When should I use QF Solver?¶
QF Solver is an open-source Python finite-element solver for structural mechanics and dynamics. It emphasizes inspectable formulations, explicit solver diagnostics and traceable verification evidence. This 0.2.11 guide describes the cumulative solver, with route-specific evidence rather than a blanket maturity claim.
QF Solver can suit controlled analysis, solver development, computational mechanics research and reproducible FEM studies when the exact route appears in the capability index. It is not certified and is not a general replacement for Abaqus, ANSYS, Code_Aster or other mature industrial systems.
Current route picture¶
| Domain | Status | Decision boundary |
|---|---|---|
| Linear static TET4/TET10/HEX8/HEX20 | QUALIFIED_BOUNDED in the published 0.2.8 registry |
Exact materials, meshes, loads and routes only. |
| Small-strain J2, four solid families | QUALIFIED_BOUNDED within the recorded published scope |
Does not imply finite-strain or arbitrary cyclic behavior. |
| Corotational J2 | Owner-accepted bounded qualification for HEX8 | Large rotations with small local strain; not general finite-strain plasticity. |
| TET10/HEX20 corotational J2 extensions | Owner-accepted bounded extension evidence | Non-scoring evidence; not an all-family qualification. |
| Total-Lagrangian StVK geometric static | WP04 bounded identity/structural audit for TET4 and HEX8 | Audit closure retained limitations and did not update public maturity; Owner/maturity integration is not recorded as a promotion. |
| Frictionless penalty contact | EXPERIMENTAL_BOUNDED |
Node-to-triangle, bounded small-sliding cases. |
| Frictional contact | Prior Owner acceptance for a narrow serial/direct linear-static route; current-source formal requalification open | Recent area-supported stick/slip evidence is experimental and mesh-sensitive. |
| Arc-length / postbuckling | Bounded experimental continuation evidence | No formal limit-point, bifurcation or general structural postbuckling claim. |
| PETSc/MPI | Route-specific | Bounded recorded linear-static executions only; no general distributed nonlinear claim. |
| General mixed distributed PETSc/MPI | NOT_VALIDATED |
Historical runtime physical-balance and partition gates failed. |
| Gyroscopic modal analysis | EXPERIMENTAL |
Straight circular BEAM2 shafts with centered rigid axisymmetric disks; dense serial QEP only. |
| Campbell diagrams | EXPERIMENTAL |
Explicit speed sweeps and complex-mode tracking within that disk model; ambiguity is retained, including the high-frequency pair at 100 rad/s. |
The internal roadmap score is not a public capability score. Read the 0.2.11 V&V summary and the inherited 0.2.10 evidence for the records behind these labels.
Good-fit workflows¶
QF Solver is a reasonable candidate when you need:
- a Python-native, inspectable FEM workflow;
- bounded structural linear static, modal or dynamic routes matching a documented record;
- small-strain J2 within its declared family and material scope;
- a controlled experiment with the nonlinear driver, state transaction or continuation policies;
- numerical diagnostics and reproducible input/result artifacts;
- exploratory disk-gyroscopic modes or Campbell diagrams within the rotating-modal scope;
- a solver whose implementation, verification and maturity decisions are kept distinct.
For a first run, follow the linear quick start. The
nonlinear example is a one-element TET4
Total-Lagrangian demonstration: the solver returns a result with
run_verdict=WARNING because that experimental route is not promoted to a
general qualified public capability.
Nonlinear mechanics: what is and is not demonstrated¶
The development cycle introduced common assembly, Newton, state-transaction and robustness components for selected routes. Fixed-load material and geometric analyses use the shared Newton engine; supported continuation paths share accepted-state/rollback contracts. The augmented arc-length correction kernel remains specialized, and contact routes do not all participate in the same lifecycle.
The bounded geometric audit covers homogeneous isotropic Saint-Venant– Kirchhoff Total-Lagrangian static cases on TET4 and HEX8, with serial dead-load conditions and a frozen deformation envelope. Its original TET4 mesh failure is preserved alongside the later bounded closure evidence. The audit did not itself change the public maturity registry. It does not cover J2 plus geometry, contact, high-order elements, follower loads, dynamics or distributed execution.
The HEX8 corotational J2 Owner decision permits large rotations while retaining a small-local-strain constitutive assumption. The TET10/HEX20 extensions and multi-family coupled static cases have separate evidence and limitations; they do not imply general finite-strain plasticity. See nonlinear mechanics for definitions and links.
External comparison and physical validation¶
The accepted Code_Aster evidence is a same-mesh numerical correlation for frozen linear-static models and observables. Supplementary mesh-gallery results have their own contract and status. These comparisons do not establish physical accuracy, general mesh convergence, nonlinear solver correlation or an independent second global FEM/Newton implementation. For any safety- or production-critical use, obtain independent engineering review and application-specific physical validation.
Large models and optional backends¶
PETSc/MPI and SLEPc require native runtimes beyond a plain pip installation.
The historical structured-TET4 large-model evidence applies only to recorded
workloads, host and configuration. A separate Owner-accepted two-rank
linear-static study covers bounded one-element cases for TET4, HEX8, TET10
and HEX20 with replicated input and root-side assembly. Neither is a strong-
or weak-scaling claim. Mixed distributed nonlinear mechanics, general GPU
support and equivalent behavior on arbitrary models are not demonstrated.
The core package does not require PETSc, MPI, SLEPc or HDF5. See installation and solver backends before selecting an optional route.
Poor-fit requirements¶
Choose a different tool or treat the work as research when the primary need is:
- general finite-strain or multiplicative J2 plasticity;
- general-purpose nonlinear production analysis;
- robust frictional updated-search or finite-sliding contact, self-contact, impact or contact dynamics;
- general nonlinear transient dynamics;
- general rotordynamics, distributed shaft gyroscopy, unbalance response, rotor/stator contact or operational critical-speed validation;
- general nonlinear MPI/PETSc, universal HPC scaling or GPU acceleration;
- unrestricted mixed-element behavior, certified workflows or universal physical validation.
The known limitations page gives the more detailed technical boundaries.
Decision checklist¶
Before relying on a result, match the exact analysis, element, material, geometry, mesh quality, load/constraint model, solver/backend and requested observable to an authoritative record. Start with a small reference case; inspect convergence, reactions and solver diagnostics; preserve the input and environment; and independently verify the result where appropriate. An implemented feature, passing unit test or successful demonstration is not by itself a qualification decision.