Research On The Fault Diagnosis System For The Vehicle Equipment Based On Neural Network And Case-based Reasoning

With the rapid development of high-speed railway transportation, the innovation of China’s high-speed railway technology, the increasing running speed, the vehicle equipment is becoming increasingly integrated. As a CTCS-3 level train control system, the reliability and safety of high-speed railways is very significant. As the core part of train control system which responsible for the running speed, real-time monitoring and speed protection. Therefore, the fast and accurate fault diagnosis of vehicle equipment has great significant to guarantee the safety of the train.With the development of vehicle equipment, the equipment’s structure becomes more and more complicated, and its integration and density is also increasing continuously. The characteristics of faults are usually stochastic, unpredicted and overlap, which increase the difficulty of diagnose and maintenance. Only rely on the knowledge and experience of maintenance personnel to diagnose, the efficiency is relatively low, the rate of misdiagnosis is high. On the basis of the traditional fault diagnosis of the vehicle equipment, neural network, rough set and case-based reasoning are introduced into fault diagnosis, which can improve the efficiency and accuracy of diagnosisFirstly this paper introduces the structure of the vehicle equipment system, the characteristics of fault, based on the type of faults. Due to the diversity of the fault types of the vehicle equipment, as well as the data collected in the operation process is not complete and random, the minimum attribute reduction algorithm combined with the neural network was applied to the train fault diagnosis.This paper introduced the basic structure of BP neural network, and constructed a three layer BP neural network, according to the common fault phenomenon and the common fault feature of vehicle equipment to build neural network data training samples. After that, the fault phenomenon membership parameters could be expressed as the fuzzy representation, improved BP algorithm has better diagnosis accuracy and error convergence speed through the Matlab simulation experiment.This paper studied the technology of case-based reasoning, including case representation, case library, case retrieval and case study. In the case sets, the data is processed by the rough set theory, and the attribute value is reduced by the algorithms based on simplified discernibility matrix, which can reduce the complexity of the data.The study of intelligent diagnose and retrieval technology. The standard BP neural network algorithm is improved by adding the additional momentum method, and the accuracy of case-based reasoning diagnosis is not high, simultaneously the reasoning process is slow. This paper set up a design combined with the two methods to meet the system requirements, neural network as a classifier and case retrieval algorithm to improve the learning speedDemonstrated the feasibility of the combination of the artificial neural network and case-based reasoning method, According to the requirements of the system, using the Visual Studio 2010 and SQL2008 database technology to develop a fault diagnosis system for vehicle equipment under the Windows operating system, the system’s diagnostic process, diagnosis module and database design, to achieve a fault diagnosis system combined neural network and case-based reasoning.At last, the advantages and disadvantages of the system are analyzed, and the limitations of the system are pointed out, and the further research is prospected.