The Identification Method Of Dynamic Model For High-speed Train

In recent years, with the rapid development of high-speed railway in our country and theincrement of train speed, aerodynamic problems are more obvious and pantograph catenarydynamic flow problems are more complex, which lead a serious deterioration in train dynamicenvironment. In this case, it is a huge risk for operation safety, stability and comfort ofhigh-speed trains. Therefore, the development of basic research by effective modeling, controland optimization for high speed trains, is significant to guarantee the high quality operationand sustainable development of high-speed railway. Due to the problem of effective modelingfor high-speed train, the main content of the paper is as follows:(1) A state-space model is established to describe the dynamic behavior of high-speedtrain. Subspace identification based on weighted nuclear norm optimization is proposed formodeling dynamic system of high-speed train. Nuclear norm form is constructed based on thesubspace identification method and the alternating direction multiplier algorithm (ADMM) isused to solve the nuclear norm optimization problem. Then the state-space model ofhigh-speed train is obtained directly from input/output data.(2)Subspace identification based on least squares support vector machines (LS-SVM) isproposed for modeling general nonlinear system of high-speed train. A general nonlinearstate-space model is established to describe the dynamic behavior of high-speed train as asingle-point-mass object. In addition, by using the method of least squares support vectormachines regression, the classical subspace identification algorithms is extended to thegeneral nonlinear model of high-speed train.(3) The simulation experiments are implemented based on the constructed state spacemodel and generalized nonlinear model for high-speed train. In order to illustrate theeffectiveness of the proposed method, model validation study of CRH3high-speed train isimplemented, and the results show the enhanced performance of the proposed method formodeling CRH3type high-speed train.