Research On PMSM Demagnetization Fault Diagnosis Method Based On Current Signal Machine Learning

With the development of power electronics technology,sensing technology,rare-earth permanent magnet materials and motor control theory,permanent magnet synchronous motor(PMSM)has been applied more and more widely in electric vehicles,airplanes,high-speed rail,etc due to its advantages of simple structure,small size,high efficiency,high torque quality ratio,better control performance and easy to heat dissipation and maintenance.However,permanent magnets as the core component of PMSM,the demagnetization fault of permanent magnet is a key problem in the popularization and application of PMSM.It is directly related to whether the permanent magnet synchronous motor can operate normally,It also has a certain impact on the performance of the entire control system,which may cause huge losses after the demagnetization failure.Therefore,it is extremely important to detect and diagnose the demagnetization fault of PMSM.The core of the PMSM demagnetization fault diagnosis is to find simple and reliable fault characteristics,and can realize fault detection in complex non-stationary state,fault isolation and determine the fault degree.Aiming at the demagnetization of PMSM,this paper studies a non-intrusive intelligent fault diagnosis method: Fault diagnosis method of permanent magnet synchronous motor based on current signal machine learning,and designs an automatic fault diagnosis system software of PMSM.Its content is mainly divided into the following three parts for in-depth study.(1)First of all,through the analysis of permanent magnet synchronous permanent magnet motor,analyzes the mechanism of demagnetization,adopt the method of finite element analyzing demagnetization of permanent magnet synchronous motor.The finite element simulation software is used to build the fault model of local demagnetization of permanent magnet synchronous motor.The key information such as current and voltage flux linkage of the fault model is obtained through simulation,and the current information is finally selected as the research object.(2)Secondly,a demagnetization fault diagnosis method based on cost-sensitive support vector machine(CS-SVM)is proposed for the single permanent magnet demagnetization fault of permanent magnet synchronous motor.The fault current signal was processed by empirical mode decomposition(EMD)method,and the intrinsic mode function(IMF)energy moment was constructed by integrating the IMF component with respect to the time axis,which was taken as the fault feature and extracted.The fault diagnosis data is classified by CS-SVMalgorithm and compared with several existing fault diagnosis methods.The results show that the proposed method is feasible.(3)Finally,for the local demagnetization fault of permanent magnet synchronous motor,a more advanced cost-sensitive large margin distributor(CS-LDM)learning method is proposed to effectively analyze and identify the fault current characteristic signal of local demagnetization fault.The experimental results show that the proposed method has high diagnostic accuracy.And based on the CS-LDM algorithm,a software based on CS-LDM for the permanent magnet synchronous motor demagnetization fault intelligent diagnosis system is developed,and the software is used to realize the online diagnosis of the permanent magnet synchronous motor demagnetization fault.