Detection of Known Railway Wheel Defects Using Sound Pressure Measurements from a Full-Scale Train

Railway wheel defects are a major source of noise and vibration and may compromise operational safety if not detected at an early stage. Although numerous condition monitoring techniques have been proposed, validation under real operating conditions remains limited because of the scarcity of datasets with independently known defect locations and characteristics. This study presents an in-situ investigation of wheel defect detection using sound pressure measurements acquired during the passage of a full-scale train equipped with both reference wheels and wheels containing pre-identified defects. Experimental campaigns were conducted across 12 distinct test conditions at train speeds ranging from 10 to 60 km/h in both travel directions to evaluate the influence of operating conditions on the acoustic response. The acquired signals were analysed in the time and frequency domains. Defects were identified by combining the local sound pressure amplitude levels with temporal fluctuation indices derived from moving standard deviation. This approach enhanced transient acoustic signatures and facilitated the localisation of wheel defect-related events. The results demonstrate that the proposed Energy-Variance Product methodology successfully isolates defect-related transient acoustic features across all tested speeds, effectively suppressing continuous background locomotive noise and providing reliable diagnostic data, while the unique experimental configuration provides reliable data for validating acoustic monitoring methods under real railway operating conditions. These findings support the use of sound pressure measurements as a practical tool for railway condition monitoring and establish a valuable experimental benchmark for future diagnostic methods.