Skip to content

QF Solver 0.2.5a0 G04 External Branch Diagnostic

Decision Boundary

This document resolves the discrepancy between the internal two-element TET4 arc-length path and the historical Code_Aster replay. It does not close 025-G04, change any acceptance criterion, or make an arc-length production claim.

Provenance item Value
Original QF numerical source 067859c1db8416758f82bc339cc5b2db4bcbaf63
Historical Owner evidence 8fdbd500e52a2c23174f66c68e19ecfaa57be0c5
Authoritative corrected-run source results/vnv_0_2_5/g04_latest/evidence_manifest.json
Gate status OPEN
Contract lowered NO

Strict Configuration Comparison

Parameter QF Solver Historical Code_Aster deck Corrected Code_Aster deck Impact
Geometry five unperturbed nodes; apex (0, 0, 0.25) same same match
Connectivity N1-N3-N4-N5, N2-N4-N3-N5 same same match
Material E=100, nu=0.3 same same match in the bounded elastic case
Kinematics Total-Lagrangian, Green-Lagrange / second Piola ELAS, GREEN_LAGRANGE same bounded formulation match
Supports N1/N2 fixed; crown UY=0 same same match
Physical load QF reference +1/3 with target factor -1: downward FZ=+1/3: upward FZ=-1/3: downward historical mismatch resolved
Control output apex node N5 UZ (global DOF 14) mean CROWN/DZ APEX/DZ historical mismatch resolved
Continuation window 80 QF steps through |u_apex|=0.94194 0..100, 800 intervals 0..0.96, 160 intervals corrected window covers the QF branch
Arc constraint QF augmented displacement/load norm Code_Aster LONG_ARC displacement control same compare equilibrium points by apex displacement

The values are a dimensionless/scaled numerical benchmark. No physical unit validation is claimed.

Diagnostic Result

The corrected Code_Aster run gives one turn at the same branch location as QF Solver. Interpolating Code_Aster's reaction-derived load factor on QF's absolute apex displacement gives:

Metric Result
QF turning point u=0.8933769433, lambda=-0.03732654177
Code_Aster turning point u=0.8940000000, lambda=-0.03732647828
Absolute turning-load difference 6.3485e-08
Maximum branch load-factor difference 4.8719e-07
RMS branch load-factor difference 2.0730e-07
Peak-normalized maximum difference 1.3052e-05

QF's reduced tangent is symmetric to numerical precision. Its smallest eigenvalue crosses from +1.1649e-03 at step 75 to -3.3334e-06 at step 76, then stays negative on the post-limit branch. The reassembled free residual near the turn is below 5e-13, and det(F) remains positive (>=0.4414 over the recorded path). This supports a physical limit-point interpretation within the bounded two-element model.

Classification

DIAGNOSIS = B - CODE_ASTER_PATH_FOLLOWING_CONFIGURATION
ROOT CAUSE = opposite physical load direction plus unmatched control observable
             and continuation window in the historical deck
QF_MODEL_VALID = YES, bounded numerical correlation only
CODE_ASTER_MODEL_VALID = YES after deck correction
PUBLISHED_REFERENCE = NOT_IDENTIFIED
CODE_CHANGE_REQUIRED = NO
G04 STATUS = OPEN

The exact two-element configuration has no linked published FEM source in the repository, and no exact public reference was identified for it. The Code_Aster manual supports LONG_ARC as a continuation method but is not a published reference for this particular geometry or branch.

Remaining G04 Blockers

  1. Link and reproduce an exact published FEM snap-through branch reference, or formally replace this custom benchmark through the requirements process.
  2. Run the required coarse/medium/fine/refined branch study only after the benchmark reference is approved.

No mesh study was run during this diagnostic. The corrected Code_Aster branch removes the historical external-deviation blocker but does not close the gate.