Predictive Maintenance of Metro Rail Station Facilities Using Tree-Based Machine-Learning Algorithms

Station facilities are essential components of metro rail systems. The effective maintenance management of station facilities is crucial for ensuring operational efficiency, passenger safety, and cost-effectiveness. Currently, the facility manager relies on preventive maintenance and corrective maintenance to perform the maintenance planning at the station. However, these approaches have a limitation, where corrective maintenance cannot prevent failures, and preventive maintenance is unable to predict the future conditions of station facilities. Moreover, existing studies have shown that the application of machine learning (ML) has been limited to railway track structures or parts such as track, rolling stock, and catenary, and the adoption of ML to enhance maintenance planning of station facilities remains unexplored. Hence, this study aims to develop an ML model that enables the prediction of the failure category, maintenance needs, and maintenance status of metro rail station facilities. Tree-based ML models such as decision tree (DT), random forest (RF), and gradient-boosted tree (GBT) were applied. The results showed that the GBT model achieved the highest accuracy in predicting failure category (74.60%) and maintenance status (88.88%). Meanwhile, the RF model performed the highest accuracy of 96.83%, compared with DT and GBT models in predicting maintenance need. The findings indicate that the developed ML models can be applied to other facility types within the station.