Braking Modeling And Parameter Identification For High Speed Train

With the continuous improvement of running speed and the complexity and changeability of running environment,the dynamic working environment of train system has deteriorated significantly.It brings great hidden dangers to the safe and stable operation of high speed trains.As an important part of ensuring the safe operation of high speed train,the braking performance of train braking system should be paid great attention.Accurate dynamic modeling is the basis of precise braking of high speed train.Through real time accurate estimation of performance parameters of model,real time performance of train can be mastered,dispatching efficiency of train section can be greatly optimized,measuring cost of train can be reduced,and maintenance efficiency of train can be improved.However,the existing research mainly focuses on the dynamic model of the train itself,without considering the impact of environmental factors and for the nonlinear nonGaussian system such as high-speed train,there is no perfect identification theory to effectively identify its braking parameters in real time.To solve these problems,on the one hand,this paper analyses the train braking mechanism,studies and establishes the environment based train braking model,and identifies the braking model under nonGaussian noise interference and the braking model with hidden variables,and puts forward the corresponding braking model identification method from the view of model point;on the other hand,the train monitoring data are analyzed and constructed.The vertical depth neural network is used to identify the braking parameters from the view of data point.The main work and research results of this paper are listed as follows:1?The mechanism of train braking is analyzed,and the single and multi particle models of train braking are established.Considering that the running state of the train will be affected by the running environment in the actual running process,the different motion characteristics of the train running on the dry track and the wet track are analyzed,and the single particle braking modeling method under different operating conditions is proposed.2?Aiming at the problem of system identification and parameter estimation of high speed train nonlinear model under nonGaussian noise,an extended Kalman filter basecd on Gaussian sum method is proposed and implemented.The nonGaussian noise of train is represented by Gaussian sum.multiple EKFs are used as sub filters,and the filter estimation of state and Parameters is obtained by parallel computing.3?Aiming at the hidden variable parameters,which are difficult to observe in the train braking model,a sliding window for data recording and counting is designed.Taking the data in the window as the research object,the condition expectation of train is constructed and maximized,and an online identification method of braking parameters based on the maximum expectation is proposed.4.Taking mass monitoring data of trains as the research object,a time varying braking parameter identification method based on BP network is proposed.Considering the time series data characteristics of train speed and friction coefficient of brake disc,a time varying braking parameter identifiation method based on LSTM network is proposed,which realizes the prediction of future changes of friction coefficient of brake disc by historical tata of train braking speed.