Abstract The demand for faster, longer, and safer railway networks has intensified the need for robust condition monitoring systems, particularly those capable of detecting rolling stock defects such as wheel flats, bearing faults, hunting movement, overloading, and unbalanced loads. Effective monitoring and timely maintenance are essential to mitigate these issues, enhancing both reliability and safety in railway operations. This paper presents a comprehensive review of wayside condition monitoring (WCM) systems. The study begins with a structured overview of WCM architectures, followed by a bibliometric analysis that highlights recent research trends in machine learning applications for railway condition monitoring. A detailed classification of rolling stock defects is provided to establish the foundation for condition assessment. The review then presents an extensive survey of commercial wayside monitoring systems currently in use, including hot axle box detectors, wheel impact load detectors, acoustic detection systems, and weigh-in-motion technologies. The various sensor types integrated into these systems are also described in detail. To bridge the gap between raw data collection and actionable insights, the paper includes a dedicated section on knowledge extraction from WCM systems. This section outlines key approaches for fault detection, diagnosis, and classification using advanced data processing methods, including signal processing and machine learning techniques. Finally, the paper identifies critical challenges such as data quality, real-time processing constraints, infrastructure limitations, and the need for generalizable models. It also discusses research gaps and suggests future directions, including the integration of edge computing, digital twin technology, and self-diagnostic capabilities to support the development of next-generation intelligent WCM systems.
Mosleh et al. (Sat,) studied this question.