Intelligent Predictive Maintenance in Urban Rail Systems: Computer Vision Approaches for Enhanced Wear Detection

Abstract

This study explores the integration of deep learning techniques for predictive maintenance within urban rail systems, emphasizing advanced automated rail head wear detection utilizing Mask Region-based Convolutional Neural Network (Mask R-CNN) and You Only Look Once (YOLO)v8 models. The increasing traffic and operational demands, especially in regions like Pakistan, amplify challenges such as rolling contact fatigue and head checks. This study showcases the effectiveness of the Mask R-CNN and YOLOv8 models, with Mask R-CNN achieving an average Intersection over Union of 0.8694 and a Dice Coefficient of 0.9300, and YOLOv8m attaining an average Intersection over Union of 0.9564 and a Dice Coefficient of 0.9777. These findings underscore the potential of these models to significantly enhance rail maintenance protocols and contribute to the sustainability and safety of intelligent urban rail transportation systems.