Skip to content

Anomaly detection using Support vector machines on overhead contact wire

Published: 21 August 2017

The Support Vector Regression (SVR) is used to model the dependency between vertical acceleration and the other factor such as uplift, train speed, height of the wire. Correlation is used to find the significant factors which influence the vertical acceleration.

Sponsor: Trafikverket/JVTC

Researchers: Yuan Fuqing, Uday Kumar

Duration: 2013

This project describes an anomaly detection method on the Overhead Contact Wire (OCW) in electrified railway system. The fundamental basic of contact wire is described. Their mechanical property and thermal property are discussed. The principle of the current collection through the overhead wire is described in brief. Some classical mechanic dynamic models between the pantograph and overhead contact wire are presented. Concentrating on the anomaly detection using vertical acceleration signal, this report proposes a support vector regression based method to detect the anomaly detection on the surface of the overhead contact wire. The Support Vector Regression (SVR) is used to model the dependency between vertical acceleration and the other factor such as uplift, train speed, height of the wire. Correlation is used to find the significant factors which influence the vertical acceleration. The SVR model is used to de-trend the vertical acceleration signal. The statistical model is proposed to find the anomaly points on the contact wire. All the method and model are implemented using Matlab. This Matlab software is GUI based and is attached with this report.