Research On The Fault Prognostics And Health Management And Reliability Assessment Of Traction Converter

PHM (Prognostics and health management) has been researched and applied in the field of aviation and aerospace, but it needs much research in the field of high-speed train, urban rail train and hybrid train which need high reliability and safety. While the IGBT has the highest failure rate and is the most fragile component of traction converter which is the core power equipment in the electric locomotive. So this dissertation mainly focuses on related research and exploration in PHM technology of the traction converter and the IGBT. The PHM technology of the IGBT, the electrical motor drive system's reliability modeling and fault diagnosis algorithms are researched respectively.During the IGBT's PHM research, firstly, the major failure modes of traction converters are classified and the reasons are analyzed. Then the FMMEA is performed on IGBTs to find out different failure modes, mechanisms and corresponding precursor parameters. Secondly, during the research of health assessment and sub-safety recognition, an improved rule-based SOM Neural Network is proposed to evaluate the degradation degree and recognize the degradation state. And the validation of the algorithm is completed based on the IGBT aging test data. In the research of RUL prediction, the traditional double exponential functions are replaced with the double gaussian functions as the state equation of particle filter (PF) model. Then the Grey Verhulst(GV) model and the PF model are combined as the fusion model to research the different aging stages. The GV model is used to research the initial slow degradation stage and the PF model is used to research the fast aging stage. Finally, the prediction results show that it's much better and more accurate than traditional methods.During the research of reliability modeling and evaluation of traction converter, both the static state and dynamic state reliability modeling of traction converter are researched. In static reliability modeling, the reliability of traction converter is analyzed based on the logic diagram modeling. And the reliability of the subsystem is calculated with the failure rates of key components and then the reliability of the system is calculated based on the system logic structure. Finally, the effects of failure rates of different device on the reliability of the system are compared and analyzed. In dynamic reliability modeling, three states of the traction converter are defined and the dynamic Markov model is then acquired. The transition probabilities between different states are calculated based on devices' failure rates and the system's state transition relationships.Finally, the dynamic reliability of the system is calculated in certain confidence. Based on the research above, the key factors affecting the reliability of the system and the weak links in the system are found out. To ensure the high reliability of the system, the control measures for the reliability of each subsystem and device are proposed.In the aspect of online monitoring and diagnostic for open circuit faults of traction converter, firstly, the output fault characteristics are compared in normal and fault conditions and fault features are also extracted with the simulation of MATLAB and dSPACE semi-physical experiment platform. The faults can be diagnosed more quickly and accurately with LM-BP Neural Network instead of traditional BP Neural Network.What's more, a safety assessment scheme is proposed to assess the effect of IGBT's open circuit faults on the converter based on the SOM Neural Network. After the design of the PHM technology architecture, the on-line state monitoring and fault diagnosis system is developed, verified and applied in urban rail trains and hybrid trains.