Western-OC2-Lab/Vibration-Based-Fault-Diagnosis-with-Low-Delay ? reverse-engineered prompt

Build me a notebook based project for vibration fault diagnosis of rolling bearings, using the same approach as this paper, where short vibration segments are turned into compact features with wavelet packet decomposition and FFT, then used to classify the machine condition with low delay.

I want separate notebooks for the CWRU, PU, and uOttawa datasets, and each one should be able to load the .mat files, prepare training and test data, extract the features, train a classifier, and show the results clearly with accuracy and basic comparison plots. Please keep the workflow easy to follow, since I want to run it in Jupyter and understand each step. If any dataset details or file formats are unclear, look up the current docs online if you need to. Also include the figures or visual summaries that help explain how the method works and how the delay and feature size stay small.