Research On The Data Analysis Method Of Online Electromagnetic Rail Defect Inspection Based On Cloud Platform

With the development of China High-speed Railway technology and the accumulation of its mileage, the matter of train safety is taken more and more seriously, which requires that the rail completeness test has become an essential part in order to guarantee safety. Currently, the Railway Department employs ultrasonic inspection vehicle as the main equipment to detect the damage. But it not only is slow and ineffective, but also gives indirect inspection results, which means it fails to meet the demand for online high-speed damage-inspection. With the background of high-speed online electromagnetic rail defect inspection technology, this essay focuses on illustrating the details of the process how the software server first collects the data from the rail defect inspection equipment, then turn it into to data with fixed form and presentation, and finally analyze the data.This essay includes the following parts. It introduces the design framework of online electromagnetic rail defect data management system. Inspection equipment examines the electromagnetic signals reflected by the rails with the front probe, and then receives demodulated data transferred by demodulation. The data-collection server program receives and records data, and shows the rail damage with real-time waves.The data processing program would format the data recorded in the database, selects the data with rail damage and reserves the fixed data span, adds the information such as the route, timestamp, coordinates of latitude and longitude to consist a complete data frame format. At last, the program would send the data to the collection program and preserve in the database.Online rail defect data management system loads the data table of real-time rail damage and the data table of history damage of formatted data structure in database, demonstrating the data in graphs on the website, so that the control staff could monitor the rail damage situation easily.The server receives the original rail damage data during experiments in the lab, withdraws rail damage features after data pre-process, and confirms the classification standards considering both the actual damage and clustering algorithm.Due to the limited damage samples in the experiment, based on the experiment data, this essay will simulate the damage data that agrees more with the real situation, pre-processes the simulated data, analyzes data with clustering, classification and dimension reduction algorithm, and finally evaluates the function of different algorithms.