Research On Electric Locomotive Fault Analysis Techniques

Railways play a very important role in the national economy.The locomotive is the traction power equipment of the railway.The use of electric locomotives in China's railway transport is increasingly widespread,HXD2 B electric locomotive is a heavy cargo locomotive.The structure of the locomotive is very complex,and the types of faults are varied.Although a number of intelligent failure analysis means have been applied in the use of electric locomotives But when the actual problem occurs,the fault analysis and diagnosis work,maintenance work depend on the experience of the maintenance personnel of the locomotive,far from meeting the requirements of railway modernization.This paper aims to study the intelligent diagnosis technology of electric locomotive through a variety of methods,so that maintenance personnel can quickly and accurately find the locomotive fault and carry out maintenance.In this paper,two kinds of fault diagnosis methods are studied: case-based reasoning(CBR)method based on rough set and BP neural network algorithm combined with particle swarm optimization algorithm(PSO-BP).This paper first reviewed the relevant information of the electric locomotive and used the driver display unit failure as an example,collected typical fault cases,fault characteristics.The expert knowledge is recorded and sorted out,and then through different methods to achieve the electric locomotive fault diagnosis technology research.CBR is an intelligent diagnostic method,which is a knowledge-based learning and problem solving method.It provides decision-making reference for new problems by reusing or modifying historical knowledge and examples,and can continuously learn and update knowledge library.The process of case-based reasoning involves four procedures: case retrieval,case reuse,case revise and case retain.In this paper,K-NN algorithm is used in case retrieval procedure.Rough set theory is used to calculate weights of the attributes and reduce the data for effective case retain.After the completion of each diagnosis,users interact with the software for case reuse and case revise by recording new case in the case base.In the algorithm of PSO-BP,the BP neural network algorithm is simple and widely used in the Research on different areas,but its convergence speed is slow and easy to fall into the local optimal value.However,the PSO algorithm is fast and has a global search function,and the combination of the two can quickly find the optimal value in the global scope.In this algorithm,the weights and thresholds of the BP neural network are replaced by the displacement of the particles in the PSO.By setting the stop criterion: fitness and iteration times,the optimal weights and thresholds are obtained in the global range search and assigned to BP neural network training,greatly accelerating the BP neural network training speed,increased accuracy.In this paper,Matlab is used to research and simulate the fault diagnosis technology,by using electric locomotive actual data,the 2 methods are completed and the results are given by the end of the paper separately.