Research On The Diagnosis Method Of Cage Asynchronous Motor Fault Based On Machine Learning

As the most common driving device in production and life,asynchronous motor plays an important role because of its small size,low cost and convenient speed regulation.If the asynchronous motor fails,it will affect the overall production efficiency,reduce the production capacity,cause economic losses,and even cause more serious accidents or casualties.In order to ensure the safety and efficiency of agricultural industrial production process,it is required to find the initial problems in the operation of asynchronous motor as soon as possible without affecting the normal production,so as to reduce the impact of motor fault as far as possible.It is of great practical significance to carry out on-line fault diagnosis of asynchronous motor.There are many kinds of asynchronous motor faults,and the fault phenomenon is complex,the initial fault is relatively slight,which is difficult to distinguish.Therefore,this paper focuses on studying a variety of machine learning methods including neural network and its application in the field of cage motor online fault monitoring:1)The paper briefly analyzes the mechanism of broken rotor bar fault and stator inter-turn short circuit fault of squirrel cage induction motor,and analyzes the fault characteristic quantity commonly used in traditional motor fault diagnosis methods.2)This paper systematically studies the principle and implementation of machine learning method,and uses random forest,support vector machine and optimized lightgbm method to train the classifier for the experimental data to evaluate the results.In addition,the weights of input features are sorted and compared to verify the feasibility of traditional electrical diagnosis methods from the data point of view.3)The feasibility of deep learning in fault diagnosis of induction motor is studied,and the performance of deep neural network is trained to compare with the above methods;Autoencoder and its derivative type: trestle self encoder and its sparse noise reduction characteristics are emphatically analyzed.4)For the first time,a fault diagnosis method of squirrel cage induction motor based on the combination of trestle autoencoder and improved lightgbm algorithm is proposed,which can simultaneously and jointly diagnose the stator inter-turn short circuit fault and broken-rotor-bar fault.The results show that the method is effective.