Font Size: a A A

Research On Comprehensive Evaluation Of Rail Condition Based On Multi-source Data Fusion

Posted on:2022-11-13Degree:MasterType:Thesis
Country:ChinaCandidate:G JiangFull Text:PDF
GTID:2492306617496424Subject:Railway Transportation
Abstract/Summary:
The rail is an important part of the railway track.It is subjected to dynamic reciprocating loads such as train load and wheel-rail force,as well as the combined action of severe cold,sand,rain and snow,freeze-thaw,corrosion and other harsh natural environments and complex geological conditions.The operating state is constantly changing during its service period.The defects of the rail in the long-term wheel-rail action process and the degradation of the material performance of the rail directly affect its service performance and seriously threaten the driving safety.With the increasing development of my country’s railways in the direction of high speed,high density and heavy load,higher requirements are put forward for the safety,stability and reliability of rails.With the increasing number of rail detection and monitoring methods,various types of detection data provide a scientific basis for the diagnosis of rail defects,and also bring challenges to the fusion analysis,trend analysis,correlation analysis and state evaluation of rail defects.How to be more comprehensive,How to more scientifically evaluate the operating status of the rails,how to more accurately grasp the actual status of the rails at the railway site,and how to guide the site to carry out targeted maintenance and repairs based on the actual status of the rails are practical problems that need to be solved urgently on the railway site.This paper first discusses the complexity of the rail state,analyzes the various factors that affect the rail state,and introduces various methods of rail detection and the detection advantages of the rail flaw detection vehicle.Secondly,the influence of various types of data such as historical damage data,environmental factors,flaw detection vehicle detection data,and total weight on the rail condition is comprehensively considered.Finally,8 characteristic parameters are selected to evaluate the rail condition.As a result of the periodic inspection,59 rail sample units of the Jingbao Line were selected with relatively concentrated rail wear,surface defects,and minor injuries,and the actual state of the site was poor.The expert survey method was used to quantitatively score the rail status of the sample units.The pair comparison matrix is constructed,the evaluation weight of experts is determined,and the subjectivity of the quantitative evaluation of the state of the sample unit is reduced.Finally,based on the BP neural network,a comprehensive evaluation model of the rail condition is constructed.The influence factors of the sample rail units and the scores of experts are used as the input and output layers of the model to train,verify and test the model,and the test accuracy is the lowest.It is 96.39% and the highest is 99.86%,which finally realizes the comprehensive evaluation of the state of any rail unit.The results show that the evaluated rail state is very close to the real state and expert experience evaluation,which can meet the rail management requirements of railway site.The main research work of this paper is as follows:From the perspective of on-site use status of rails and the complexity of rail conditions,this paper proposes a method to analyze the historical data of rails,focusing on the concentrated areas of various minor injuries and the rail conditions under the joint influence of various defects that have been neglected in previous research and rail management work.,detection monitoring result data,environmental observation data and other multi-source heterogeneous data fusion evaluation method,established a comprehensive evaluation model of rail condition.Using the characteristics of input normalization of BP neural network,data of different types and structures are fused,and the evaluation of sample rail units by experts in public works department is accurately simulated through machine learning algorithm,and the state evaluation of any rail unit is carried out with examples.,realizes the quantification of the rail status,and provides the basis for the precise management of the rail life and effective guidance for the on-site maintenance and repair of the rail.
Keywords/Search Tags:Rail, Multi-source data fusion, Rail flaw detection vehicle, Comprehensive evaluation, Hierarchical analysis method, Neural network
Related items