Research On Fault Diagnosis Method Of Permanent Magnet Synchronous Motor In Ammunition Loading Syste

With the development of modern intelligent weapons,automatic loading of ammunition has become an inevitable trend.Due to the advantages of high efficiency,low noise and small volume,more and more large caliber gun automatic loading systems are driven by permanent magnet synchronous motor.The permanent magnet synchronous motor(PMSM)in the ammunition loading system of large caliber artillery will have various kinds of faults and even secondary damage if it works in the bad environment of the battlefield for a long time,it directly affects the exertion of artillery operational efficiency.Therefore,the fault detection of the motor is particularly important.In actual use,the particularity of ammunition loading system makes the number of fault samples smaller and the number of serious fault samples smaller than that of other conventional motor systems,in order to solve this problem effectively,the method of motor fault diagnosis based on deep learning is difficult to achieve high diagnosis accuracy,in this paper,a new method of deep learning motor fault diagnosis is proposed,which combines the generation of the antagonistic network and the transfer learning:1.Fault characteristic analysis of permanent magnet synchronous motor.The common faults of permanent magnet synchronous motor(PMSM)and their mechanism are analyzed,and the stator winding faults,including one-phase inter-turn short-circuit fault,three-phase inter-turn short-circuit fault and open-circuit fault,are mainly studied,the normal motor simulation model and fault motor simulation model of permanent magnet synchronous motor are built by using MATLAB/Simulink software.2.Research on interturn short-circuit fault diagnosis of PMSM based on GAN.To study the network structure and data generation mechanism of the GAN,and to study the network structure and working principle of the one-dimensional convolutional neural network(1D-CNN),in order to improve the accuracy of fault classification of PMSM networks,a set of generated countermeasure networks is used to expand the PMSM Inter turn short circuit fault data set,and a set of expanded data sets is used to train the 1D-CNN.3.Research on inter-turn short-circuit fault diagnosis of PMSM based on deep migration learning.First,a one dimensional convolutional neural network is trained on the PMSM simulation data set,then the trained network is migrated to the real motor fault data set,and the L1 regularization and Cost-sensitive loss function are used to optimize the network,in order to improve the network in the small sample noise data set on the diagnosis effect.4.Software design of motor state detection and fault on-line diagnosis.Based on the network model in this paper,the software of motor state detection and fault on-line diagnosis based on 1D-CNN is designed.The visual interface is designed by QT framework,and the on-line diagnosis and real-time state detection of permanent magnet synchronous motor are realized by using Pytorch ml library to provide on-line diagnosis deep learning model support.