Detecting and Classifying Rail Corrugation Based on Axle Bearing Vibration

• Published in: ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) pp. 2752 - 2756
• Main Authors: Alten, K., Fuchs, A.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.05.2019
• Subjects: condition assessment; corrugation; machine learning; vibration features
• ISSN: 2379-190X
• DOI: 10.1109/ICASSP.2019.8683317

Vienna's tram network is regularly surveyed by an inspection tram that uses a laser light section method to measures wear of the rail head and a microphone to detect curve squeal. In order to expand the vehicle's inspection capabilities and encompass more condition indicators for maintenance purposes, the vehicle's accelerometers which were hitherto only employed as an inertial measurement unit are now used to monitor the vibration of the axle bearings as the tram travels across the network. The inspection tram is equipped with four triaxial accelerometers on the axle bearings of the middle bogie and another one on the bogie itself. In the first stage of the current research project, the aim is to automatically detect and classify periodic unevenness of the rail head, known as corrugation, by training a classifier on a range of vibration features. Applying this machine learning algorithm during post-processing allows the condition state of the rail head to be compared across the network.