Analysis Of EMU Wheels Application And Testing Data

With the higher speed of the train, dynamic interaction between the train and the line in high-speed condition significantly increase. On the one hand, the damaging effects of the high speed train on the line was aggravated; on the other hand, the influence of line conditions on vehicle driving safety and ride comfort are more obviously. With years of high-speed railway operating and development, the vehicle maintenance department has accumulated a large number of vehicles relevant data. It is necessary to dig the knowledge behind the data. Get wear law of wheel or the relation between wheel profile parameter and the vehicle stability. Assess or predict the matching state among the vehicle and the rail. To guide the vehicle maintenance department how to use and maintenance the wheelset.In this paper, for the CRH3C EMU running on the Wuhan-Guangzhou high-speed railway occurs lateral instability problem(bogie lateral acceleration alarm). Collected and collated the bogie instability data from 2014 March to August. Summarizes the LY System (Wheel shape detection subsystem in wheel fault dynamic detection system) test data and related EMU maintenance data. Use the BASIC programming language, developed the data structure transformation software, transform the unstructured vehicle diagnostics text data of operation and maintenance data into structured data. Use data warehouse theory and methods build the of LY test data, bogie instability data and EMU operation and maintenance data to a new wheelset data system. By analyzing the data and operating the new system of wheelset multidimensional data, obtained the data sample set for further data mining.According to the data mining business goals, that is to find LY system detects the law between data and frame instability. Determined the basic idea of mining the wheelset detection data sample set, according to the classification method of data mining classified each data object according to whether occurs frame shake. According to this idea, as 1:6 ratio choose the data both occurred and did not occurred frame shaking integrated wheel data mining sample set. There are 54 data attributes of each data object. Also check and improve the errors or missing data of the sample set. Use WEKA data mining platform pretreatment the data sample. We screened a number of different sizes subset of attributes. By rule-based classification algorithm, decision tree classification algorithm and integrated learning, established the model on different attributes subset. Compared the different subset of the sample properties and classification algorithm modeling training error and generalization error. Obtained the maintenance department needs and easy to understand rules-based classification model.Analyzed the above classification model and the conclusions as follow:large flange thickness can significantly affect frame lateral instability; the change of wheel inside distance (or rail central distance) and a larger tread wear parameters has great effect on the framework lateral instability; the A frame wheel diameter difference forms (in-phase or reverse) and too large inverted wheel diameter difference of non-power car B frame is also affects the frame lateral instability;Finally, this paper describes the role of classification model for wheelset operating quality control and maintenance decision-making. Analyzed the shortage of wheel inspection probe data during mining process, and put forward some suggestion on make better use of the accumulated historical data in the future and provide service on site production managemen.