抄録
For preventive maintenance of track, low-cost monitoring systems have been developed based on on-board sensing devices and applied for regional railway lines. However, it is difficult to deal with a large volume of automatically collected data. To enable efficient computational processing of collected data, diagnosing and predicting track conditions need to be automated. In this study, an algorithm for identifying the track conditions from car-body acceleration was constructed using machine learning techniques and verified in simulation study. In addition, we examined the possibility of diagnosing track conditions for an actual regional railway line. The field test results show that the track faults can be detected automatically by the proposed machine learning techniques.
