Abstract This paper presents the development and validation of an automatic onboard monitoring system installed on a locomotive for railway infrastructure condition assessment. The system integrates a strategic layout of accelerometers mounted on axles, bogies, and the carbody, combined with triggering mechanisms for accurate structural monitoring of track sections. Real-time solutions for data acquisition, storage, and transmission enabled long-term deployment under operational conditions. A case study was conducted on a 450-m track section on Ferrovia Tereza Cristina, in Brazil, comprising both intact rails and areas with spalling, squats, and severe wear with plastic deformation. Two defect detection methodologies were evaluated: one based on time-frequency domain features using continuous time wavelet transform, and another relying on time-domain autoencoder modeling. Both methods effectively distinguished damaged from intact sections and demonstrated sensitivity to defect severity, with higher damage indices observed for severely damaged rail sections. The results confirm the robustness of the monitoring system and the potential of data-driven approaches for accurate and continuous railway condition assessment.
Bragança et al. (Mon,) studied this question.