Dynamic Time Warping to Enable Health Monitoring of Legacy Pneumatic Railway Door Systems

Sammanfattning

To ensure optimal efficiency, operational safety, and reliability for its passengers, a train wagon’s door system is a crucial subsystem. To enhance operational safety and reliability, it is essential to categorize and monitor the faults and failure modes of the door system. This paper delves into the door system of older rolling stock utilised in Swedish traffic by SJ. The feasibility of a health monitoring system based on differences in time series measured by the dynamic time warping algorithm (DTW) is studied, developing the system on a digital twin for use on physical doors. A case study was conducted, focusing on the pneumatic door system and its opening and closing sequences through defining common, generalisable fault modes, chosen as increased resistance and cylinder leakage. Data was gathered from the physical doors on a wagon provided by SJ and further failure modes were studied through the use of a digital twin. Neural networks were then trained on data from the digital twin and validated on data gathered from real doors. The findings highlight the feasibility of DTW as a basis for health monitoring, achieving a classification accuracy of about 88% on simulated data. For real-worlddata, however, the findings are inconclusive due to difficulties validating the digital twin to the physical doors and signal noise, affecting values calculated by the DTW algorithm.