[1-Bort-1]

Do not merge ahead of #415: this pull request closes when #415 merges. #415 makes the default solve agree with the Python VSM v2.3.0, and that also turns this correction on. It takes the quarter-chord flow by Python's rule, which conflicts with the rule here, and only one of the two can stand.

With the old default, one velocity both read the polars and turned the force, which keeps lift but loses induced drag: the straight elliptic AR 4 wing reached an Oswald efficiency of 0.73 against a theoretical 1. This branch turns the force by the flow at the quarter chord (Gaunaa, Li & Pirrung 2026) and reaches 0.998. In that flow it leaves out only the section's own bound vortex, where the existing code left out every section's.

Against the paper's vortex-lattice reference, this rule is the closer one on wings that are only curved: within 0.8 %, where #415 is off by 2.3–2.5 %. It is no better once the wing is swept, as a kite is. On the 15 % swept wing and the swept, anhedral wing it overshoots by 13.5 % and 7.8 %, where #415 undershoots by 8.2 % and 8.0 %. The paper's own model stays within 5.6 % and 0.5 % there. On the 25 % swept wing this branch gives NaN, and #415 stays finite but does not converge.

I judge that #415 should win: the project has asked for agreement with Python, this rule also moves the LLT model away from Python, and it has no accuracy advantage on a swept wing. What that costs is about 2 % of Oswald efficiency on a purely curved wing. Whoever wants this rule instead should say so on #415 before it merges.

On the TU Delft V3 kite the default change keeps CL and the 14° stall, and lowers drag at 10° from 0.0998 to 0.0742. Anyone calibrated to the old drag sees about 25 % less of it, and #415 moves that drag again.

The vortex-segment changes here exist because this rule evaluates bound filaments on their own axes, which the Python rule never does, so nothing needs them once this closes. #408 is built on this branch and has been told to move.

Verification

• Rule comparison: every wing from the paper's appendix, inviscid, 120 sections, α = 10° by constant twist, force on the chordwise legs on, run on this branch and on The default solve now gives the same results as the Python VSM v2.3.0 #415's head 242880a.
• The appendix does not fully state the geometry. I took the reading that brings CL closest to the reference: chord elliptic in projected span, twist about the local tangent, CL on blade area. That gives CL within 1.5 % on the dihedral, anhedral and combined wings. On the swept wings the chord direction stays an assumption, and The default solve now gives the same results as the Python VSM v2.3.0 #415's CL is 3.0 % low at 15 % sweep.
• Neither branch reproduces the paper's 12 % effect of the chordwise-leg force on the swept, anhedral wing: it moves e by 0.2 % here and 0.1 % on The default solve now gives the same results as the Python VSM v2.3.0 #415.
• Straight elliptic AR 4, old default: e = 0.7283. test/verification/test_verification.jl red with the old default (22/23), green now (24/24), including a pin of the old setting below 0.8.
• Local full suite on 602fcd5: PASS (9 min, Julia 1.13.0, one matrix cell). GitHub CI: docs PASS; the test matrix is skipped while this is a draft.
• Docs build: not run locally. No new public symbols.
• Benchmark: inconclusive; the box was saturated. Allocations on the V3 kite: 6 allocations, 208 B per solve with either setting.
• Risk: the swept-wing ranking rests on an assumed chord direction.

Opened by 1-Bort-1, an AI agent working for @1-Bart-1.
Closes #318 · task VortexStepMethod.jl-318

[1-Bort-1]

Independent review (advisory)

Verdict: APPROVE WITH COMMENTS · 2 inline, 0 off the diff

Good

• Matches the plan: the default flips in both SolverSettings and Solver, and the Eq. 2 fix is the one-line model == LLT && icp == jring change at body_aerodynamics.jl:480, as the card says
• Closed-form core is correct: I derived Γ/(4π ε² ℓ²)·d_sum·(r1×r0) from the core-boundary velocity scaled by h/ε, and the sign matches, since r1×r2 = −r1×r0
• Slot reuse is safe: work_vectors[8:9] are only otherwise used by induced_velocity_at, which is never nested in calculate_AIC_matrices!, and the filament kernels only touch slots 1–5
• body_aero.cache[5:6] are otherwise unused in src, so the allocation fix does not overwrite any buffer
• Bug-fix tests fail before the fix, as the card reports: the curved-wing LLT matrix test checks norm(velocity) > 1e-3 for i≠j, so it separates Eq. 2 from zeroing every bound filament; the ForwardDiff slope test gave 0 before
• The test_polar_vectors.jl and test_stability.jl edits tighten or correct the tests and loosen nothing: the 1e-5 CMy assertion is kept and the bisection tolerance now bounds CMy below it
• Comment trims and the uniform_AIC! helper are behaviour-preserving cleanup around the change, and the card names them

Not good

• src/settings.jl:202 — With true as the default, every VSM solve! (and every perturbation inside linearize) now builds a second N×N×3 aero-centre AIC, but the card gives no timing and only shows the change stays allocation-free. Simulators calling solve! every step will pay a cost nobody measured.
• src/filament.jl:115 — The zero-core cutoff scales with norm3(r1), the distance to the filament start, not with the filament length or epsilon. A point far along the extended axis therefore gets a looser 'on axis' band than one near x1, and neither the docstring nor a test pins down which is intended.
• velocity_induced and U_2D are still zeros(MVector) while their siblings moved to work_vectors; the zero-alloc result depends on escape analysis, and the code does not show why two moved and two did not
• The card's table says 'default core radius', but the new verification test goes through lift_drag_polar, which fixes core_radius_fraction=1e-20; the tested case is not the one tabulated
• vsm_settings_coarse.yaml changes its live value from false to true; the card calls this collapsing a duplicate, not a behaviour flip for users of that file
• Now that false is no longer the default, no test pins the correct_aoa=false path (LL-3/4); anyone who sets it back to get the old drag has no regression guard
• The 'LLT matrix leaves out only…' test rebuilds the AIC from the filament kernels, which is close to the parity-test pattern §2 warns about; a hand-computed expectation for one off-diagonal entry would be independent of the kernels
• The swept-wing no-core test only asserts isfinite; it would also pass if the fallback zeroed entries it should not

_{claude, rubric CLEAN_CODE.md. A different lab from the implementer
on purpose: a reviewer sharing its blind spots would not flag its mistakes.}

[1-Bort-1]

Local full suite: PASS (26 min, Julia 1.13.3, one cell of the matrix)