CNN architecture for arc fault detection

[“

Claude responded: Open it in Google Colab with File → Upload notebook, then Runtime → Run all.

Ran 2 commands, created a file, shared files

Ran 2 commands, created a file, shared files

Open it in Google Colab with File → Upload notebook, then Runtime → Run all.

\n

It walks through:

\n
\n- synthetic normal, dimmer, SMPS, and arc signals
\n- FFT and spectrogram views of each
\n- a tiny strided Conv1D with global average pooling, compared against a pooling version
\n- per-load false-trip rates and k-of-m trip logic
\n- int8 quantization, with a C header you can drop into firmware
\n
\n

There are exercises at the end. I tested the signal and dataset cells. The TensorFlow cells couldn’t run in my environment, so if one errors in Colab, paste the error here and I’ll fix it.



Afci conv1d starter

IPYNB 

”,”

See CNN architecture for arc fault detection - How deep to learn new methods.

See CNN architecture for arc fault detection - Avoiding rabbit holes.