Analysis On Demagnetization Fault Of Permanent Magnet Synchronous Motor Based On Image Of Current Signal

Permanent magnet synchronous motors are extremely widely used due to their small volume,reliable operation,high working efficiency.Permanent magnets are the most expensive among the part of PMSM,but they are prone to demagnetization failures due to factors such as harsh environments and complex working conditions,which will reduce the performance of the motor.Therefore,in the early stage of the occurrence of demagnetization faults,demagnetization failures detection is of vital significance to the safe operation of permanent magnet synchronous motors.However,in the past,when performing demagnetization fault diagnosis,most of the characteristic signals were analyzed,which was easily affected by factors such as electromagnetic environment and noise.And subject to the constraints of signal decomposition,the extraction of signal features and the identification of fault types were very difficult,and methods that rely on manual feature extraction are no longer applicable.Therefore,a method of permanent magnet synchronous motor demagnetization fault diagnosis based on current signal imaging is proposed in this thesis,which converts the original current time domain signal into image samples through image processing,introduces image recognition and artificial intelligence technology.It employs deep convolutional neural network extracts the converted image features,then diagnoses the fault state under different types of demagnetization and different degrees of demagnetization.The main research contents of this thesis are as follows:(1)The formation mechanism of the demagnetization failure of the permanent magnet material of the permanent magnet synchronous motor is analyzed.By studying the magnetization characteristics of permanent magnet materials,the causes of different types of demagnetization of permanent magnets in the motor are analyzed.For the permanent magnet synchronous motor studied,a mathematical model and a finite element simulation model under the permanent magnet demagnetization fault are established,and the magnetic flux density change of the motor and the change law of the current under different types and degrees of demagnetization are analyzed.(2)The current signal image processing is implemented,and a conversion for the current signal imageization is proposed,and the current signal data is converted into an image sample.The convolutional neural network model was studied,and the causes and solutions of the over-fitting of the convolutional neural network ware analyzed.The convolutional neural network can directly input two-dimensional image samples,and use the image samples converted from the current signal of the permanent magnet synchronous motor as the input of the network.(3)A permanent magnet synchronous motor demagnetization failure experiment platform is built,and a diagnostic model based on image samples and convolutional neural networks for verification established.A signal data set based on the signal analysis method is set up,and three network models of probabilistic neural network(PNN),general regression neural network(GRNN)and least square support vector machine(LS-SVM)are established and compared.The experimental results shows that the method proposed in this thesis achieves fast feature extraction and can improve the recognition rate.This method is suitable for both types of uniform demagnetization and local demagnetization,and can realize the identification of different demagnetization degrees,and realizes the rapid and effective diagnosis of permanent magnet synchronous motor demagnetization fault.