Klassifikation von Oberbaufehlern am Beispiel Weichen

Abstract

By extending a recently published method for forecasting of single failures in track, based on continuous track monitoring data within the AutoMain project, the German Aerospace Center (DLR) presents a method, which proofs, that different kinds of single failures can be separated from each other. While state of the art assessment of track in Germany is based on a limit value consideration about the failures amplitude, the failures type is neglected. This work shows that with a given reference dataset of a known type of failure, a classification can be achieved, which aims to optimise the previously presented forecasting algorithms by developing type specific prognostic models for each failure type. This method can contribute to improved maintenance efficiency in terms of better prognostics and optimised preparation for repairs by anticipating the failure type before seeing the worksite.