This study introduces a predictive modelling framework aimed at assessing vehicle downtime risk through the application of advanced machine learning techniques. Minimising unplanned downtime is crucial for enhancing operational efficiency and reducing costs within transportation and logistics systems. Utilizing maintenance records from a South African coal transportation enterprise, we developed and evaluated several machine learning algorithms, namely Random Forest, Logistic Regression, Support Vector Machine, and XGBoost, to predict high- and low-risk downtime events. The dataset, which consists of 201 records collected over a three-month period, was preprocessed and balanced using the Synthetic Minority Over-sampling Technique (SMOTE) to improve classification accuracy. Among the models assessed, XGBoost delivered the highest predictive performance, achieving an AUC of 0.97, a macro F1-score of 0.87, and a weighted F1-score of 0.88, demonstrating a strong ability to differentiate between risk categories. Feature importance analysis revealed that Failure Mode and Depot were the most influential predictors of downtime, providing valuable insights for targeted maintenance planning. Additionally, the study outlines a deployment strategy for integrating the model into operational systems via local and web-based platforms, facilitating data-driven maintenance scheduling. These findings underscore the transformative potential of machine learning in predictive maintenance and establish a framework for leveraging AI-driven analytics to enhance reliability, resource allocation, and operational resilience within the vehicle and railway sectors.