Predicting Wheel Slippage in Railways using Bidirectional Recurrent Neural Networks

This paper shows the application of Artificial Intelligence to predict wheel slippage in railways.The prediction was conducted with a Bidirectional Recurrent Neural Network, and the data was produced using an experimental railcar and track.The objective is to predict slippage by only measuring the train acceleration without recording the train's speed and wheel velocity.Eighty experimental data were applied to train and validate the Neural Network, and the results show that it was possible to predict slippage using only the train's acceleration data.The method presented in this paper can be used to design a monitoring system and predict wheel wear and polygonization.