Study On Performance Parameters Estimation Of High Speed Train

Service environment, such as increased vibration, accelerated wear and tear, enhanced noise etc, have become more complicated with the continuous improvement of speed because the dynamic characteristics of train at high speed are significantly different from some at low speed, especially, wheel-rail interaction force, air role and bow net-effect have further strengthened.The stability, reliability, security of high-speed train in service will put forward higher requirements in response to these environmental factors,at the same time, The maintenance and nursing of high-speed trai have proposed new requirements. Therefore, the deepening study of service performance under its evolution is of great significance. Since the mechanical properties of high-speed trains depend on structural parameters, to some extent, the study of service performanc evolution is equel to study parameters evolution. During the process of studying relationship between arameters with their evolutio and Miles of service, between dynamic parameter and performance degradatio, the most important step is parameter identification.At present, the relevant study of parameter identification of high-speed train at home and abroad is in the initial stage, the parameter identification theory and technology is also not mature, there is less research and thesis. Many key technologies, such as reasonable simplication of mechanical model, conversion from high-dimensional mechanical model to high-dimensional probability model, analysis of parameter identification and structural features of the probabilistic model and information flow, probability theory for parameter identification, suitabilities between parameter identification model and identification method did not elaborate on the system. At the same time, the key performance indicators, as the accuracy, quantity, timeliness of parameter identification etc put forward higherand newer requirements. Therefore, this study identifies the specific process of parameters identification for clues, regards the relevant model of parameter identification, parameter identification theory and its derivation, parameter identification model, theory and experimental verification of parameter identification theory and its model as the main study contents, regards performance indicators improvement of adaptability, identification accuracy, speed and etc as the main targets.In response to these key issues, this paper carries out related research work, mainly reflected in:(1) The characteristics, advantages and disadvantages of different physical parameters identification method are analyzed on the base of research status about physical parameters and method at home and abroad. The parameter identification of high-speed train is a multiparameter, high-dimensional state, the initial value sensitive, complex multi-modal nonlinear problems, through analysis of the structural characteristics of high-speed train. The main development direction that parameter identification of high-speed train is based on mixture filtering theory is pointed out, which can make full use of the unique advantages sloving linear state inference by the Kalman filter and unique advantages of particle filter, sliced Gaussian mixture filter for nonlinear Problem, initial treatment of sensitive issues.the result is the formation of complementary advantages.(2) The mechanical model of high-speed train is constructed by mechanics,On the base of technical features analysis of the four discrete methods for continuous, the defined method is highly adaptable mechanical model of high-speed train and further to determine the discrete frequency,fianally the state transition equation is established. The supporting evolutionary equation of parameter identification is introduced. The observation equation is constructed by sensor configuration. On this basis, probability model of parameter identification is a high-dimensional conditionally linear model through analysis of information flow structure of the probabilistic model.(3) The methods and models of parameter identification based on extended kalman filfter and Rao-Blackwellised particle filter are proposed on the base of the analysis of the basic theory and technical characteristics.the methods and theory of parameter identification are tested and verified from the view of theoretical and experimental poin.In another aspect, performance comparison is carried ou.(4) The method and model of parameter identification based on modified Rao-Blackwellised particle filter is proposed on the base of modification and improvement of traditional Rao-Blackwellised particle filter.An example to verify the validity of the method is executived. Parameters identification performance based on modified Rao-Blackwellised particle filter are compared to traditional Rao-Blackwellised particle filter.(5) The parameters identification method and model based on the sliced gassian mixture filter modified from the low dimension to high-dimensional case by the dirac-mixture tree and gaussian mixture reduction on the base of carding of basical Gaussian mixture filter.The numerical results verify the effectiveness of the method. The parameters identification performance based on the sliced gassian mixture filter is compared to Rao-Blackwellised particle filter and modified Rao-Blackwellised particle filter.