Arc Detection and Classification in Pantograph Catenary Systems by the Use of Clustering Techniques

In this paper the authors employ advanced signal processing techniques, namely clustering algorithms, to the analysis of currents and voltages collected by high speed trains in order to detect the presence and the quantity of electric arcs in the pantograph – catenary current collection system. The analysis is performed on a set of data recorded during different test runs; the data include voltage, current and a phototube signal detecting the presence of electric arcs. The proposed analysis groups the data in different clusters, which are then related to the real presence of arcs by the use of the phototube data. The results show the physical meaning of the clusters and the potential of the technique for its use into a preventive maintenance technique.