Research On Fault Diagnosis Method Of Asynchronous Motor Driven By Frequency Conversion

As a kind of mechanical equipment to realize energy conversion,asynchronous motor is characterized by simple structure,high operation efficiency and wide application.In order to effectively reduce power consumption,prolong service life and realize variable frequency speed regulation of asynchronous motor,frequency converter asynchronous motor drive system is widely used in industry.Under the harsh conditions of high temperature,humidity and vibration,the rotor of asynchronous motor is prone to failure,mainly including broken bar and eccentric fault.In order to reduce the loss caused by motor fault in the production process,this paper studies the diagnosis methods of two types of rotor fault of asynchronous motor under variable frequency drive.Firstly,this paper analyzes the fault mechanism of broken bar and eccentric fault of motor from the perspective of electromagnetic field,and obtains the calculation formula of characteristic frequency induced by various fault stator current signals.Build a three-phase asynchronous motor model in Maxwell in the simulation software ANSYS,change the conductivity of the guide bar and move the rotating shaft of the guide bar to turn the normal motor into a fault motor,and then conduct joint simulation with the SPWM control circuit in Simplorer to obtain the stator current signal,conduct FFT analysis,obtain the fault characteristic frequency under different frequencies,and verify the accuracy of the simulation model.Secondly,this paper introduces three decomposition methods of EMD,EEMD and ceemd.The analysis shows that ceemd has better decomposition effect and can effectively suppress modal confusion.The fault feature is extracted by combining sample entropy and arrangement entropy.The obtained fault feature data set is identified by convolution neural network(CNN).The simulation results show that the fundamental frequency in the original signal is filtered first,and then ceemd-cnn method is used for fault diagnosis,The correct rate of diagnosis is 97.56%,which is 8.07%higher than that of fault diagnosis directly using the original signal.Finally,the experimental platform is built,the data obtained from the experiment are FFT,and the authenticity of the data measured by the experimental platform is verified.Then the eemd-cnn and ceemd-cnn methods are used for fault diagnosis.It is concluded that the ceemdcnn method has better fault identification effect,and the accuracy of the measured data reaches 95.94%.