Research On Fault Diagnosis Of Linear Synchronous Motor Based On Neural Network

Linear synchronous motor is one of the research hotspots in recent years.It is widely used in industrial production.If the linear synchronous motor fails,it will cause unpredictable losses,so the requirements for the reliability and safety of linear synchronous motor operation are getting higher and higher.Therefore,it has very important engineering value for the fault diagnosis of linear synchronous motor.The paper "Research on Fault Diagnosis of Linear Synchronous Motor Based on Convolutional Neural Network" is based on the National Natural Science Foundation of China: "Research on Operation Mechanism and Control Strategy of Maglev Feed Platform for Controlled Excitation Linear Synchronous Motor（Project Approval No.: 51575363）".The main contents of this study are as follows:（1）Research on the structure and operation mechanism of linear synchronous motor.Based on the strong coupling and nonlinear characteristics of the linear synchronous motor,the relationship between the winding inductance and the air-gap magnetic potential is analyzed,and the winding function of the winding that changes with the position is deduced.Combined with the voltage equation and the flux linkage equation of the armature circuit and the excitation circuit of the linear synchronous motor,the equation is established.Mathematical model to derive linear synchronous motor inductance,resistance and electromagnetic thrust.（2）Establish a simulation model of the linear synchronous motor when it is normal and when it is faulty,and build a data set of the normal and faulty conditions of the motor.On the basis of the mathematical model of the linear synchronous motor,the simulation of the normal and fault conditions of the linear synchronous motor based on the Matlab/Simulink software is established,and the data sets are constructed by comparing different time-frequency domain transformation methods.（3）Linear synchronous motor fault diagnosis based on convolutional neural network.The convolution layer is used to extract deep features,the Re Lu activation function is selected to deal with the gradient dissipation problem,the loss function is reduced by comparing different learning rates,the maximum equalization pooling layer is used to reduce the amount of computation and control over-fitting,and then the global maximum The equalization pooling layer converts the output into one-dimensional,and then inputs it to the fully connected layer,reduces the degree of overfitting through the average pooling layer,and finally inputs it to the Softmax classifier to identify the motor state,and finally realizes the convolutional neural network.GoogLeNet network is applied to the fault diagnosis of linear synchronous motor.