Research On Key Technology About Inspecting And Monitoring Data Analysis Of GSM-R For High Speed Railway

With the advent of big data, technologies for data mining andinformation sharing have become significant approaches for knowledgediscovery, wealth creating and efficiency improving. In the course ofrapid development of high speed railway technology, GSM-R network has beenevolved into an indispensable part for transmission of train controlcommunication information. This paper focuses on the application of datamining technology in GSM-R network inspection and monitoring data on highspeed railway. Through in-depth analysis on GSM-R network inspection andmonitoring data, it provides effective solutions to comprehensiveevaluation of GSM-R network QoS inspection results, how inspection datareveal the principle of train control communication, and automaticallydiagnosing of C3(CTCS-3) degrading classification. Main contributionsare listed as follows:Firstly, it provides a comprehensive evaluation method of GSM-R QoSinspection results compared to current independent evaluation ofinspection indicators respectively which lacks the comprehensive view.Using K-means clustering algorithm and KNN classification algorithm, oneevaluation model is constructed based on historical inspection data. Atthe same time, another model which completely based on technicalspecification is built. And then, give the final evaluation by combingthe two models’ evaluation results. Experiments better illustrate thisprovided method could reflect the real situation of inspectionreasonably.Secondly, data analyzing on original GSM-R QoS inspection data:(1)A relationship model which can be used to calculate the real train controldata transmission delay caused by interference time is proposed based onthe similarities between CSD interference test data transmission and realtrain control data transmission.(2) According to the relationship model,the communication differences of three most popular terminals adopted in CTCS-3train control system are compared through CSD interference testdata. And effective suggestions are provided both on terminal choosingand GSM-R network dynamic inspection.(3) Problems of frame continuousloss in CSD transmission delay test are solved, and aiming at the flawsin current GSM-R communication inspection system, some effectiveimprovements are proposed.Thirdly, based on researches on application of data mining technologyin C3degrading, the framework and process of the C3degrading automaticdiagnosis system are proposed, and solutions to three key technologiesare provided for system implement.(1) C3degrading expression model:19typical on-board and groundcommunication disconnection types and their principles are summarizedthrough plenty of research on the C3degrading cases, which lays thefoundation of C3degrading classification. The whole C3degradingattributes extracted from interfaces monitoring data is considered asconditional attribute set. And the set of decision is formed of the C3degrading classification. Both the conditional attribute set and thedecision attribute set formulate the C3degrading expression model basedon interface monitoring data.(2) C3degrading automatic diagnosis model: the automatic diagnosismodel based on C4.5decision tree is introduced. It could be optimizedto improve classification accuracy rate by implementing an algorithmnamed decision table reduction based on conditional information entropy.The automatic diagnosis model is trained by real C3degrading cases, andexperiments illustrate that the classification accuracy rate can reach98.579%.(3) C3degrading automatic diagnosis model renew: the automaticdiagnosis classification model renew could be achieved easily throughrebuilding decision tree using updated training sets of the C3degrading.In this paper, through researches on GSM-R network inspection andmonitoring data analysis, it provides new approaches on comprehensive evaluation of GSM-R network QoS inspection results, and diagnosing C3degrading classification automatically.