Research On Fault Diagnosis System Of Permanent Magnet Synchronous Motor Based On Support Vector Machine Multi-classifiers

As the kernel transmission part of mechanical system,motor is widely used in industry,agriculture,daily life and other fields.However,due to overload operation,bad working environment,natural aging of coils and other reasons,the motor is prone to a myriad of faults.As an important part of the system,the failure of the motor can easily produce chain reaction,leading to the paralysis of the whole system.Therefore,it is of great significance to study the condition monitoring and fault diagnosis of motors.When the motor fails,there are a lot of abnormal signal singularities.These abnormal singularities can be used to detect and diagnose the motor fault.So,the extraction and analysis of the fault signals are very important.In fact,motor fault diagnosis is to recognize the pattern of operation state on the basis of known motor fault characteristics.This paper studies the fault diagnosis system of permanent magnet synchronous motor(PMSM)and takes the current signal and vibration signal characteristics as the research object.A fault diagnosis method is proposed which combines the improved wavelet packet transform and support vector machine(SVM)multi-classifier,and an experiment platform of fault PMSM diagnosis is built based on LabVIEW environment.The main work is as follows:Firstly,after analyzing the basic structure,fault types and mechanism of PMSM,the fault diagnosis method based on signal processing is analyzed.The fault characteristic frequencies of current signal and vibration signal of fault PMSM are concluded.Secondly,according to the analysis of fault characteristic frequencies in motor signal,a fault PMSM diagnosis experiment platform is designed.The functions of signal acquisition,storage,analysis,processing and fault diagnosis are realized,which will be used to verify the diagnosis algorithm.Thirdly,based on the improved wavelet packet transform,the current and vibration signals of the motor are processed.The improved wavelet packet transform can eliminate the frequency confusion and obtain accurate signal fault characteristics.90TDY115-2B PMSM platform is used to conduct an inter-turn short-circuit fault and the experimental results verify that the acquired signals contain corresponding fault characteristics.Finally,four common multi-classification algorithms of SVM are analyzed,which are adopted to realize multiple fault classification diagnosis experiments conducting on low speed 90TDY115-2B PMSM.The experimental results show that the "one-to-one" SVM multi-classifiers has the best comprehensive classification effect.As a classification algorithm of motor fault diagnosis system,a fault diagnosis experimental system of PMSM is designed to realize its operation status monitoring and multi-fault classification diagnosis.