intel/Predictive-Assets-Maintenance ? reverse-engineered prompt

Build me a working end to end predictive maintenance demo for elevator sensor data.

I want a notebook that loads historical time series data, cleans it, prepares it for training, then trains a forecasting model to predict future vibration or other sensor values, and finally runs anomaly detection to flag likely failure points. Please make it feel like a complete walkthrough, from raw data to predictions to anomaly results, with clear cells and comments so I can follow along.

Use the same overall flow described in the repo, with time series preprocessing, forecasting, and threshold based anomaly detection. If there are example settings that work better for this kind of data, use reasonable defaults and explain them briefly in the notebook. Include a simple way to visualize the results so I can see the forecast and the detected problem points. If you need to check current library docs online to make sure the notebook runs, please do that.