Files
OneCAD-Simulation/docs/parity.md
T
40f98f0af9 Implicit solver: fem kernels (tet4/10, hex8, hex20), sparse Cholesky, block Lanczos, solve.* commands (M4) (#23)
* Renumber the layer table for solver crates: fem at L1, solver at L2, fsi at L3

L0 model -> L1 geometry, mesh, fem, decks, render -> L2 solver (explicit, lbm later)
-> L3 fsi (later) -> L4 doc -> L5 io -> L6 engine -> L7 ui-egui, mcp -> apps.
UI_MIN_LAYER moves to 7, the testkit sits above it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm

* modelcraft-fem: element kernels for tet4, tet10, hex8 (full, incompatible modes), hex20 (full, reduced)

Shape functions, Gauss and conical-product quadrature, stiffness, consistent and HRZ-lumped
mass, body forces, consistent pressure loads on faces, integration-point stresses and their
least-squares extrapolation to nodes, isotropic and orthotropic elasticity. The element table is
one row per kind, so wedges, shells and beams are added locally.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm

* modelcraft-solver: linear static and frequency analysis with an analytic verification suite

Assembly from the resolved model on the node graph, boundary conditions by elimination,
consistent pressure, gravity and body-force loads, nested-dissection ordering, sparse up-looking
Cholesky with a factor-size cap, reactions, integration-point and nodal-averaged stresses, von
Mises; shift-invert block Lanczos with full reorthogonalization for natural modes, participation
and effective mass. ADR 0001 records the own-linear-algebra decision (determinism, wasm).

Verified against closed forms: displacement and traction patch tests for every element, cantilever
tip deflection (Timoshenko) with convergence rates, Lamé cylinder, Kirsch hole (Kt -> 3),
cantilever and free-free beam frequencies with six rigid-body modes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm

* Engine: solve.static, solve.frequency and solve.probe; results kept in the session

Results are stored per step, restored on a failed command and cleared when the model changes.
Caps (degrees of freedom, factor entries, Lanczos vectors) may only be lowered; hostile parameters
are tested against a solvable session.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm

* docs: implicit solver architecture, roadmap M4, parity and references

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm

* Simplify the solver: shared face integration, block assembly, cached extrapolation

- model: one face integrator (nodal::face_normal_weights) for tractions and pressures
- fem: mid-side tables from ElementKind::edges, orthotropic via inverse3, symmetric
  accumulation without a symmetrize pass, incompatible-mode recovery without K_uu, model's
  coordinate check, dead table rows removed
- solver: one pattern search per node pair (block, scalar and diagonal adds), extrapolation built
  once per element type, reactions from the prescribed rows only, matrices dropped after
  restriction, the free-body shift keyed on the factorization error, step selection in the solver,
  Error::Element naming the element
- engine: the solver owns the default step and the shift check; the model comparison runs once

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm

* docs: results are stamped with the session revision

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NmAttHhUZTokgPeEJia8sm

---------

Co-authored-by: jyviko <2501648+jyviko@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-13 17:44:21 -04:00

9.0 KiB
Raw Blame History

Capability checklist (reference: Abaqus/CAE)

What an analyst does in a full CAE pre-processor, grouped by milestone, with ModelCraft's status. Scope comes from the public documentation of Abaqus/CAE and public analyst discussion, described in our own words; nothing was taken from the product itself (AGENTS.md: no commercial products are run, no documentation text is copied). Status: done, started, planned. HEX marks items centred on hexahedral meshing, the core goal.

Where we aim to be better

# Opportunity Why analysts want it ModelCraft approach
1 Hex meshes without manual partitioning Partitioning a part into sweepable cells by hand is the most cited chore Sweep and mapped meshing with automatic decomposition that proposes SolidCraft splits (M5); octree all-hex for anything else (M6)
2 Undo, redo, replay, crash recovery No general undo; replay files are hard to reuse; corrupted model files Every action is a journalled command with rollback, undo and redo; autosave with recovery after a crash
3 Sets that survive changes Sets and loads break when geometry regenerates or is remeshed Sets bind to SolidCraft's persistent face, edge and body names and resolve after meshing (M3)
4 An API that is not a bundled scripting runtime Version breakage, slow scripts, geometry picked by position JSON commands over CLI, control channel and MCP
5 Located diagnostics Generic error messages Errors name the element, face or keyword; mesh.quality lists failing elements
6 Lossless deck round trip Unsupported keywords dropped on import Deck readers keep unknown keywords and write them back (M9)
7 Safe, diffable model files Binary model databases that corrupt or lock to a version .modelcraft: text, line-diffable definition and journal, checksummed binary mesh, versioned format tag, atomic saves with backup, autosave (M3, done)
8 Geometry checked before meshing Slivers and tiny faces fail late Surface preparation reports defects with their location (M1)
9 No size caps, any solver Node limits in learning editions; one solver per tool Free, open source, three deck formats
10 Clear step history Loads and BCs propagate between steps implicitly An explicit step × load/BC table (M3)

Checklist

Area Capability Milestone HEX Status
Geometry Open SolidCraft designs, bodies, persistent face names, assemblies placed M1 done
Geometry Face-tagged mesh import (3MF) M1 done
Geometry Weld, closed/manifold/orientation and self-intersection checks, feature edges M1 yes started
Geometry Partitions and datums as SolidCraft features (split body/face, construction planes) M5 yes planned
Geometry Midsurface extraction for shells M8 planned
Mesh Size field: global, per face and edge seeds, bias, curvature and proximity refinement M2 yes planned
Mesh Free tet meshing (tet4, tet10 with mid-side nodes on the geometry) M2 started
Mesh Quality metrics, certified validity, verify per set M1–M2 yes done (metrics, gate)
Mesh 2D quad meshing of faces M5 yes planned
Mesh Mapped hex of six-sided regions M5 yes planned
Mesh Sweep hex (extrusions, revolutions) M5 yes planned
Mesh Automatic decomposition into sweepable and mappable regions M5 yes planned
Mesh Conforming interfaces, tet–hex transition with pyramids M5 yes planned
Mesh Octree/grid-dual all-hex for any part M6 yes planned
Mesh Hex-dominant by recombination; polycube M6 yes planned
Mesh Smoothing, untangling, boundary padding, layers through thin walls M6 yes planned
Mesh Mesh editing (move, merge, split, collapse, remesh region), imported orphan meshes M6 yes started (node/element add/remove)
Model Sets and surfaces bound to geometry names M3 yes done
Model Materials: elastic (isotropic, orthotropic), density, unit-system check M3 done
Model Materials: plasticity, hyperelasticity, damage, thermal M8 planned
Model Sections: solid, shell, beam profiles; assignment by set M3 done
Model Composite layups, material orientations, continuum shells M8 yes planned
Model Element formulation per region (integration, incompatible modes, hybrid) mapped per solver M3 yes done (model; per-solver mapping with the deck writers)
Model Boundary conditions: fixed, displacement, symmetry M3 done
Model Loads: force and moment, pressure, gravity, body force M3 done
Model Reference points, kinematic and distributing coupling M3 done
Model Steps: linear static, frequency; output presets M3 done
Model Steps: nonlinear static, explicit dynamics, heat transfer; amplitudes M7–M8 planned
Interaction Tie between non-matching meshes M7 planned
Interaction Contact (general and pairs), friction, automatic pair detection M7 planned
Interaction Bolt pretension, springs, point masses, connectors M7 planned
Decks Abaqus input writer (deck.write, flat deck; reports what Abaqus cannot express) M3 done
Decks Nastran bulk data writer (deck.write, SOL 101 or 103 per deck; reports what Nastran cannot express) M3 done
Decks LS-DYNA keyword writer (implicit; explicit in M8) M3 done (implicit static and eigenvalue; beam sections planned)
Decks Readers keeping unknown keywords M9 planned
Solve Own linear static and modal solver, verified against analytic solutions M4 yes started (solid elements: tet4, tet10, hex8, hex20)
Results Deformed shape, contours, mode shapes, probes, reaction sums M4 started (results in the session, solve.probe, reaction sums; display planned)
Results XY and path plots, cuts, display groups, free-body forces, animation M9 planned
Results Readers for public result formats (Nastran OP2, LS-DYNA d3plot, Abaqus .fil) M9 planned
Results Mesh convergence studies M9 planned
App Desktop viewport, model tree, commands, control channel continuous done
App Undo/redo on the command journal, autosave, .modelcraft file M3 done
App MCP server, browser build continuous planned

Deck notes for the writers

  • Node order: the model stores Abaqus/VTK order. Nastran orders the mid-side nodes of 20-node hexes and 15-node wedges differently; LS-DYNA takes tets and wedges as degenerate hexes and triangles as degenerate quads. Every mapping is tested by the sign of the element Jacobian after writing.
  • Each format's field widths, continuation rules, number forms and label limits are respected by the writer and checked by golden-file tests on our own examples.
  • Abaqus: the model's node order and face numbers are Abaqus's, so connectivity passes through; element types come from the element kind, the section and its formulation (table in crates/decks/src/abaqus.rs). The step table is bridged to Abaqus's propagation: zero-valued boundary conditions active in every step go to the initial state, and steps restate their boundary conditions and loads with OP=NEW when the active set changes.
  • Nastran (modelcraft_decks::nastran): entries with a real use large (16-column) fields, the rest small fields; a real is exact when its shortest round-trip form fits 16 columns, else rounded to at least ten significant digits. Linear solids map full integration to PSOLID TWO/THREE + FULL, hex20 reduced to TWO + FULL, hex8 incompatible modes to BUBBLE; one-point reduced linear solids, hybrid solids, reduced-integration or one-point shells, body forces, gravity on part of the model, orthotropic beams and rods, and orthotropic shells off planes of constant Z are reported as unsupported. Static and frequency steps need separate decks (deck.write steps).
  • LS-DYNA (modelcraft_decks::lsdyna, deck.write {"format": "lsdyna"}): long format (20-column fields; every value with up to 13 significant digits, and 17-digit values from 0.01 to 1e17, are written exactly; the rest is rounded to at least 13 digits and counted). The 10-node tetrahedron follows the manual's figure: mid-side nodes 8, 9, 10 on edges 2-4, 3-4, 1-4 (the model has them on 1-4, 2-4, 3-4). Incompatible modes map to ELFORM −18 (eight-point enhanced strain). Several steps are written as *CASE analyses. Refused with the item named: beam sections (not yet), body forces per volume, partial-DOF kinematic couplings, constraints on a coupled reference point, gravity on different sets or on parts of sections, hybrid other than tet4, reduced quadratic elements, materials without density, more than 10 shell integration points.
  • Open item (LS-DYNA tet10 node order): the writer follows the manual's figure (mid-side nodes 8, 9, 10 on edges 2-4, 3-4, 1-4), which disagrees with an earlier reading as the Abaqus order. It needs one confirmation run by a licensed user; agents never run commercial solvers.