LU2-WP08: Mixed, WEDGE15, PYRAMID5 and HEX8 decisions

This is a governance decision record, not an implementation or qualification record. It is based on the repository state at 8ef34e345f970879548a4dfdce4ac5ba32c11bda. The machine-readable source is lu2_wp08_decision_matrix.json, and the state record is lu2_wp08_state.json.

Decision summary

Axis	Observed state	LU2 decision	Priority	Public boundary
Mixed TET/WEDGE/HEX	PARTIAL technical paths, no end-to-end qualification	DEFER	P1	Not a qualified mixed route
WEDGE15	NOT_SUPPORTED	DEFER	P2	No active capability
PYRAMID5	NOT_SUPPORTED	DEFER	P3	Negative fixtures are not implementation
Existing HEX8	SUPPORTED_WITH_LIMITATIONS	KEEP_EXISTING_ROUTE	P1	Bounded route and WP19 diagnostic only
HEX8R	NOT_EVALUATED	RESEARCH_ONLY	P2	No production claim
SRI	NOT_EVALUATED	RESEARCH_ONLY	P2	No production claim
B-bar	NOT_EVALUATED	RESEARCH_ONLY	P2	No production claim
Hourglass control	Not applicable to current full integration	DEFER	P3	No reduced-integration claim

Mixed meshes

The Gmsh importer recognizes a bounded set of same-dimensional solid families, and individual element routes exist. That is not an end-to-end mixed-mesh qualification. Interface assembly, shared material and DOF contracts, boundary-face selection, load transfer, post-processing and cross-family equilibrium need a dedicated contract and V&V corpus. The LU2 decision is therefore DEFER, rather than IMPLEMENT_NOW or TARGETED_RESEARCH.

Candidate new solids

WEDGE15 has a plausible industrial value for quadratic prismatic meshes, but would require a new kernel, high-order face/load mapping, quadrature, quality, mass/modal and independent external evidence. PYRAMID5 is useful mainly when transition meshes demonstrate a concrete need; its apex quality and interpolation risks make it a larger V&V commitment than its current evidence justifies. Neither family has an active descriptor, importer path or registry capability here. Both are deferred beyond LU2.

Future evidence must include shape-function identities, Jacobian/orientation, rigid-body modes, affine/patch tests, rank and energy checks, distortion, loads/faces, deterministic replay and formulation-compatible external results.

HEX8 next generation

WP19 remains authoritative for the existing bounded HEX8 route. Its compatible C3D8 displacement study is a diagnostic and does not prove locking, universal accuracy, or a replacement formulation. No production HEX8R, SRI or B-bar route was evaluated. Those alternatives are RESEARCH_ONLY; a future prototype must keep a separate route, prove its formulation and compare locking benefit against distortion, rank, energy, patch and backward-compatibility risks. Hourglass control is subordinate to a reduced-integration decision and is deferred with it.

Scope and next steps

No element, formulation, mixed-mesh framework or large benchmark was added in this WP. Existing public maturity is unchanged. The best future element investment is WEDGE15 after interoperability and V&V foundations are funded; the lowest-ROI standalone LU2 item is PYRAMID5. These are prioritization decisions, not implementation commitments.

The gate closes as PASS_WITH_LIMITATIONS: the required decisions and future V&V boundaries are explicit, while deferred and research routes remain absent from the active qualified registry. WP04's independent supervised retry remains the next operational action; it is not reclassified by this document.