QF Solver

An inspectable Python finite-element solver for structural mechanics.

QF Solver provides white-box finite-element formulations, numerical diagnostics and reproducible verification evidence. Support is always scoped by element, analysis, material, mesh, loading and solver route. A passing example is not a universal qualification.

Capabilities

The active combination matrix is in docs/verification/0_2_7/0_2_7_capability_matrix.md. It is the authoritative guide to what a particular combination means.

Installation

For the stable source release:

git clone https://github.com/emptiesvoid-cloud/QF_solver.git
Set-Location QF_solver
git checkout v0.2.7
python -m pip install .
qf-solver --version

When the package is available from the package index, the equivalent user installation is:

python -m pip install qf-solver
qf-solver --version

Optional development and integration extras are documented in docs/getting-started/installation.md. PETSc, MPI and SLEPc are optional integrations and are not required for the core import or the standard small examples.

Quick start: CLI

From the repository root, run the maintained TET4 example:

qf-solver check-mesh --input .\examples\tet4_static.json
qf-solver solve --input .\examples\tet4_static.json --output .\results\tet4.json

The full first-calculation guide is docs/getting-started/quickstart.md.

Quick start: Python

Use the public qf_solver namespace:

from qf_solver import check_mesh, load_model, save_result, solve_model

model = load_model("examples/tet4_static.json")
check_mesh(model)
result = solve_model(model)
save_result(result, "results/tet4.json")

The historical solveur namespace remains available for compatibility. New applications should use qf_solver; see docs/reference/api_stability.md.

Elements and analyses

The public element summary is in docs/elements/index.md, and the analysis summary is in docs/analyses/index.md.

The current release includes bounded routes for TET4, TET10, HEX8 and HEX20, along with case-bounded shell, beam and discrete paths. The status is explicit: WEDGE6 static remains experimental. WEDGE6 modal qualification does not transfer to static, nonlinear or other dynamic analyses.

Measured performance

The published performance evidence is bounded, not a universal scaling law:

No claim of GPU, general HPC, hardware-independent scaling, mixed-mesh support or a general nonlinear performance claim is made.

Main limitations

• WEDGE6 static is experimental; WEDGE15 and PYRAMID5 are not supported.
• Mixed TET/WEDGE/HEX workflows and next-generation HEX8R/SRI/B-bar paths are deferred or research-only.
• Finite-kinematic J2, generalized nonlinear, contact and finite-sliding routes remain experimental or outside the qualified scope.
• 5M Gold and deeper 10M scaling analysis are deferred.
• Code_Aster correlation is bounded to comparable recorded cases. CalculiX is NOT_COMPARABLE where conventions or observables do not match strictly.
• Linux and Windows evidence is available in the recorded test matrix; macOS and some Python versions remain unverified and are not claimed as tested.

Documentation and verification

• Getting started
• Elements
• Analyses
• Solvers and backends
• Public roadmap
• 0.2.7 verification summary
• API stability
• Detailed changelog

QF Solver distinguishes IMPLEMENTED, TESTED, VERIFIED, EXTERNALLY_VALIDATED, QUALIFIED and EXPERIMENTAL. The detailed evidence pack preserves the exact inputs, outputs, manifests and source references used for each recorded result.

No claim of certification or universal physical validation is made.

Contributing, license and citation

Development setup and quality checks are described in CONTRIBUTING.md. The software is distributed under the Apache License 2.0; documentation and original examples are under CC BY 4.0. Third-party terms are listed in THIRD_PARTY_LICENSES.md. See CITATION.cff for citation metadata.

Documentation contributors can build the controlled evidence locally with:

python .\scripts\build_docs.py --profile engineering
python .\scripts\build_technical_latex.py