A neural network approach to estimate the top speed and trainset configuration of a high-speed train

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 609
  • Download : 532
The prediction of top speeds of a high-speed train and the configuration design of train formations has been studied using a neural network with backpropagation. High-speed trains, such as KHST-K, KHST-11 and KHST-20 which will be operated in Korea, are used for the teaching data of the neural network. Other high-speed trains of other countries, such as ICE, Shinkansen and TGV, are also used for the data of neural networks. The input data for top speed prediction consist of values that are closely related to traction force, propulsion resistance and train formation. KHST-20 test trains are composed of power cars and passenger cars and have modified train formations. The top speeds have been predicted using neural networks. The configuration design of a train formation has also been performed using neural networks with backpropagation, which determine the basic train formation, such as weight, power, the number of cars, the number of motor cars and the number of traction motors in a high-speed train. The train formation of the KHST-20 test train is configured by the neural network and compared with the actual train formation. The design of Shinkansen 700 and KHST-20 is also configured using neural networks. In the configuration design of KHST-20, two types of train were used for the teaching data of the neural network. These were the power concentrated train and the power distributed train. Neural networks of each set of data estimate very different train formations from the same input value.
Publisher
PROFESSIONAL ENGINEERING PUBLISHING LTD
Issue Date
2004-03
Language
English
Article Type
Article
Keywords

DESIGN

Citation

PROCEEDINGS OF THE INSTITUTION OF MECHANICAL ENGINEERS PART F-JOURNAL OF RAIL AND RAPID TRANSIT, v.218, no.1, pp.41 - 49

ISSN
0954-4097
URI
http://hdl.handle.net/10203/1751
Appears in Collection
ME-Journal Papers(저널논문)
Files in This Item

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0