Adapting passenger vehicles for railways operation: A data-driven tribological study for tire-rail contact

Repurposing disused railway lines for lightweight passenger vehicles offers a cost-effective pathway to sustainable rural mobility. A critical technical challenge is maintaining reliable friction at the tire–rail interface under varying and degraded surface conditions. This study investigates the tribological behaviour of rubber-steel contact using a data-driven approach that combines friction measurements with 3D texture characterization of rail samples. Three experimental protocols were designed to simulate dry, wet, and greased contact conditions. A texture-aware k-fold cross-validation strategy was implemented to train and evaluate Random Forest (RF) regression models, including dimensionality-reduced variants by applying Rang correlation and Principal Components Analysis (RF-Rang and RF-PCA). Results show high prediction accuracy across all models and conditions, with residuals following near-normal distributions. Feature importance analysis reveals that specific surface descriptors, such as autocorrelation length, directionality, and valley aspect ratios, consistently influence friction. Their relevance shifts with environmental conditions, confirming that multiscale topography plays a central role in contact mechanics. This work provides quantitative insight into key surface parameters governing friction in rubber-steel interfaces and supports the design of adapted maintenance strategies for lightweight railway operations.