Statistical identification of geometric parameters for high speed train catenary

Abstract

Pantograph/catenary interaction is known to be strongly dependent on the static geometry of the catenary, this research thus seeks to build a statistical model of this geometry. Sensitivity analyses provide a selection of relevant parameters affecting the geometry. After correction for the dynamic nature of the measurement, provide a database of measurements. One then seeks to solve the statistical inverse problem using the maximum entropy principle and the maximum likelihood method. Two methods of multivariate density estimations are presented, the Gaussian kernel density estimation method and the Gaussian parametric method. The results provide statistical information on the significant parameters and show that the messenger wire tension of the catenary hides sources of variability that are not yet taken into account in the model.