Three-regime relay consolidation: CC → opt_passive → ff+PID across all VGOSWEC flap variants

[Copilot]

Finalizes the simulation study as a clean three-controller regime relay — CC (short periods) → opt_passive (resonance band) → ff+PID (long tail) — across all five flap variants (VGM-0/10/20/45/90), and prunes the degenerate fixed-passive arm. All figures regenerate from committed CSVs via --plot-only.

Degenerate passive pruned

B_pto = B55(ω₀) is 10⁴–10⁵× smaller than |Z_intrinsic(ω₀)| for every flap (VGM-0 at 3.2e-7 N·m·s/rad in the pitch-radiation notch). Passive captures ≈ 0 W across the full band — it adds no signal to the study.

• analysis/passive/ and analysis/passive_vs_optpassive/ deleted
• config/vgoswec_{0,10,20,45,90}_passive.yaml reverted: B_pto: 0.5 placeholder + TODO: tune with tank-test data, wave.period: 1.5. Controller type remains usable in code.

New: scripts/three_regime_comparison.py

Modeled on cc_vs_ffpid_comparison.py. Loads CC (analysis/cc/), opt_passive (analysis/opt_passive/), ff+PID (analysis/passive_guarded/) CSVs and produces:

• Per-flap 3-way figures — shaded CC/opt_passive/ff+PID regime bands with interpolated crossover markers; masked=true regions hatched
• Cross-flap summary figures — power and efficiency overlays
• Master operating-envelope figure — upper hull of P_capture over all (controller, flap-angle) at every period; annotated by winning regime band

python3 scripts/three_regime_comparison.py --plot-only
# → analysis/three_regime/figures/{three_regime_VGM*.png, operating_envelope.png, ...}
# → analysis/three_regime/operating_envelope.csv

Resonance hump marches with flap angle (key result)

Flap	opt_passive peak	peak T	ff+PID at resonance	winner
VGM-90	0.509 W	2.50 s	~0.55 W	ff+PID edges
VGM-45	0.479 W	3.00 s	~0.63 W	ff+PID edges
VGM-20	0.755 W	3.25 s	~0.73 W	opt_passive edges
VGM-10	0.772 W	3.50 s	~0.75 W	opt_passive edges
VGM-0	0.681 W	4.75 s	~0.68 W	tie

opt_passive matches a tuned feedforward controller at resonance with a single tuning-free coefficient — it does not universally win. ff+PID edges it on high-angle flaps; the honest framing is "ties at resonance."

Docs

• analysis/FINDINGS_3REGIME.md (new) — three-regime relay table, per-flap peak table, honest opt_passive vs ff+PID framing, appendix on passive degeneracy
• docs/CONTROLLERS.md — added three-regime relay section; deprecated passive-vs-opt_passive study
• analysis/FINDINGS.md, README.md — updated reproduction commands and framing