Railway Wheel Damage: Mechanisms, Monitoring, and Prediction

Abstract

Abstract: Wheel damage remains a major constraint on railway safety, dynamic performance, asset life, and maintenance efficiency. Despite extensive studies of individual damage modes, understanding remains fragmented across mechanisms, monitoring, prediction, and maintenance application. This review provides an integrated synthesis of damage formation, measurable evidence, predictive modelling, and maintenance relevance. It examines four major damage modes, namely polygonal wear, wheel flats, rolling contact fatigue, and hollow wear, and clarifies the mechanisms and influencing factors governing their initiation, coupling, and evolution. Bench testing and in-service monitoring are assessed as complementary evidence sources for mechanism validation, parameter calibration, diagnosis, prediction, and maintenance assessment. Prediction studies fall into three paradigms: physics-based, data-driven, and hybrid approaches. Physics-based approaches retain mechanistic value for long-term profile evolution but depend on accurate contact representation and impose high computational cost. Data-driven approaches support efficient diagnosis and short-term prediction but remain constrained by data quality, label scarcity, drift, and cross-domain transfer. Hybrid approaches combine physical structure with data adaptability but still require field validation and maintenance-oriented evaluation. The main barrier is weak integration among mechanism interpretation, sensing evidence, predictive capability, and maintenance criteria. Future progress depends on benchmark datasets, transferable monitoring and prediction frameworks, uncertainty-aware evaluation, and stronger links between model outputs and maintenance decision-making. Conflict of Interest Statement The authors declare no conflicts of interest.