Design And Implementation Of Train Network Alarm Mining System

With the continuous increase of high-speed rail operating mileage and the number of EMUs,the operation and maintenance of EMUs is becoming increasingly heavy.How to improve the maintenance efficiency has become an urgent problem to be solved.In order to improve the maintenance efficiency,more and more equipment has added the state detection function to realize the alarm of fault state.However,due to the transitivity of the alarm in the train network,the root cause alarm will produce a large number of secondary alarms through network transmission,the efficiency of fault detection according to the alarm information is still not high.Therefore,it is of great practical significance to carry out alarm correlation analysis,reduce the number of alarms and dig out the root causes of alarms,so as to improve the efficiency of maintenance and reduce the cost of operation and maintenance.Based on the analysis of the alarm data generation mechanism and data characteristics,a Train Network Alarm Mining System is designed in this dissertation.The main research work is as follows:Firstly,in view of the unbalanced importance and time synchronization of train network alarm data,the weight fusion scheme based on analytic hierarchy process(AHP)and entropy weight(EW)method is adopted to balance the importance of alarm.The sliding window algorithm is used to synchronize alarm time,and the alarm transaction set suitable for association analysis algorithm is constructed.Secondly,due to the low quality of rules generated by the association analysis algorithm under the traditional support-confidence framework,the FP-Growth algorithm based on the weighted support-full confidence framework is designed in this dissertation,which can improve the proportion of a small number of important alarms and filter a large number of cross-support modes,thereby improving the quality of rule mining.Thirdly,train network alarm mining system based on.NET + MS Access architecture is designed and developed.The module design idea is adopted to design the functions,software programming and interface optimization of each module.Among them,the function of data management module is based on MS Access;the alarm data mining module uses the data preprocessing scheme and the improved FP growth algorithm to complete the correlation analysis of the alarm data;the function of alarm rule visualization module is realized by using the data visualization algorithm integrating ’ tag network ’ and ’ social network graph ’.Finally,taking an actual alarm data of a EMU as the test sample,the operation performance of each functional module under different parameters is tested,and the results meet the actual functional requirements;combined with the rule visualization module,the correlation analysis of alarm rules is carried out,and the validity of the rules is verified.After verification,the system can compress the alarms by approximately 80%;and establish a small number of rules with guiding significance,incorporation engineering experience,a complete alarm knowledge base can be formed to realize intelligent maintenance prompt and intelligent filtering of alarm information.To summarize,the proposed system can significantly improve the maintenance efficiency.