The Research And Application On Condition Monitoring And Fault Diagnosis Technology Of EMU Active Operation And Maintenance Services

In recent years, with the large-scale of China’s high-speed railway trains putting into use, EMU monitoring data also continues to grow. How to make full use of the monitoring data to build active operational service system to the EMU equipment, and effectively support the EMU management in the aspects of operations and maintenance have become the EMU research focus in the field of management. This article focuses the research on the proactive maintenance service system status monitoring and fault diagnosis technology. Designed and implemented the EMU traction motor condition monitoring and fault diagnosis subsystem which rely on the PSOM-SVM (Parallel Self-Organizing Maps&Support Vector Machines) algorithm.However, the traditional SVM algorithm for large sample has poor training effect and slow convergence speed and other weakness. Aiming at these deficiencies, this paper combined SOM algorithm and SVM algorithm, preliminary solved the large sample and dirty data sample’s training problem. To further improve the performance of the algorithm, through the parallel can solve the problem of large sample’s long train time. In this paper, the specific work includes:(1) The paper describes the basic theory and learning method of SVM algorithm and SOM algorithm. In this paper, writer researches on some deficiencies of SVM algorithm in the system and designs SVM improved algorithms. SOM-SVM algorithm steps are designed to solve large samples of training data and dirty data problems. It also introduces parallel SOM-SVM algorithm. And from the mathematics point of view and the experimental view, this paper proves the effectiveness of the algorithm.(2)The basic characteristics of the EMU active operational service is introduced in the paper, and especially expounds on logic designing and module designing of the emu traction motor subsystem’s condition monitoring and fault diagnosis. PSOM-SVM algorithm is applied to the system. SVM algorithm kernel function and related parameters are selected by designing experiments. Writer researches also state the data preprocessing work in the process of system implementation and the training process of PSOM-SVM algorithm. Finally, taking traction motor of CRH2as example, this paper shows the condition monitoring and fault diagnosis of PSOM-SVM algorithm. Experiments show that the algorithm is accurate and the fault diagnosis result is good.