1103 ニューラルネットワークを用いた車輪・軌道連成系の物性値同定(車輪/レールの接触と動特性)(OS20 車輪-レール系)(シンポジウム : 第7回鉄道技術連合シンポジウム(J-RAIL 2000))

概要

論文の詳細を見る
Identification of parameters used in the wheel/railway dynamic interaction analysis is attempted. The proposed method is constructed by means of the neural network (NN). In the present problem, the unknown parameters, such as damping constants of the rail and sleeper pads, and shear moduli of the ballast and concrete foundation, are considered as the output data of the NN. As the input data, the spectrum ratios of sleeper/rail and foundation/rail are employed. Learning of the NN is accomplished based on the solutions obtained by numerical model. The wheel/track interaction analysis method is constructed with integral equation and finite element methods. Through numerical experiments, validity of the proposed identification method is investigated.
2000-12-12