A Neural Network-based Prediction Approach of Relationship between Rail Wear and Gross Traffic Tonnage

In order to study the relationship between rail wear depth and gross traffic tonnage on the Beijing-Shanghai high-speed railway, and to determine the rail grinding period more accurately, the rail wear-gross traffic tonnage (RWGTT) neural network prediction model based on the TensorFlow was established firstly, to realize the prediction of the gross traffic tonnage by inputting the data of the wear depth of rails, the speed of trains, axle load and wheel-rail profile. Then the simulation model of rail wear was established, including the dynamics simulation module, the wear calculation module and the rail profile update module. Moreover, the calculation results of the simulation model were employed as the training and test data of the prediction model. The results show that: the RWGTT prediction model can accurately predict gross traffic tonnage by inputting parameters such as rail wear depth, with the prediction accuracy ranging from 93.28% to 97.98%. The prediction model can provide a theoretical basis and reference for the accurate determination of the grinding period of rails on the Beijing-Shanghai high-speed railway. By comparing and analyzing the training results of prediction models with different numbers of neurons in hidden layers, it can be concluded that the mean square error (MSE) of the prediction model fluctuates with the change of the number of the neurons in the hidden layer. Consequently, in order to improve the prediction accuracy of the neural network, reasonable selection of the number of neurons in the hidden layer is required. With the increase of gross traffic tonnage, the rail wear depth and the radius of curvature of the rail head increase accordingly. © 2021, Department of Journal of the China Railway Society. All right reserved.