Intelligent condition monitoring of railway catenary systems : A Bayesian Network approach

This study proposes a Bayesian network (BN) dedicated for the intelligent condition monitoring of railway catenary systems. It combines five types of measurements related to catenary condition, namely the contact wire stagger, contact wire height, pantograph head displacement, pantograph head vertical acceleration and pantograph-catenary contact force, as inputs based on their physical meanings and correlations. It outputs an integrated indicator of catenary condition level. The BN parameters are learned from historical measurement data. Preliminary results shows the applicable ability of the BN to integrate multiple types of parameter while make sense of the output to facilitate maintenance decision making.