Machine learning algorithms for the prediction of the mechanical properties of railways’ rail pads

Abstract

Train operations generate high impact and fatigue loads that degrade the rail infrastructure and vehicle components. Rail pads are installed between the rails and the sleepers to damp the transmission of vibrations and noise and to provide flexibility to the track. These components play a crucial role to maximize the durability of railway assets and to minimize the maintenance costs. The non-linear mechanical response of this type of materials make it extremely difficult to estimate their mechanical properties, such as the dynamic stiffness. In this work, several machine learning algorithms were used to determine the dynamic stiffness of pads depending on their in-service conditions (temperature, frequency, axle-load and toe-load). 720 experimental tests were performed under different realistic operating conditions; this information was used for the training, validation and testing of the algorithms. It was observed that the optimal algorithm was gradient boosting for EPDM (R2 of 0.995 and mean absolute percentage error of 5.08% in test dataset), TPE (0.994 and 2.32%) and EVA (0.968 and 4.91%) pads. This algorithm was implemented in an application, developed on Microsoft .Net platform, that provides the dynamic stiffness of the pads characterized in this study as function of material, temperature, frequency, axle-load and toe-load.