Monitoring and tracking of a suspension railway based on data-driven methods applied to inertial measurements

Abstract In this study, the dynamic response of the suspension railway system Skytrain Dusseldorf is exploited in order to localize the train and monitor the infrastructure based on the proposed structural health monitoring methodology. The Skytrain is a driverless monorail train, which operates at the Dusseldorf airport in Germany and connects the railway station with the main terminals. An inertial measurement unit is mounted on the bogie of the Skytrain that continuously acquires in-service data about the vehicle movement. Based on the train–track interaction, the dynamic response will be directly transmitted to the sensor and unique signal features can be extracted. In this context, the dynamic behavior of the train is used to identify turns and stops along the track. Computational intelligence learns from the operational data and can recognize features of the characteristic track profile. In order to localize the vehicle precisely, the proposed methodology applies a k-means clustering algorithm to label the field test data and an artificial neural network to classify the individual track sections. Each prediction is analyzed by an autoencoder in order to detect anomalous vehicle movements, which will be removed from the damage evaluation to avoid false alarms. Additionally, infrastructure monitoring is conducted on a side track of the Skytrain Dusseldorf. The Skytrains test track is modified with a broken bolt connection, which affects the dynamic response of the bypassing train. A damage index is introduced to evaluate the different track conditions and detect the damage along the side track. Furthermore, the results are compared to a one-class support vector machine, which represents a generalizable method for damage detection based on unsupervised anomaly detection. The outcome of the proposed work can be used to optimize the maintenance planning and ensure a high level of reliability and safety.