Research Note A New Model for Predicting Vehicle Derailment in Switch Track Using Neural Network

Switches in railway tracks have discrete and complicated structures. Due to this, these points are one of the most dangerous parts of a railway track. As an operational and maintenance approach it is important to evaluate and predict derailment probability in these points. As a train-track interaction problem, this phenomenon is influenced by switch geometry, switch structure and dynamic characteristic of railroad vehicle like velocity and axle load. Therefore it is important to develop a new model that contains these factors and be able to predict vehicle derailment with reasonable accuracy. In this research, by using the sensitivity analysis in described parameters, a derailment model is presented. For this goal, dynamical modeling of the chosen both freight and passenger wagon in switch track using ADAMS/RAIL is developed and the effect of the switch geometry, switch structure and operational parameters of the vehicles are studied in vehicle derailment. With the help of the neural network, the optimum mathematical model that describes the vehicle derailment and effect of the main parameters is determined.