Rogowski Coil Modeler

This is a self-contained browser app for studying how Rogowski coil installation geometry can create current measurement error, then comparing installation methods with Monte Carlo simulation.

What it models

The app starts with a fast engineering model:

I_meas = I_true * F_offset * F_tilt * F_shape * F_skew * F_neighbors

Where:

• F_offset ~= 1 - k_o (r / R)^2
• F_tilt ~= cos(theta)^n
• F_shape ~= 1 - k_s (1 - b/a)
• F_skew ~= 1 - k_a |skew| / R
• F_neighbors is a simple first-order adjacent-conductor coupling term

This is intended for concept screening and method comparison, not as a final physics claim. The fitted coefficients should be tuned against bench measurements.

What the app includes

• Deterministic geometry view for one install case
• Side-by-side Monte Carlo comparison for two installation methods
• Editable distributions for offset, tilt, ovality, skew, closure bias, and motion
• Summary metrics including mean error, P95 absolute error, and threshold exceedance rates
• Installation Quality Index P(|error| < target)

Running it

Because the app uses JavaScript modules, serve it from a local web server:

python -m http.server 8000

Then open:

http://localhost:8000

Suggested workflow

1. Set your reference geometry and fitted coefficients in Model Settings.
2. Use Deterministic Geometry Case to understand how one placement changes the response.
3. Define distributions for Method A and Method B based on observed field installs.
4. Run Monte Carlo and compare:
   • mean absolute error
   • P95 absolute error
   • fraction of installs above 0.5%, 1%, and 2%
   • Installation Quality Index
5. Calibrate k_o, k_s, the tilt exponent, and the neighbor term using bench data.

Calibration ideas

For each installation method, measure a sample of real installs or bench setups and record:

• coil offset from conductor center
• tilt angle
• major and minor loop axes
• axial skew
• adjacent conductor spacing and current
• latch support or hanging behavior

Then tune the model coefficients until simulated and measured error distributions line up reasonably well.

Next extensions

• Add CSV export of Monte Carlo samples for reporting
• Add saved method libraries for different coil or panel types
• Add a more physics-grounded numerical geometry model
• Add calibration mode that fits coefficients from measured data