Bond Graph Based Modeling And Fault Diagnosis For CRH5 EMU Traction System

Due to the convenient and environmental benefits of high-speed railway, it has become the main way of transportation in China. Also as an important role of China's foreign economic trade, high-speed railway has received lots of attention both at home and abroad. High-speed train is the closest part of relationships with passengers in the high-speed railway, and the traction drive system is one of the most important subsystem of high-speed train. So real-time fault diagnosis technology of traction drive systems is the core of normal train operation and passengers' safety. Moreover, it will be the important basis to develop the next-generation intelligent high-speed train. Based on CRH5 EMU(Electric multiple unit), this paper carries out the studies of fault modelling and real-time fault diagnosis for traction drive systems.Firstly, according to the characteristics of traction drive systems with multi-energy domains, using Bond Graph modeling technology, the physical mechanisms of inverter, traction motor and gear box are analyzed from the perspective of energy transfer and conversion. Then the respective Bond Graph models are established, and finally the system Hybrid Bond Graph model with corresponding with the actual physical structure of the system is completed. The simulation results show the reliability and accuracy of the system model, and the advantages on analyzing the system mechanism.Secondly, based on the Hybrid Bond Graph model of the system, according to the physical effects and failure characteristics of the system components, the mathematical form of the element incipient fault is established. A feasible and effective fault injection method of traction systems is proposed to accurately present fault characteristics of the actual system, which provides the basis for the study of fault propagation analysis and fault diagnosis.Finally, this paper studies the Global Analytical Redundancy Relationship based fault diagnosis of traction systems. Based on the system Hybrid Bond Graph model, the Global Analytical Redundancy Relationships and Fault Feature Matrix are derived according to the internal causality. Then detectability and isolability of elements can be analyzed. Moreover, a method of threshold selection is proposed to compare with the system residuals for fault location. The simulation results show that the proposed fault diagnosis method can rapidly and accurately diagnose a variety of abrupt fault and incipient fault.