PREDICTIVE MAINTENANCE FOR FREIGHT TRAINS: A MACHINE LEARNING APPROACH TO OPTIMIZING WHEEL REPLACEMENT SCHEDULES : Machine Learning Approach for Predictive Maintenance

Efficient maintenance planning is crucial for railway operators to minimize costs and operational disruptions. Green Cargo, a major rail freight provider, currently replaces train wheels based on scheduled maintenance or reactive repairs, leading to inefficiencies. This thesis explores the application of machine learning to predictive maintenance by analyzing sensor data from DPC detectors mounted along railway tracks. These detectors measure force impacts as the train moves, providing key indicators of wheel wear. Using supervised learning techniques and grid search optimization, predictive models for left and right wheel damage were developed and evaluated. The left wheel model achieved a test R2 of 1.0000 with an RMSE of 0.08 kilonewton, while the right wheel model achieved a test R2 of 0.9873 with an RMSE of 0.36 kilonewton. Forecasting trends over a three-year horizon identified one axle requiring replacement, with left wheels typically failing first. Maintenance scheduling logic was implemented to replace both wheels on an axle when either exceeded predefined thresholds. This research demonstrates the potential of predictive maintenance to enhance efficiency and reduce costs in railway operations by leveraging advanced machine learning methods and historical data from DPC detectors. Future work will focus on integrating real-time monitoring and refining predictive capabilities for broader industrial applications.