E01 OPERATING MODELS - LATEST REPAIRED-VOLUME EDITION

Current summary/data: results-volume-v2/E01_summary.json and E01_web.json.
Current figures: figures-volume-v2/E01_driver_response.png/.svg and
figures-volume-v2/E01_captive_impact.png/.svg.
Current exact package list: publication-v2/PUBLIC-FILES.json.
Original results, failed controls, figures and source remain preserved.

This is a completed numerical research package WITH PRESERVED FAILURES.
It is not a built/validated acoustic exciter, fracture device or safety rating.
The latest composite acoustic_numerical_checks_passed flag is FALSE. Do not
replace that with an all-tests-passed headline. See the specific ledger
failure below; successful refinement comparisons do not erase it.

WHAT RAN
Original matrix:972 algebraic frequency points and40 trajectory integrations
(24 acoustic,16 impact),137.11 seconds.17 author tests passed, zero skips.
Those tests separately included5 short integrations, outside the matrix.
Corrected volume:486 new algebraic points and12 new ND64 integrations,
36.094 seconds. The current972-point/40-integration view contains12 new and
28 inherited integrations. Historical matrix total=40+12=52, not80.
There are zero physical trials. No impact trajectory was rerun for the CAD fix.

WHY THE VOLUME CHANGED
CAD checking found four bosses merely tangent to the wall. Increasing their
radius6to7mm made them connected solids. The net air-volume proxy changed from
0.000706091410953079 to0.0007039780004475063 m^3 (0.7039780004475063 L).
The two current ND64 volume assumptions are that CAD proxy and one half of it.
The ND91 reference uses hypothetical1 and2 L net boxes. These are actual
named-driver parameter sets, not scaled copies. Actual cavity volume,
permeability, wiring, gaskets, leaks and driver parameters require measurement.

PRESERVED NUMERICAL FINDINGS
The corrected ND64/CAD/20Hz coarse run has returned-supply-energy minimum
-3.1023928854649686e-12 J at0.0125 s, below its frozen-1e-12 J lower bound.
Its fine counterpart is-3.551261472306674e-14 J and passes that same bound.
Both fit the linear frequency solution and energy balance closely, but the
coarse author's nonnegative-ledger gate remains failed. Independent review
also preserves14 stricter monotonicity findings across the current24 acoustic
traces. The report does not claim exact monotone cumulative energy.

The post-hoc diagnostic in diagnostics/ uses the closed-form linear state and
64/128-node positive-weight quadrature to examine the first12.5ms. Both give
zero returned energy in that interval. This supports a dense-output sampling
explanation but does not prove the exact internal solver cause. It changes
no trace, threshold or failure flag and performs no new ODE integration.

Impact I04's deliberately soft capture crosses the hard5mm boundary at both
step sizes; both failed rows remain. Seven of eight fine cases finish within
the model's domain/numerical screens. Only two are numerically near rest over
the final50ms. Guide, contact and buffer laws are hypothetical; no physical
capture reliability or hard-stop survival is established. Manual release
is one impact. Powered reset and programmed repeated BPM remain future work.

REPRODUCE
Tested runtime: Python3.12.14, NumPy2.5.3, SciPy1.18.1; Matplotlib3.11.2 for
plots. The modules use Python3.10+, NumPy1.26+, SciPy1.11+ syntax/APIs.
Run from this directory; every output directory must be NEW.

  python -m unittest test_operating_models -v
  python publication-v2/verify_package.py --require-traces
  python run_study.py --output results-original-rerun-001
  python run_volume_revision.py --output results-volume-rerun-001
  python make_figures_v2.py --results results-volume-rerun-001 --output figures-volume-rerun-001

The correction deliberately inherits the exact frozen results/ parent from
this package, not a new rerun whose timestamps/hashes differ. Obtain all52
NPZ files using publication-v2/TRACE-INDEX.json for complete trace verification.
The core ZIP excludes those files. Without them, verify_package.py with no
--require-traces flag reports core-only verification and explicitly counts
omitted traces. Author tests and the algebraic browser comparison do not
require the downloaded trace payloads.

Read publication-v2/REPRODUCE-CORRECTED-VOLUME.txt for detailed lineage and a
copied per-detail evidence_origin label clarification. Summary-record origin
fields correctly identify the12 actual corrected-volume integrations.

No hardware control, amplifier calibration, thermal protection, measured SPL,
acoustic focus, diamond/rock fracture or universal large-object capability is
provided. All source/input/result hashes and failed controls remain available.
