Bearing fault detection

Improving on NASA's work with induction motor bearing fault detection using RNN-powered smart sensors.

Walkthrough

For starters, you'll want to run source setup_venv.sh to automatically setup a Python virtual environment under bearing_venv. You may want to experiment with different versions of analysta (to be found in the anomaly_detection submodule) to make sure training works properly.

Then:

• To preprocess the NASA data, download it from here and unpack it into bearing-fault-detection/data, then cd bearing-fault-detection and run python3 preprocess_data.py.
• To train the AI, cd bearing-fault-detection and run analysta -vv model single -c lstm_config.json.
• To view the trained model and a some stats, take a look at bearing-fault-detection/lstm_results.
• To view spectrograms of the raw data, head over to bearing-fault-detection/spectrograms - to generate them, you could cd bearing-fault-detection and run python3 spectrogram.py.