Research On Turnout Monitoring Data Visualization Oriented To Electric Operations And Maintenance

Turnout switch machine,as an important basic equipment of railway signal system,is one of the weak links of track,which has decisive significance for train safety and transportation efficiency.At present,many railway operators and equipment manufacturers have accumulated a large amount of turnout monitoring data.Scholars in the industry have done a lot of research on how to use turnout monitoring data to accurately evaluate the status of line equipment and predict equipment failure.However,the research on data visualization is insufficient.Most of them directly use simple charts to show the simple time series relationship of monitoring data.Lack analysis of historical data.Visual display of the historical state information of turnout equipment can provide important reference for fault analysis and maintenance decision-making of turnout equipment.As an unsupervised learning,clustering analysis can express the difference between abnormal data and other normal data without knowing the fault types of turnout monitoring data clearly.In practice,the working state of turnout is affected by many factors,such as internal and external environment,which may lead to new fault types,which can also be reflected by clustering analysis.The main work of this thesis is as follows:Firstly,the clustering results can directly reflect the spatial distribution of data through scatter plots,so as to understand the overall situation of turnout monitoring data from a macro perspective.The typical clustering algorithm K-means can quickly realize the scatter plot of turnout monitoring data,and the defects of K-means algorithm easily affect the clustering results.In this thesis,K-means algorithm is deeply studied and its shortcomings are improved.To solve the problem of how to determine the number of clusters K before clustering,an improved algorithm based on quartile and local density is proposed.To solve the problem of how to determine the initial clustering center,an improved algorithm based on spherical hash algorithm is proposed.In addition,two improved algorithms are experimentally analyzed,and UCI knowledge set is used to compare the algorithms,and the effectiveness of the improved algorithm is proved.Secondly,a feature extraction method of turnout monitoring data based on high-order cumulants is proposed,and the advantages of using high-order cumulants compared with traditional methods,FrFT methods and DWT methods are proved by experimental analysis.Thirdly,features are extracted and visualized by the method of subsection statistics,and the changing trend of features and its causes are analyzed in detail.From the macro point of view,the improved algorithm is used to cluster and visualize the monitoring data based on fourth-order cumulant,to understand the distribution of the monitoring data as a whole,and to correspond the abnormal data with the monitoring log data.From the micro point of view,the evaluation scheme of the working state of the turnout based on the adjusted cosine similarity is proposed,and statistical analysis of the state of all turnouts is carried out.Finally,the influencing factors of turnout state are summarized,and the correlation analysis of weather factors is emphasized.