Dynamic Time Warping and Spectral Clustering Based Fault Detection and Diagnosis of Railway Point Machines

Abstract

Point machine plays a very important role in metro operation. Among these infrastructure failures in urban rail transit systems, vast majority of them are triggered by railway point machines. Thus, fault detection and di- agnosis should be well concerned to ensure traffic safe- ty. In this paper, we propose to employ dynamic time warping and spectral clustering to handle this problem. Firstly, dynamic time warping method is used to com- pare unequal sequences with phase-shifted shape. Sec- ondly, spectral clustering method is applied to deal with the classification problem without training steps. At last, simulation results demonstrate well performance of the proposed scheme.