Methodologies to Distinguish Locomotive Types Based on the Measured Geometry

This chapter presents methods to sort locomotives and other certain types of vehicles by types using geometry measured by a measurement station. The approach is implemented using a series of artificial locomotives created by adding normally distributed noise to each distance between adjustment wheels of locomotives that are in exploitation in Norway. Two approaches are studied. The first approach, Gaussian Mixture Clustering, is presented and shown to be suitable to obtain geometries of the main locomotives in the set. The second approach, based on the comparison of norms of differences between geometries of measured and known locomotives, is found quite accurate to use for sorting measured locomotives by the predefined types. The methods are presented on geometries of locomotives, but they also can be used for distinguishing different types of wagons and multiple units.