Skip to content

Known limitations

These limitations apply to the cumulative QF Solver 0.2.11 surface. Every result is bounded by its element family, formulation, mesh, loads, boundary conditions, constitutive model, solver, and evidence decision. The limitations below are technical boundaries, not a complete list of input checks.

Nonlinear mechanics

  • A shared nonlinear driver, residual/tangent interface, and state transaction do not make every material/geometry/contact/backend combination available or qualified. Some contact active-set/recovery loops and arc-length correction remain specialized.
  • General finite-strain multiplicative plasticity is not claimed. The accepted corotational J2 route supports large rotations with a small-local-strain assumption and is bounded to HEX8; it is not a general finite-strain model.
  • The Total-Lagrangian St. Venant–Kirchhoff audit covers selected static serial TET4/HEX8 cases and records GO_WITH_LIMITATIONS, not a maturity promotion. It excludes contact, coupled J2/geometry, dynamics, MPI/PETSc, high-order families, follower loads, and reduced/hourglass HEX8 routes.
  • Owner-accepted coupled evidence covers bounded static cases for TET4, HEX8, HEX20 and TET10. It does not cover friction, finite sliding, dynamics, MPI/PETSc, Code_Aster correlation, or an independent global FEM/Newton solve.
  • Adaptive stepping, cutback/retry, line search and arc-length are convergence policies, not guarantees for arbitrary models. Current arc-length evidence does not establish general limit-point tracking, bifurcation, or postbuckling. General nonlinear transient dynamics is not covered.
  • Checkpoint/restart and deterministic state digests apply only to the routes named in their records. Do not infer frictional-contact, distributed, or nonlinear-dynamics restart support from shared transaction code.

Rotating modal and Campbell

  • The rotating capability is experimental and restricted to serial dense gyroscopic modal analysis of straight, collinear, circular-isotropic BEAM2 shafts with centered rigid axisymmetric disks, constant signed spin, a fixed global axis, and an undamped, unprestressed small-perturbation model.
  • Campbell diagrams only orchestrate single-speed modal solves and track their complex modes. They do not calculate forced response, unbalance amplitude, operational risk, instability, or validated critical speeds.
  • At 100 rad/s, the high-frequency pair remains ambiguous. The tracker keeps the gap; it must not force continuity or interpolate across the ambiguity.
  • GYRO-06 is internal mesh-convergence evidence, not independent physical validation. Both analysis maturities remain EXPERIMENTAL; numerical PASS does not itself qualify the capability.
  • Distributed shaft gyros, variable speed, speed-dependent stiffness/mass, general damping, bearings, centrifugal stiffening, contact/rubbing, nonlinear rotors, forced unbalance response, PETSc/SLEPc and MPI are outside this scope.

Contact

  • Frictionless penalty node-to-triangle contact is experimental and bounded; it is not a general contact formulation.
  • Frictional stick/slip has prior Owner-accepted bounded serial evidence for a narrow route, but formal requalification of the current source is not established. The accepted studies report mesh sensitivity; they do not establish mesh convergence.
  • General updated search, finite sliding, self-contact, and impact/contact dynamics are not established by the available evidence.

Elements, dynamics, and backends

  • Nonlinear evidence is narrower than linear registry availability, especially for TET10/HEX20 and other higher-order routes. A linear element record does not imply a nonlinear capability.
  • Modal, Newmark, harmonic, and linear-buckling boundaries are route-dependent. Mixed Newmark/harmonic records are experimental and do not establish general nonlinear dynamics.
  • PETSc/MPI Owner-accepted evidence is limited to two-rank linear-static one-element cases with replicated input and root-side assembly. No strong- or weak-scaling, distributed assembly, nonlinear MPI, contact, or dynamics claim follows. Structured-TET4 large-model observations are workload- and environment-specific, not general performance guarantees.
  • PETSc, MPI and SLEPc are optional external runtimes. Their absence must be reported for a backend-dependent run and does not invalidate core import.
  • The .inp importer is a provisional bounded subset, not full Abaqus or CalculiX compatibility. Mixed HDF5 support is opt-in and does not establish restart, parallel-HDF5, or general scalability.
  • WEDGE15 is unsupported; PYRAMID5 remains internal/research-only. HEX8-SRI is separate experimental evidence; no blanket HEX8R, B-bar, or hourglass-control production claim is made.

Verification, validation, and release

  • Code_Aster results are same-mesh numerical correlations for recorded linear-static observables and selected families. They are not experimental physical validation, general mesh convergence, general stress-field equivalence, or nonlinear solver correlation.
  • Internal observable recomputation is not an independent global FEM/Newton solve. A test pass is not itself a maturity decision.
  • No certification, Abaqus equivalence, industrial-grade guarantee, or universal physical validation is claimed.
  • Selected wheel/sdist audits are bound to their exact source and release contract. A package-scoped result does not clear the full-repository G03 archive gate or the automatic GitHub source archives. G03 remains FAIL and WP14 remains on HOLD.

For exact published element-analysis boundaries, see the 0.2.8 consolidated registry. For evidence see the 0.2.11 V&V summary, the separate 0.2.10 records and the capability index.