Research And Application Of Inter-turn Short Circuit Fault Of Permanent Magnet Synchronous Motor Based On Signal Analysis And Machine Learnin

Permanent magnet synchronous motor(PMSM)is an indispensable electromechanical component in contemporary society,and its good characteristics can also promote the development of new energy applications.As a common fault in synchronous motor,although its own impact is not significant,if the fault occurs,the winding coil heats up which causes other faults such as vibration of other motors,which seriously affects the working condition of synchronous motor and the safety of related personnel.When the fault occurs,the motor parameters change in two cases with the offset of time,stable change and sudden change.Using this feature,the signals of different fault states of PMSM are divided into smooth and non-smooth waveform signals.Different signal analysis methods and classification methods can accurately determine whether the motor fault occurs and the severity of the fault,and provide fault information for maintenance personnel as early as possible.For the different characteristics of the two signals,the following two methods are studied:(1)Wavelet Packet Decomposition(WPD)and K-Nearest Neighbor(KNN)algorithm are combined to diagnose the short circuit between turns of PMSM.The model addresses the characteristics that the signal remains smooth during normal motor operation,decomposes the signal using the discrete WPD algorithm,determines the optimal decomposition method based on mean square error,and calculates the energy entropy as a feature value.The I＿w algorithm is written for optimizing the feature values and assigning appropriate weights,and the KNN classification algorithm is combined to realize the diagnosis of faults under smooth signals.(2)A model of short-circuit fault between turns of a PMSM is constructed by combining Variational Mode Decomposition(VMD)and Convolutional Neural Networks(CNN).The model decomposes the signal in a non-stationary state when a fault occurs in the motor.Conbining the model,I use VMD to decompose the signal,use the mean square error,Pearson correlation coefficient and running time to determining the optimal decomposition method,use fifteen different time-domain feature values to extract the primary features,and use CNN to extract the secondary features.The softmax classifier in the CNN is used to classify the faults and achieve the fault diagnosis of synchronous motors.A fault diagnosis system for inter-turn short circuit of PMSM is built,which contains two parts: signal analysis and signal diagnosis.In the module signal analysis,the input signal is visualized and the corresponding partial feature values are calculated.In the module signal diagnosis,two models are input,and the signals to be diagnosed are input into the model for decomposition and diagnosis to achieve the diagnosis and classification of motor faults,the corresponding results are saved.The Thesis establishs the simulation model of the turn-to-turn short circuit of the PMSM,and simulate the actual motor state and signal waveform to analyze and diagnose the fault for different signal characteristics respectively.The proof-of-concept experiments and comparative experiments show that both models demonstrate high fault diagnosis rate and good guidance value in mechanical diagnosis application.