What's New in QF Solver 0.2.11

This page summarizes changes from QF Solver 0.2.10 to 0.2.11. For the cumulative solver capabilities, use the capability index. QF Solver 0.2.11 was published on 2026-10-07. Its exact-version DOI is 10.5281/zenodo.23214487. The selected wheel, sdist and checksum manifest are the audited distribution artifacts; GitHub-generated source archives are not part of that selection.

Overview

Version 0.2.11 builds on the existing structural mechanics, dynamics, nonlinear routes and reproducible V&V system. Its main new capability is a first experimental gyroscopic modal route for a tightly bounded shaft-and-disk model, followed by an experimental Campbell speed sweep and complex modal tracking. This release does not broaden the maturity of the pre-existing 0.2.10 capabilities.

Gyroscopic modal analysis

The new rotating_modal route solves one prescribed signed speed at a time for a linear, straight, connected, unbranched BEAM2 shaft carrying centered rigid axisymmetric disks. It uses a fixed global rotation axis, small perturbations, an undamped and unprestressed model, and a serial dense SciPy generalized QEP. Disk inertia supplies the gyroscopic contribution; the signed spin speed is applied once in the QEP. Complex eigenvalues and mode shapes remain available in the result.

This is not a general rotating-machinery solver. Distributed shaft gyros, variable speed, speed-dependent stiffness or bearings, centrifugal stiffening, forced unbalance response, contact, nonlinear rotors, and PETSc/SLEPc/MPI are outside this scope.

Campbell sweep and tracking

The campbell route orchestrates explicit, strictly ordered spin speeds by calling the single-speed solver and associating complex modes. A mass-weighted complex MAC, global one-to-one assignment and subspace-aware handling of degeneracy provide branch lineage where the data support it. When an association is unclear, the result retains candidate matches or a gap; the plotting helper cannot assign tracks or fill missing points.

At 100 rad/s, the high-frequency pair remains ambiguous. The tracker preserves the ambiguity rather than forcing branch continuity. This limitation remains visible in the structured result and the plotted example.

An optional 1× line is a frequency reference, abs(Omega)/(2*pi). A crossing is only a frequency coincidence or candidate critical-speed location in this linear modal model; it does not predict forced-response amplitude, unbalance response, operational danger, or instability.

Verification evidence

The verification evidence contains six bounded gyro checks. GYRO-01 checks zero-speed recovery against the classic modal route; GYRO-02 checks the local disk mass, gyroscopic matrix, frame covariance, signed-speed convention and ownership; GYRO-03 uses an independent analytical disk-oscillator oracle; and GYRO-04 checks signed-speed spectral splitting. GYRO-05 exercises analytical Campbell branches, degeneracy, phase/order invariance and controlled ambiguity. GYRO-06 checks internal mesh convergence for a BEAM2 shaft with a centered disk.

For GYRO-06, the maximum relative frequency change between the 16- and 32-element meshes was 0.00011433 (about 0.0114%), below the prospectively fixed 0.5% criterion; the maximum QEP residual was 1.6174e-9 against 1e-8. This is internal mesh-convergence evidence, not independent physical validation. Neither numerical PASS nor this evidence promotes maturity: rotating_modal and Campbell both remain EXPERIMENTAL.

See the 0.2.11 V&V summary and the frozen Campbell contract for provenance, thresholds, evidence availability and the full boundaries.

Compatibility and citation

Version 0.2.11 keeps the existing qf_solver namespace and 0.2.x compatibility surface. The new routes and their complex result contracts are experimental and provisional, not stable downstream API commitments. See the installation guide for exact-version installation. Cite this release with version DOI 10.5281/zenodo.23214487, and use the concept DOI 10.5281/zenodo.22697897 for the project across versions. The preceding version DOI 10.5281/zenodo.23106744 identifies 0.2.10 only.

Historical G03 remains FAIL_PRESERVED, the whole-repository archive is not cleared, and WP14 remains HOLD_NOT_PROMOTED.