Research On Rough Set Of Traction Converter Fault Diagnosis

Converter is a very complex device, including rectifier, middle circuit and inverter. It can turn the single phase AC the overhead lines into three-phase ones whose frequency and amplitude can be adjusted to feed motor. It is a key component of electric locomotive during energy conversion. However, due to continuous and long working state of the traction converter, it is easy to appear fault. Once a fault is occurred, the safety of people’s life and property will be threatened. Therefore, more and more attention from all walks of life focus on the question of rapid diagnosis and repair about traction converter.Firstly, the concept of fault diagnosis is introduced in the paper, secondly, the development trend of locomotive fault diagnosis and methods of the traction converter fault diagnosis are described both at home and abroad, then, the present situation of research on rough set method is analyzed, a kind of traction converter fault diagnosis method based on Rough Set is put forward. The rough set theory is applied to fault diagnosis, no need of any prior knowledge of fault data and knowledge, a variety of data which is incomplete and uncertain can be processed, and the implicit knowledge can be discovered, the potential rules can be revealed.The internal structure and working principle of the traction converter on CRH5 are also introduced in the paper, and by analyzing the output current waveform of the traction inverter we found current have serious distortion under the fault condition It can be served as fault feature parameters. MATLAB/simulink is used to build the inverter model of traction converter, the output current waveform are produced under various fault conditions and the wavelet transform is applied to extract the energy value of the fault waveform,getting the fault data, building a decision table through data discretization and attribute reduction, the fault diagnosis of the traction converter is carried out correctly.The research result indicates that this method has good real-time performance and high accuracy. the dimension of fault data and the workload of fault diagnosis are reduced after attributing reduction, the efficiency of diagnosis is improved with good prospects of development.