Developing and Extending Status Prediction Models for Railway Tracks Based on On-Board Monitoring Data

Abstract

Assigning inspection trains to monitor track quality is a standard procedure for maintaining railway system safety. The main challenges lie in lacking time and resources to perform the inspections because of the increasing traffic nowadays. To overcome these challenges, many consider adopting the on-board monitoring (OBM) technique for performing the inspections. This technique assigns commercial trains, instead of traditional track recording vehicles (TRVs), to monitor the track status, allowing railway operators to perform more inspections without affecting the traffic and using expensive inspection trains as well. However, compared with TRV data, the new OBM data are of lower data quality and have fewer features, although they can be recorded more frequently. Therefore, new methods should be developed for effectively applying the new data. This study develops four models, namely the linear regression model, Markov model, ordinary Kriging model, and Kalman filter model, for predicting the track status based on the OBM data. Data collected from the Switzerland railway network are used for verifying the models. Results show that the proposed models can effectively predict the degradation of the track status in different ways and, therefore, assist railway operators in scheduling maintenance tasks.