Experimental Measurement and Numerical Correlation of the Brake Stopping Distance on a Local Railway

Railways are complex systems, and the braking performance of trains is crucial to ensure the line’s safety. The assessment of the stopping distance can be obtained by empirical formulas or by more sophisticated numerical models which simulate the train’s longitudinal dynamics. In both cases the level of estimation accuracy may vary a lot, depending on the numerical values attributed to several parameters related both to the train braking system and the railway. The correct identification of such parameters might be an issue in local railways. On the one hand, some widely used empirical formulas are mainly intended for national railways, thus it is critical to determine whether they are appropriate for local railways. On the other hand, the development of a good predictive simulation model requires the identification of parameters not always known for the trains in service. Starting with measurements taken from braking tests performed on a local railway, this research aims to propose an experimentally correlated dynamics model based on the train’s equation of motion that can accurately estimate the stopping distance with a reduced amount of input parameters obtainable from measurements. Several braking tests have been performed on the track to identify the model’s parameters and to enhance the numerical–experimental correlation. Meanwhile, the applicability of stopping distance empirical formulas to the case of local railways has been evaluated, and the parameters of these formulas have been identified to reduce the gap with respect to measured distance values. Even if both simulation approaches led to an accurate estimation of the stopping distance, this work highlights some distinctions. In order to match the measured distance values, some empirical formulas required the definition of doubtful input parameters values, and suggest skipping their use for local railways application. Conversely, the proposed dynamics model led to a good balance between accuracy level and the effort required for parameter identification from testing, with it being more easily applicable to different scenarios and open to the implementation of additional features in future studies.