Data-driven model development using Support Vector Machine for railway Overhead Contact Wire maintenance

Abstract

The fault on Overhead Contact Wire (OCW) causes traffic disruptions in railway transportation. The inspection and maintenance on OCW is inconvenient as it locates highly above the ground. Condition monitoring system has been installed to monitor the status of the OCW automatically. Huge amount of data have been collected by these systems. However, a general problem of these systems is the analysis on these data is weak. The condition monitoring system has been devalued due to this. This paper aims to develop data driven Support Vector Machine (SVM) model to extract useful information from the huge amount of data. This data driven model evaluate the OCW status without requiring understanding the physical mechanism of the fault. Maintenance points are suggested based on the output of the SVM model.