Identification Methods About Typical Fault Of Rotor System Based On Motor Current

Rotor system is the core component of rotating machinery,and it's the most prone to be out of order.The normal operation of the rotor system not only directly affects the operation of the whole mechanical system,but also leads to loss of the economic and property,even endangers the personal safety of workers,so many experts have devoted themselves to the theoretical and experimental research of typical faults about rotor system at home and abroad.Vibration signal analysis is one of the main fault diagnosis methods of the rotor system at present,and has been made great achievements.However,for some equipment where the sensor can't be installed,the acquisition of vibration signal is not convenient.In order to overcome this problem,the application of the motor current signal analysis method(MCSA)to the identification of typical faults of rotor system is proposed.When the rotor system is attacked by fault excitation,the load torque of rotor system will fluctuate because of fault excitation,as the same time,the electromagnetic torque of the motor will be affected by the fluctuation of load torque,which will cause the change of motor current through the stator flux.In this paper,the above transmission path was taken as the theoretical basis.According to the Lagrange equation and electromagnetic theory,the electromechanical coupling simulation models under typical fault excitation(unbalance,angular misalignment fault and parallel misalignment fault)were established based on the relationship between electromagnetic torque of motor and load torque of rotor system,and then the simulation models were simulated in MATLAB/Simulink software.In order to validate the simulation results,the test scheme was designed and the related experiments were carried out to collect the current signals exhibited by the typical faults(unbalance,angular misalignment fault and parallel misalignment fault)of the test bed.At last,the coupling relationship between the side frequency component of current signal obtained from simulation and experiment and the fault excitation was studied by the mean of Fourier transform.The result shows:the modulation frequency of|50 ± f?|(f? is the rotation frequency of the rotor system)will appear in the current spectrum,The peak value of the side frequency component changes obviously with the increase of the unbalance and parallel misalignment,but the peak value is not basically affected by the angular misalignment.In order to realize the accurate identification of fault types,the feature vectors of motor current signal exhibited by the typical faults(unbalance,angular misalignment fault and parallel misalignment fault)were extracted by using wavelet packet energy method through five layers wavelet decomposition,then the vectors of motor current signal were input to the BP neural network and support vector machine based on genetic algorithm(GA-SVM)for fault pattern recognition,and the recognition rate was 83.33%and 88.33%.The results of comparison shows:the support vector machine is more suitable than the neural network for the identification of typical faults of rotor system based on motor current.In view of the weakness because of too much feature vector dimension extracted by the wavelet packet energy method,the motor current signal under the typical fault excitation was decomposed by the empirical mode decomposition(EMD).The method about feature vectors' extraction of motor current signal was put forward based on IMF component's energy and kurtosis though EMD decomposition.Then the feature vectors were input to(GA-SVM)for fault identification,and the rate of recognition was 95.0%and the recognition effect was good.Therefore,it can be used as an effective method to identify the typical faults of the rotor system based on motor current.