Research On Prediction Model For Rail Track State

Security is always the subject of rail transport, and the track state is one of the most important influence factors for running safety. Rail track geometric irregularities are the concrete embodiment of track structure deterioration, and give an expression to the comprehensive performance of track members. To guarantee the security of railway transportation, the research on the change state of railway track must be thoroughly put in practice. As a result, the future value of track irregularity would be predicted and the future quality status of the track would be accurately grasped.The main researches and results are as follow:First of all, the original track irregularity detected data is preprocessed. For outliers and mileage drift two problems existed in original detected data, a reasonable way for excluding outliers and proofreading mileage are explored and determined. The outliers are excluded by the method of absolute average. Two batches detected data with mileage drift is proofread according to that the square sum of difference of two batches detected data is minimal. Through preprocessing, we can effectively reduce the error in the track irregularity detection data and provide a correct data source for reasonable and effective research of the prediction model. Then, the prediction model for track irregularity state is established. Based on linear prediction model proposed by Associate Professor Xu of Tongji University, the author propose piecewise linear prediction model and the piecewise linear-Markov prediction model by learning and analyzing a large number of the original test data in a certain railway line deeply, then establish the TQI prediction model using data detected in the line and examine the model. Through in-depth analysis of the prediction model of track state home and abroad, based on the full study through linear prediction model, the disadvantage of this model is fund. The disadvantage is that the linear model is less applicable for those sections where TQI sequence is with abrupt features. To conquer it, piecewise linear prediction model is adopted to predict and analyze the track irregularity. The development and change of track irregularity under the comprehensive influence of many complex factors tend to have a great fluctuation. Against to this character of the development of track irregularity, residual modification model is adopted to modify the piecewise linear prediction model and make it close to the development change rule of true TQI. At last, the improved linear prediction model proposed by the author is compared with gray prediction model and gray-Markov assembled prediction model. Variation characteristics of track geometric irregularity reveal the general regularity of track deterioration. A large amount of track geometric irregularity dynamic test data is used in this thesis to examine TQI prediction model by practical example, and the experimental results show that the improved linear prediction model proposed by the author has a better prediction effect.