Preliminary development, simulation and validation of a weigh in motion system for railway vehicles

The development of efficient Weigh In Motion (WIM) systems with the aim of estimating the axle loads of railway vehicles in motion is quite interesting both from an industrial and an academic point of view. This kind of systems is very important for safety and maintenance purposes in order to verify the loading conditions of a wide population of vehicles using a limited number of WIM devices distributed on the railway network. The evaluation of the axle load conditions is fundamental especially for freight wagons, more subjected to the risk of unbalanced loads which may be extremely dangerous both for the vehicle safety and the infrastructure maintenance. In this work the authors present the development, the simulation and the validation of an innovative WIM algorithm with the aim of estimating the axle loads $\widehat{N}$ of railway vehicles (the axle loads include the wheelset weights). The new estimation algorithm is a general purpose one; theoretically it could be applied by considering as input different kinds of track measurements (rail shear, rail bending, sleepers with sensors, etc.) and could be easily customized for different kinds of signals. In the paper a benchmark case based on rail bending measurements is proposed in which the longitudinal deformations ε _xx measured on the rail foot through strain sensitive elements are used as input. The considered input is affected by noise and bandwidth limitations and, consequently, is a good benchmark to test the robustness of the new algorithm. To estimate the axle loads, the algorithm approximates the measured physical input through a set of elementary functions calculated by means of a single fictitious load moving on the track. Starting from the set of elementary functions, the measured signal is then reproduced through Least Square Optimization (LSO) techniques: in more detail, the measured signal is considered as a linear combination of the elementary functions, the coefficients of which are the axle loads to be estimated. Authors have also developed a physical model of the railway track. The model consists of the planar FEM (finite elements method) model of the infrastructure and of the two-dimensional (2D) multibody model of the vehicle (the effects of lateral dynamics are treated as disturbances) and takes into account both the coupling between adjacent loads moving on the track and the vehicle dynamics. The physical model of the track and the innovative WIM algorithm (both considering possible measurement errors) have been validated by means of the experimental data kindly provided by Ansaldo STS and have been implemented in the Matlab and Comsol Multiphysics environments. In particular the model of the railway track has been developed expressly to test the WIM algorithm with a suitable simulation campaign when experimental data are not available; in other words it provides simulated inputs to test the WIM algorithm when there are no experimental inputs.