Prediction of Radio Signal Failures of Communication Based Train Operating Systems by Machine Learning Methods

Abstract

The use of rail systems in urban public transportation has become a necessity for reasons such as time saving, travel comfort and operating costs, especially in cities with high population and road traffic. Communication based train operating systems (CBTC) are used for the safe use of urban rail systems and the maximum capacity of the railway line. In this study, in line with the data collected from the trains on a railway line operated with CBTC, the status of the radio signals that enable the wireless communication of the trains with the trackside signaling equipment was evaluated by machine learning methods, and the situations that may have negative effects on the train operations of the problems at the signal level were evaluate. The problems on the antennas which receives signals from trackside above trains, the poor connection related with fiber optical and LAN cables, the trackside transmitter antenna orientation problems causes decrease on signal levels. It is aimed that to take actions about the problematic signal levels without any negative impact on the passenger comfort and the operation yet. The radio signal losses cause unexpected trains stops and delays. A decision support model has been developed that will offer early solution suggestions to system maintainers in order to intervene first. In conclusion, since it is the first study related with failure prediction by using radio signal levels data on railway signaling system, this study presents an important innovation in terms of literature.