The operation of a tram fleet involves constant wear of the rail infrastructure. Resulting damage must be identified and repaired depending on its urgency in order to prevent further deterioration of the rail condition and consequential damage to the vehicle. To this end, costly manual inspections of the infrastructure are required in accordance with BOStrab. On-board sensor technology offers great potential to support and automate existing processes. This contribution examines how on-board acceleration measurements and inspection data can be analysed to detect infrastructural damage. These investigations are based on sensor data from the Measurement Tram 2.0, which is operated in regular service. Classical methods from the Fourier and Cepstrum domains are employed to extract relevant features of the sensor signals [1]. With the help of a reduced representation [2], these features are made available to data-driven methods from the field of machine learning. In addition, statistical data on travel speeds and acceleration processes of streetcars are evaluated for individual track sections in order to obtain utilization profiles of the rail network and identify areas with high loads. A fleet of additional trams with a reduced sensor setup are used to monitor tram crossings of these track sections. Relevant sensor data statistics are combined with finite element simulations to estimate contributions to rail fatigue. Finally, opportunities for improvement in the area of data collection during inspections are discussed in order to improve the data foundation for the application of machine learning approaches. REFERENCES [1] Niebling, J., Baasch, B. and Kruspe, A. „Analysis of Railway Track Irregularities with Convolutional Autoencoders and Clustering Algorithms“. Dependable Computing - EDCC 2020 Workshops. Hrsg. von Bernardi, S. et al. Cham: Springer International Publishing, S. 78–89, 2020 [2] Heindel, Leonhard, Hantschke, Peter and Kästner, Markus. „Fatigue monitoring and maneuver identification for vehicle fleets using a virtual sensing approach“. International Journal of Fatigue, 2023