Experimental implementation of an autonomous drive-by monitoring system for freight railway systems

This paper presents the development and experimental validation of an automatic drive-by monitoring system installed on a locomotive for railway infrastructure condition assessment. The system integrates twelve accelerometers distributed across axles, bogies, and the car-body. To ensure precise localization and efficient data management, a triggering mechanism based on Ultra-High Frequency Radio-Frequency Identification technology was implement-ed. Passive tags installed on track sleepers communicate with a locomotive-mounted anten-na to automatically initiate data acquisition upon approaching target sections, while simul-taneously enabling vehicle speed estimation. A case study was conducted on a 450-meter track section of the freight railway line Tereza Cristina, in south Brazil, comprising both intact rails and areas with spalling and severe wear. An unsupervised defect detection meth-odology was evaluated based on time-frequency domain features using the Continuous Wavelet Transform and convolutional autoencoders. The methodology, trained exclusively on healthy data, effectively distinguished damaged from intact sections and demonstrated sensitivity to defect severity, with higher damage indices observed for severely damaged rail sections. These results confirm the system's robustness for continuous condition assessment under real operational conditions.