Research On The Theory And Key Technologies Of Railway Locomotive Equipment Portrait

The railway locomotive system is an important department of railway transportation system.It is mainly responsible for the operation organization,maintenance,repair and comprehensive overhaul of all types of railway locomotives.As important railway transportation and production equipment,locomotive's transportation and production efficiency,equipment quality status,maintenance and repair ability,safety management level have an important impact on the steady improvement of railway transportation production capacity and the steady development of operation management.With the wide application of various monitoring equipment and various information management systems,numerous locomotive data in various forms have been accumulated.The data increment and data quality have been greatly improved,and the data value has been increasingly reflected.Therefore,the railway industry has a very urgent requirement to improve locomotive health management.At present,there are some problems in the process of locomotive health management in the railway locomotive system,such as less analytical methods,insufficient big data mining,weak scientific management decision,,and lack of comprehensive analysis platform.Research on the theory and related key technologies of railway locomotive equipment portrait is an important means to realize the health management of locomotives.It is committed to strengthening the integrated utilization of locomotive data resources,comprehensively and accurately depicting the quality safety status of the locomotive through an objective,visual and scientific label system.Then,based on the label system,through in-depth mining of potential data value,association analysis of locomotive accident and fault,early warning and control of safety status,predictive analysis of quality safety situation,differentiated maintenance and repair,cost saving and efficiency improvement,robust and reliable management decisions and other goals can be realized.This can support the scientific,digital and intelligent development of railway transportation production and quality safety management.In this paper,the railway locomotive equipment portrait theory and some key technologies have been studied and applied,and the following innovative achievements have been achieved:(1)Railway locomotive equipment portrait theory is put forward.By sorting out the meaning and research significance of locomotive equipment portrait,the importance and positioning of the railway locomotive equipment portrait theory are clarified.Based on this,the definition and connotation of the railway locomotive equipment portrait theory are put forward and explained in detail.Then,the compositions of the railway locomotive equipment portrait theory which meets the needs of the current stage of locomotive transportation production management is sorted out,and the research methods of a series of key technologies and the logical relations between them are elaborated in detail.Furthermore,the corresponding application framework is designed,and the six main contents including core application,enabling application,technical support and overall objective are introduced in detail.This provides a new theory and method support for the systematic research on locomotive health management.(2)Railway locomotive portrait label system based on equipment portrait is constructed.By summarizing the related theoretical knowledge of label technology and integrating the multi-dimensional locomitive data,the technical framework of the three levels label system for locomotive equipment portrait is proposed.The technical framework is analyzed in data acquisition layer,label library layer and label application layer.The specific contents of each level of the label system are explained in detail.The construction method of locomotive portrait label system is formed.In the clustering acquisition mode of labels,the selection method of initial centroid of K-means clustering algorithm is improved to enhance the accuracy and stability of clustering.Finally,an objective,accurate,complete and reliable locomotive portrait label system is obtained by carrying out practical application research of locomotive equipment portrait in a railway bureau.(3)A multi-minimum support association rule mining method for railway locomotive accidents and faults based on Ms Eclat algorithm is proposed.In order to obtain association rules related to locomotive accidents and faults from railway locomotive big data,an optimized Ms Eclat algorithm is proposed.The Eclat algorithm is unable to mine association rules in the case of multiple minimum support.To solve this problem,an improved Eclat algorithm,namely Ms Eclat algorithm,is proposed.By constructing the minimum support index table,the new data set is reconstructed based on the support of each item,and then the idea of vertical mining is used to obtain the frequent item sets related to different items.Then,in order to further improve the execution efficiency of the Ms Eclat algorithm in big data analysis scenarios,the Boolean matrix and Map Reduce are applied to the calculation process of the algorithm to obtain an optimized Ms Eclat algorithm,and the frequent item set mining steps are designed and explained in detail.The comparison of different algorithms shows that the Ms Eclat algorithm and its optimization algorithm have a great advantage in computing efficiency for mining association rules with multiple minimum support.As an application practice,taking the big data of locomotive operation,maintenance and repair of a railway bureau as an example,the optimized Ms Eclat algorithm is used to mine the association rules of locomotive accidents and faults,some association rules are explained in detail.The application example verifies the effectiveness,high efficiency,accuracy and good application prospect of the algorithm.(4)A BP neural network model for locomotive quality safety situation prediction based on PSO+DE hybrid optimization method is designed.By summarizing and comparing the theorems of back propagation(BP)neural network,particle swarm optimization(PSO)algorithm and differential evolution(DE)algorithm,as well as their advantages and disadvantages,the BP neural network prediction model based on PSO+DE hybrid optimization method is designed.The PSO algorithm and the DE algorithm are combined to optimize the parameters of BP neural network in the way of time-varying probability.Based on this,the training steps of this prediction model are elaborated in detail.In addition,a mechanism to prevent the precocity of the PSO algorithm is used to make the prediction model more accurate.Then,based on the locomotive quality evaluation scheme of a railway bureau,the grey relational analysis method is used to select seven evaluation items that have a greater impact on the locomotive quality evaluation,such as the number of locomotive operation faults,the number of trivial repair problems,and the number of temporary repair problems and so on.These evaluation items are selected to predict the quality safety situation of locomotives in the next three months.Through the actual comparison of different prediction models,the new prediction model has better convergence ability,and the prediction accuracy of the locomotive quality evaluation grade and the score variation trend can reach more than 98 % and 91 % respectively.Finally,the actual prediction application and analysis are carried out,which provides a better technical method for mastering the locomotive quality safety situation scientifically.(5)The application of railway locomotive health management based on the theory of railway locomotive equipment portrait is designed.By sorting out the relationship among the railway locomotive health management application,the railway locomotive equipment portrait theory and locomotive system big data,the overall architecture of“N+1+3” and its technical architecture of the railway locomotive health management application based on the railway locomotive equipment portrait theory are designed.Then,from three aspects of equipment,personnel and comprehensive management,seven typical application scenarios are introduced,such as locomotive operation organization,locomotive maintenance and repair,aided decision-making analysis and so on.At the same time,based on the railway locomotive equipment portrait theory and related data mining technologies,the data mining analysis ideas and frameworks of these application scenarios are introduced,which lays an important foundation for the solid application of the railway locomotive equipment portrait theory.Finally,the engineering practice of the railway locomotive health management containing the relevant research results of this paper is carried out in a railway bureau.By building a man-machine friendly application system,a series of big data mining and analysis algorithm models of the railway locomotive system are encapsulated,and a number of functions are realized,such as locomotive data management,generation and analysis of locomotive portrait labels,association analysis of locomotive accidents and faults,evaluation and analysis of locomotive quality,prediction analysis of locomotive quality safety situation and so on.In the process of practical engineering application,the innovation practice of the railway locomotive equipment portrait theory and its key technologies have been realized,and good results have been achieved.There are 56 figures,21 tables and 267 references in this paper.