NUMERICAL PROOF OF CONCEPT -- NOT A PHYSICAL TEST

Run from this directory with Python 3.10+ and NumPy installed:
  python generate_proof.py
Or run only the verification checks:
  python test_resonance_model.py

All outputs are created under results/. Every generated data file begins
with NUMERICAL or SYNTHETIC so it cannot reasonably be mistaken for acquired
specimen data. The generator overwrites only these defined simulation files.
No hardware is opened, no waveform is transmitted and no internet is used.

WHAT IS CHECKED

The custom NumPy-only finite-element code assembles two-node cubic-Hermite
Euler-Bernoulli stiffness and consistent mass matrices. It removes clamp
displacement and rotation, transforms K phi = omega^2 M phi using Cholesky,
and solves the symmetric eigenproblem. A uniform bare cantilever is compared
with the independent continuum result:
  f_n = beta_n^2/(2*pi) * sqrt(E*I/(rho*A*L^4)).
The first roots beta are approximately 1.875104, 4.694091 and 7.854757.

The nominal CAD coupon is 400 x 120 x 3 mm with 100 mm inserted into a clamp:
the modeled FREE length is 300 mm. E=69 GPa and rho=2700 kg/m^3 are assumed,
not measured. Beam thickness is the bending direction. Frequencies do not
include actuation, supports, cable loading, damping or fixture behavior.
At 120 mm width the actual coupon is not an ideal narrow beam; plate and
Poisson effects can invalidate the one-dimensional approximation. Numerical
agreement with beam theory is verification of the solver, not experimental
validation of this coupon or assembled apparatus.

An attachment sensitivity calculation adds a hypothetical rigid 90.5 g
point mass 150 mm from the clamp and an assumed 2 g receiver at 250 mm.
The 90.5 g value is the published WHOLE exciter mass, not its moving mass or
effective frequency-dependent mechanical impedance. This lumped model omits
motor suspension, rotary inertia, attachment footprint, adhesive compliance,
base response and other coupling. It therefore must not be presented as the
predicted frequencies of the actual assembled demonstrator. Weighing a head
alone does not resolve those limitations.

The separate analytic damped oscillator begins from rest under sinusoidal
force. It checks that accumulated input work equals stored energy plus
viscous dissipation within numerical integration error, and that a finite
constant force reaches a finite damping-limited response. The illustrative
mass is 1 kg, natural frequency 200 Hz, Q=20 and force peak 1 mN. These are
mathematical test parameters, not estimates of the actuator or coupon.

SYNTHETIC ELECTRICAL DATA

SYNTHETIC_voltage_sweep.csv contains frequency_hz,input_peak_v,response_peak_v,
gain_v_per_v,phase_deg. It models a dimensionless second-order transfer with
f_n=200 Hz, Q=20 and arbitrary static gain 0.01 V/V. At f_n its gain is 0.2 V/V
and its phase is -90 degrees. It is INDEPENDENT of the beam FE model.
The applied input is 20 mV peak and added phasor noise is assumed, not a
specified acquisition noise floor. The file is a toy settled sine sweep.

SYNTHETIC_voltage_waveform.csv contains time_s,drive_v,response_v at 20 kS/s
for two seconds of 200 Hz steady-state sinusoidal drive. There is no drive
onset or sweep transient in this file. The random seed is fixed for exact
reproduction. SYNTHETIC_ground_truth.json records the parameters.

Voltage ratios and phase do NOT constitute mechanical force, acceleration,
displacement, energy, strength, acoustic pressure or material failure limits.
A real piezo voltage measurement requires calibration and has a frequency-
dependent mounting response. A sampling guard cannot remove aliased content
from an actual recording; suitable acquisition anti-alias filtering is still
required. No diamond, boulder, steel pole or fracture performance is inferred.

REFERENCE BASIS

MathWorks, Dynamics of a Damped Cantilever Beam:
https://www.mathworks.com/help/pde/ug/dynamics-of-a-damped-cantilever-beam.html
This supports the analytical fundamental-frequency comparison and damped
modal/static/transient framework; our NumPy implementation is independent.

Eric Raymond Johnson, Virginia Tech, Aerospace Structures, chapter 18:
https://eng.libretexts.org/Under_Construction/Aerospace_Structures_(Johnson)/18:_Introduction_to_flexible_body_dynamics
This supports the mass/stiffness eigenproblem and consistent mass approach.

Parts Express, DAEX25Q-4 entire-head mass:
https://www.parts-express.com/Dayton-Audio-DAEX25Q-4-Quad-Arm-25mm-Exciter-20W-4-Ohm-295-228

The meaningful next verification step is physical measurement of a known
reference specimen and repeatability after remounting. No such test has been
performed for this package, and no novelty or patentability is established.
