. PYRAMID5 remains internal/research-only. The generic mixed PETSc/MPI runtime remains NOT_VALIDATED; its WP13-01C initial FAIL_RUNTIME, R1 STILL_FAIL_RUNTIME, R2 diagnostic PASS_DIAGNOSTIC_FIX_REQUIRED, and R3/R4 STILL_FAIL_PHYSICAL_CONVERGENCE history is not overridden by architecture readiness or partial two-rank results.

This record audits the 46 active combination records from the public 0.2.7 capability registry at the frozen baseline 9a0d2ab8e8fc3509f0f7ea53bb93db0bd7d99ca5. It is a planning disposition, not a qualification-state change. The 0.2.7 registry and all its historical evidence remain unchanged.

The machine-readable source is qualification/0_2_8/wp01_maturity_matrix.json. It preserves the exact source-registry digest, every combination identifier, the required next proof, test, oracle, tolerance policy, exit gate and risk.

Disposition policy (WP01 historical context)

• KEEP_QUALIFIED: retain the current bounded 0.2.7 state; WP01 adds no claim.
• PROMOTABLE_WITH_EXISTING_EVIDENCE: reserved for a record with complete evidence and an explicit existing Owner decision. No record meets that bar.
• NEEDS_MINOR_VNV: an existing bounded implementation and evidence basis require a frozen replay, predeclared tolerance and Owner decision.
• NEEDS_MAJOR_VNV: a missing comparable oracle, route-level evidence or scope expansion prevents a small administrative closure.
• KEEP_EXPERIMENTAL: a deliberate technical boundary remains out of scope.
• NOT_QUALIFIED_TO_CLOSE: the active record has an explicit unresolved qualification gap.

Every future numerical tolerance must name its observable, numerical value and provenance before a run. Case-dependent values are allowed only when justified by mesh, formulation and oracle. Retuning after a result is forbidden.

Matrix summary (WP01 historical context)

Disposition	Count
KEEP_QUALIFIED	20
PROMOTABLE_WITH_EXISTING_EVIDENCE	0
NEEDS_MINOR_VNV	9
NEEDS_MAJOR_VNV	8
KEEP_EXPERIMENTAL	8
NOT_QUALIFIED_TO_CLOSE	1

Combination dispositions (WP01 historical context)

Family	Routes and WP01 disposition
BEAM2	static, modal: NEEDS_MINOR_VNV; Newmark, harmonic: NEEDS_MAJOR_VNV.
DISCRETE	static, modal: NEEDS_MINOR_VNV; Newmark, harmonic: NEEDS_MAJOR_VNV.
MITC3	static: NEEDS_MINOR_VNV; modal, Newmark, harmonic: NEEDS_MAJOR_VNV.
MITC4	static, modal, Newmark, harmonic: NEEDS_MINOR_VNV.
TET4	static, modal, Newmark, harmonic, linear buckling: KEEP_QUALIFIED; nonlinear load control and geometric nonlinear static: KEEP_EXPERIMENTAL.
TET10	static, modal, Newmark, harmonic, linear buckling: KEEP_QUALIFIED; nonlinear load control and geometric nonlinear static: KEEP_EXPERIMENTAL.
HEX8	static, modal, Newmark, harmonic: KEEP_QUALIFIED; linear buckling: NOT_QUALIFIED_TO_CLOSE; nonlinear load control and geometric nonlinear static: KEEP_EXPERIMENTAL.
HEX20	static, modal, Newmark, harmonic, linear buckling: KEEP_QUALIFIED; nonlinear load control and geometric nonlinear static: KEEP_EXPERIMENTAL.
WEDGE6	modal: KEEP_QUALIFIED; static: NEEDS_MAJOR_VNV.

Evidence findings and priority candidates (WP01 historical context)

The source registry confirms that each active record has a declared implementation path, tests and recorded evidence references. It also records where those references are inherited from earlier releases rather than a new 0.2.8 decision. That distinction prevents an administrative promotion.

1. BEAM2 and DISCRETE static/modal have existing element, workflow and external/analytical evidence routes. Their remaining closure is a frozen 0.2.8 replay manifest, predeclared observables/tolerances and an Owner decision. Their transient and harmonic routes need a comparable history or frequency-response oracle and remain major work.
2. MITC4 has the strongest shell basis: existing static, modal, transient and harmonic V&V paths need an explicit frozen replay and Owner decision, not a new formulation. MITC3 static has a similar narrow path.
3. MITC3 modal, transient and harmonic lack the comparable independent decks stated by the inherited record, so they are major V&V rather than a maturity relabel.
4. WEDGE6 static remains major: its 0.2.7 Owner decision explicitly kept it experimental. The existing Code_Aster PENTA6 affine same-mesh policy is reusable only within its declared 1e-6 scope; refinement and distortion need separately approved pre-execution tolerances.
5. HEX8 buckling remains not qualified. The prior diagnostic evidence is not a closure: it needs a three-level refinement, comparable independent oracle, replayed eigenpair evidence and Owner decision.

The detailed candidate records in the machine-readable matrix name the exact test file to add, oracle/reference, tolerance policy, exit gate and risk. No candidate changes public maturity until its stated gate passes.

WP03 BEAM2 and DISCRETE V&V

The completed technical campaign for BEAM2 and DISCRETE static/modal routes is documented in WP03 V&V. It records the frozen analytical oracles, tolerances, invariants, two replays and bounded decisions. The four dynamic routes were intentionally not decided by that WP03A record; their separate bounded analytical campaign is documented in WP03B dynamic V&V. WP03B preserves every WP03A static/modal decision. The resulting bounded maturity decisions are recorded by the final WP03 Owner gate: seven routes are QUALIFIED_BOUNDED; DISCRETE Newmark remains EXPERIMENTAL.

WP04 MITC3 and MITC4 V&V

The WP04 frozen contract, executable replays and bounded decision delta are recorded in WP04 MITC V&V and its machine-readable evidence and WP04 maturity matrix. The final decisions are recorded in the WP04 Owner gate and machine-readable Owner record. MITC3 static, MITC4 static, MITC4 Newmark and MITC4 harmonic are technical QUALIFIED_BOUNDED within their bounded scopes. MITC3 modal/Newmark/harmonic remain EXPERIMENTAL because no executable independent shell oracle is present. MITC4 modal also remains EXPERIMENTAL: its deterministic replay misses the frozen relative-residual gate, and that tolerance was not relaxed. Across the 46 source combinations: 32 are QUALIFIED_BOUNDED, 13 are EXPERIMENTAL, and COMB-HEX8-linear_buckling is the sole NOT_QUALIFIED combination. WEDGE6 static is distinct and is qualified only within the bounded scope recorded by WP05.

Explicit boundaries

The eight nonlinear combination records remain KEEP_EXPERIMENTAL: this WP does not expand nonlinear scope, finite-kinematic J2, friction, GPU or distributed nonlinear execution. No mixed-mesh, PYRAMID5, HEX8R, SRI, B-bar or hourglass production claim is created by WP01.

WP05 WEDGE6 static V&V

WP05 closes the technical V&V campaign for the WEDGE6 linear-static candidate identified by WP01 as NEEDS_MAJOR_VNV. The bounded technical decision, predeclared contract, current replays, face-load checks, geometry failure paths and the audited Code_Aster evidence are recorded in WP05 WEDGE6 V&V, its machine-readable evidence and maturity delta. The decision is QUALIFIED_BOUNDED for the exact static scope after the separate Owner gate; the public maturity is now QUALIFIED_BOUNDED within that scope, with the limitations recorded in the WP05 Owner gate. The machine-readable Owner record is wp05_owner_gate_final.json. Modal, Newmark and harmonic WEDGE6 routes are not promoted.

WP06 HEX8 linear-buckling assessment

WP06 audits the sole NOT_QUALIFIED combination, COMB-HEX8-linear_buckling, using the frozen HEX8 buckling contract and executable campaign evidence. The technical decision remains NOT_QUALIFIED: the Euler factor and three-level refinement gates failed in both declared boundary configurations, the invalid-orientation case did not fail closed, current CalculiX execution was unavailable, and the two complete replay digests differed. The passing preload, Kg, eigen-residual and mode-shape sub-checks do not override those failed closure gates.

The WP06 maturity delta records no promotion and reconciles the unchanged global state as 32 QUALIFIED_BOUNDED, 13 EXPERIMENTAL and one NOT_QUALIFIED, exactly COMB-HEX8-linear_buckling. WEDGE6 static remains distinct. No 0.2.7 evidence or numerical source was changed, and no Owner promotion gate is requested while the technical decision is negative.

WP06B HEX8 buckling root-cause remediation

WP06B starts from be8fbd99d8a3d3b8e4ee3b8d867842389b063061 and reuses the WP06 scope and tolerances without retuning. The WP06B contract classifies the Euler/refinement failures as model/scope limitations, the original invalid-orientation probe as a test-oracle problem, and the replay tail variation as sparse-solver start nondeterminism. The minimal remediation fixes only the latter two test/replay issues: a fixed ARPACK start vector is used, and a true mirrored HEX8 connectivity is tested.

After remediation, invalid geometry, robustness and two complete deterministic replays pass. The Euler factor and mesh-refinement gates still fail in the unchanged scope, and the current CalculiX oracle remains unavailable. The WP06B evidence and WP06B maturity delta therefore retain COMB-HEX8-linear_buckling as NOT_QUALIFIED; no public promotion or Owner gate is applied. No 0.2.7 evidence is changed and WP07 is not started.

WP07 mixed TET4/WEDGE6/HEX8 linear static

WP07 audits the existing family-generic architecture for one conforming mixed-solid linear_static workflow. The frozen WP07 contract, executable WP07 evidence and WP07 candidate matrix record the subsystem audit, affine/interface/load gates, Gmsh path, explicit failure contract and two deterministic replays.

All declared technical gates pass, producing QUALIFIED_BOUNDED_CANDIDATE for the exact conforming TET4/WEDGE6/HEX8 linear-static scope. The separate WP07 Owner gate approves that workflow as QUALIFIED_BOUNDED with limitations; the corresponding Owner record keeps it separate from the 46 element-analysis combinations. The affine refinement is a consistency check, not a general accuracy-convergence claim. Dynamic and other analysis routes remain out of scope, no external solver correlation is claimed, and no 0.2.7 evidence or prior WP01-WP06B record is changed.

WP08 mixed TET4/WEDGE6/HEX8 modal

WP08 audits the existing mixed consistent-mass modal route separately from the 46 element-analysis combinations and from the WP07 static workflow. The frozen WP08 contract, machine-readable WP08 evidence and WP08 matrix cover conforming fixed-base TET4/WEDGE6/HEX8 pairwise and three-family cases, the first six positive modes, an independent dense local-matrix oracle, mass conservation, residuals, orthogonality, interface failure paths and two deterministic replays.

The mixed cases and the independent modal oracle pass. The first-frequency refinement trend is finite and positive, but the predeclared shared-node mode-matching gate fails on the first refinement transition (MAC min ≈ 0.103 against a 0.5 gate). WP08 therefore records SUPPORTED_WITH_LIMITATIONS, applies no public promotion and requests no Owner gate. No external industrial correlation is claimed, no modal stress claim is made, and no 0.2.7 evidence or WP07 static claim is changed.

WP08B mixed modal refinement follow-up

The WP08B root-cause audit shows that the failed WP08 refinement compared changing mechanical domains: only the first WEDGE6/HEX8 segment was retained beyond level one. WP08B leaves that record and its 0.5 gate unchanged, then uses a new predeclared contract with the full invariant physical chain and a fine-mass mapped MAC comparator.

All new mixed gates pass: volume/bounds are invariant, the first six modes are globally matched without permutations, and the minimum mapped MAC is 0.9037. The resulting WP08B evidence and WP08B matrix record QUALIFIED_BOUNDED_CANDIDATE only. It remains a separate mixed-workflow candidate, with no public maturity relabel or change to the 46 combination records, pending a dedicated Owner gate.

The dedicated WP08B Owner gate approves the workflow as QUALIFIED_BOUNDED with limitations. The qualification is limited to modal linear analysis of the tested fixed-base conforming TET4/WEDGE6/HEX8 chain, consistent translational mass, and the first six positive modes at levels 1/2/4/8. Shape-function prolongation and the fine consistent-mass MAC are the valid cross-mesh comparison; the original WP08 shared-node MAC gate and its failure remain unchanged. No external correlation is claimed, and no modal stress, Newmark, harmonic, nonlinear, contact, nonconforming, higher- order-family or large-performance claim is created. The Owner decision is recorded in the WP08B Owner record, without changing the 46 element-analysis combinations or starting WP09.

WP09 PYRAMID5 feasibility gate

WP09 records a bounded internal feasibility gate for a five-node linear pyramid. The frozen formulation, contract, evidence, and decision matrix cover the collapsed-coordinate kernel, bounded geometry validation, affine patch, loads, post-processing, Gmsh type-7 import, a conforming HEX8/PYRAMID5/TET4 transition and two deterministic replays.

The decision is FEASIBLE_CONTINUE, not a qualification or maturity promotion. PYRAMID5 remains absent from the public compatibility descriptor and the 46-combination element-analysis registry, so public dispatch remains fail-closed. No coherent h-refinement or external-solver correlation is claimed. WP07/WP08 scopes remain TET4/WEDGE6/HEX8 only, and WP10 is not started.

WP10 HEX8 selective reduced integration research gate

WP10 evaluates a separate HEX8 SRI research kernel under the frozen formulation and contract. The machine-readable evidence and matrix compare the unchanged standard HEX8 with a deviatoric 2×2×2 plus volumetric centre integration variant.

The bounded campaign records measurable fine-level displacement-error reduction on both declared bending families, with passing rank, rigid-body, symmetry, energy, geometry and replay gates. The result is EXPERIMENTAL_BOUNDED_CANDIDATE, not QUALIFIED_BOUNDED. A diagnostic affine nodal-force action difference remains, so no general patch-force or production accuracy claim is made. The Owner gate records APPROVE_EXPERIMENTAL_BOUNDED as a separate documented research capability; SRI is not added to the 46 element-analysis combinations and does not replace standard HEX8. HEX8 standard and all prior WP01–WP09 records remain unchanged. No WP11 large mixed workflow is started by this WP10 record.

WP11 mixed large-scale evidence

WP11 records a mandatory 300,003-DOF mixed linear_static run using equal counts of TET4, WEDGE6 and HEX8 in repeated conforming chains. The small baseline, two large replays, residual/equilibrium/energy gates and environment metadata pass. See the WP11 evidence report, the frozen contract, the machine-readable evidence and the decision matrix.

The bounded claim is only recorded large-scale mixed TET4/WEDGE6/HEX8 evidence for this specific conforming linear-static route and environment. The repeated chain topology is not an arbitrary connected industrial mesh. No accuracy, production, modal, dynamic, nonlinear, PYRAMID5, HEX8-SRI or universal scalability claim is made. The optional 1M and 3M targets were not run and remain non-blockers; no 0.2.7 evidence or prior qualification record was changed.

WP11B HEX8 buckling experimental-readiness gate

WP11B adds a separate pre-publication experimental-readiness record for HEX8 linear eigenvalue buckling. Its frozen benchmark is a clamped solid cantilever block under uniform end-face compression, not the historical pinned Euler model. The three declared refinements produce finite positive factors with UZ-dominant modes, passing prestress, geometric-stiffness, eigen-residual, scaling, robustness and replay checks. Refinement is characterized honestly and is not claimed to converge to Euler.

The result is EXPERIMENTAL_CANDIDATE, proposing EXPERIMENTAL exposure only after a separate Owner gate. The historical WP06 Euler and mesh-convergence FAIL results remain unchanged, no external correlation is claimed, and no 0.2.7 evidence or numerical source was modified. See the WP11B report, contract, evidence and matrix.

WP11B HEX8 buckling Owner gate

The WP11B Owner gate approves the separate HEX8 linear-buckling route as EXPERIMENTAL, not QUALIFIED_BOUNDED. The bounded scope is linear-elastic prestress, valid clamped cantilever solid geometries, the first positive mode and exploratory or research use only. The solid-block benchmark passes physical scaling, mode, residual, robustness and deterministic replay checks.

The refinement factors remain a characterization (41.9397, 21.5200, 17.8649) and are not presented as general convergence or Euler evidence. The WP06/WP06B Euler and convergence FAIL records, and the unavailable external oracle, remain explicit. The 46 element-analysis records now reconcile to 32 QUALIFIED_BOUNDED, 14 EXPERIMENTAL, 0 NOT_QUALIFIED; only COMB-HEX8-linear_buckling transitioned. No numerical source or 0.2.7 evidence was changed, and no WP12 work is started.