The Prediction of Derailment Coefficient Using NARX Neural Network
- 16 June 2011
- conference paper
- Published by American Society of Civil Engineers (ASCE)
- p. 2235-2244
Abstract
In railway system, track inspection vehicles regularly measure the vertical and lateral irregularities of each rail in order to ensure the safe running of railway. Derailment coefficient is an important indicator of the vehicle safety when the vertical wheel-rail force and lateral wheel-rail force are integrated together. In this paper, NARX neural network is proposed to predict derailment coefficient by measuring the track irregularities. In order to improve the generalization of the neural networks, the Bayesian Regularization algorithm is employed to train the neural networks. The experiments are carried out and the results show that the NARX neural network with Bayesian Regularization algorithm can predict derailment coefficient accurately.Keywords
This publication has 12 references indexed in Scilit:
- State-of-the-Art Railway Vehicle Design with Multi-Body SimulationJournal of Mechanical Systems for Transportation and Logistics, 2010
- Railway track quality assessment method based on vehicle system identificatione+i Elektrotechnik und Informationstechnik, 2009
- Long-term time series prediction with the NARX network: An empirical evaluationNeurocomputing, 2008
- A new approach to the analysis and presentation of vertical track geometry quality and rail roughnessWear, 2008
- An inverse railway wagon model and its applicationsVehicle System Dynamics, 2007
- A new monitoring method of train derailment coefficient2006
- Bayesian regularization neural networks for optimizing fluid flow processesComputer Methods in Applied Mechanics and Engineering, 2006
- Improving performance in pulse radar detection using Bayesian regularization for neural network trainingDigital Signal Processing, 2004
- Simulation of wheel–rail contact forcesFatigue & Fracture of Engineering Materials & Structures, 2003
- Gauss-Newton approximation to Bayesian learning2002