A One‐Class SVM‐Based Approach for Crossing‐Gate Rod Breakage Detection in a Railway Telemeter System

Abstract

AbstractShikoku Railway Company (JR Shikoku) has installed a telemeter system network across all railway lines in the Shikoku area for collecting real‐time equipment data. In a previous study, we proposed a supervised machine learning‐based method to detect crossing‐gate rod breakages from the big data collected by the telemeter system. However, this method requires past data, including breakage cases of each crossing‐gate rod, for training. To avoid the past‐data problem, we propose a method that applies a one‐class support vector machine (SVM) with unsupervised learning to the rod breakage detection. The detection results were evaluated based on performance on real‐word railway data. © 2023 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.