Research On Fault Diagnosis Method Of Stator Winding Interturn Short Circuit Based On Data Drive

Induction motor,as the most important driving equipment in modern industrial society,plays an important role in production and life.Induction motor running state health or not will have a direct impact on equipment performance,product quality and personal safety.The short-circuit fault between stator turns is an important fault type of induction motor.The fault is manifested as the gradual change of winding insulation ability.Therefore,the diagnosis of the interturn short-circuit fault of induction motor can help to find the fault at the early stage of short-circuit and take corresponding measures in time,so as to ensure the safe operation of the motor.With the development of fault diagnosis technology,the diagnosis method of inter-turn short circuit fault has been developed from the traditional expert-based method to the data-driven one.Through feature extraction of fault current signal,artificial intelligence network model is trained to realize fault diagnosis and classification.In this thesis,according to the different degrees of induction motor stator turns short circuit fault,first build the fault simulation model,output the simulation fault current data;Then,the traditional machine learning method is used to classify and diagnose the real fault data.In order to further improve the classification accuracy,the deep learning method is used for classification;Finally,the simulation fault current data is used for transfer learning to solve the problem of insufficient real fault samples in the process of actual motor fault diagnosis.The research content of this thesis mainly includes the following four points:(1)Build the simulation model of interturn short circuit fault.Short-circuit current with different fault degree can be generated by setting short-circuit resistance and shortcircuit turns ratio.In view of the gap between the model output current and the real fault current,a steady-state fault model was built,and the least square method was used to identify the model parameters and narrow the gap between the model parameters and the real motor parameters.By building a fault model,simulation fault samples were generated to provide data support for subsequent migration learning.(2)The fault diagnosis method based on machine learning is studied.By manually extracting the time-frequency domain features of real fault data,random forest and XGBoost were used respectively to achieve different degrees of inter-turn short-circuit fault diagnosis.Experimental results show that the fault stator current data can be used as signal source for feature extraction.(3)A fault diagnosis method based on deep learning is studied.Aiming at the problem that traditional machine learning requires manual feature extraction and the accuracy of fault classification is insufficient,a Res Former model fused with Res Net neural networks and Transformer is built,and statistical features and depth features are extracted from fault signals.Experiments have confirmed that the Res Former model using depth fusion can effectively realize the diagnosis of stator winding inter-turn short-circuit faults.(4)Aiming at the problem of insufficient real fault samples in actual motor fault diagnosis,the application of transfer learning in fault diagnosis is studied.A network model based on CORAL feature migration,a network model based on DANN adverssion migration and a network model based on CORAL-DANN adverssion migration were built respectively.The network model was trained to learn fault knowledge from the simulation fault sample data set and applied to the real fault diagnosis field,and the fault diagnosis between turns short circuit was realized under the condition of small samples.