ON07 – Predictive Maintenance in Train Bogies by Analysing Track Vibrations by AI

This project presents an innovative approach for detecting defects in train wheels within the MTR network, leveraging vibration detection at railway points. The current methods of in- workshop wheel defect inspection, such as visual inspection and ultrasonic techniques, are limited by their interval-based scheduling and partial coverage of the wheel tread area. To address these limitations, this project uses artificial intelligence to extend the functionality of existing point vibration monitoring systems to detect abnormalities in train wheels and predict maintenance needs. The proposed system is expected to optimize maintenance resources, reduce unscheduled downtime, and advance the MTR’s operational capabilities through AI-driven predictive analytics. This project underscores the potential of this technology to revolutionize train wheel maintenance by providing a more efficient, accurate, and cost-effective solution. The effectiveness of this invention has been proven by cross-checking with existing maintenance logs of train wheels. Subject to data quality, our invention could predict wheel turning maintenance need for train wheels in the next five days with an accuracy of around 60%. It was found that our invention was most effective in detecting wheel flats and cavities.