Fault Diagnostic Of Asynchronous Motor Based On Rough Set And Neural Network

After the reform and opening up, the economy is developing fastly, and thepenetration of electrification is increasing, motor has been applied to all walks of life. Ifthe motor breaks down on the process of application, it will bring our production anddaily life a lot of trouble, not only damage the motor itself, but also bring hidden dangerto people security, cause collective incalculable economic losses. Therefore, thediagnosis of motor faults also become a hot issue of common concern for scholars inboth domestic and overseas.In this paper, based on the study of rough set and neural network technology, tookadvantage of two algorithms to complement each others, presented a new method formotor fault diagnosis-based on rough sets and neural networks for motor fault diagnosis.Although this method has been repeatedly applied to other devices fault diagnosis, ithas been rarely applied in motor fault diagnosis. The article is mainly for research inthis area.Above all, this thesis describes the significance, current situation and developmentdirection of induction motor diagnosis; secondly, the basic principles of the inductionmotor and the common failure mechanization has been analyzed, the key point focusedon the type and diagnosis style of stator failures, rotor faults and bearing faults. Thirdly,neural networks rough set theory made a key presentation, and a detailed description ofthe thinking characteristics and algorithms of rough set and neural network ensemblemethod are given, it provides a strong theoretical support for the simulation in the finalpart. Finally, by using the collected data of motor fault, BP neural network simulationwere completed and the type of motor failure was diagnosed. Then, another motor faultdiagnosis simulation which is based on rough set and neural network approachcombining has been finished with the same data. The experimental result shows that themotor fault diagnosis method which combine rough set theory with neural network hashigh feasibility, and compared with other methods, it only need a small amount ofcomputation, have high accuracy and some other advantages.