Research On Fault Diagnosis Method Of Motor Bearing Based On Optimized SVM

The motor is a kind of mechanical equipment widely used in modern industrial production,and its running state is related to the safety and stability of the entire system.Since the bearing is the core component of the motor,bearing fault not only affects production efficiency,but may also lead to safety accidents.It is of great significance to study the fault diagnosis method of the motor bearing and make a timely and accurate diagnosis in the early stage of the bearing fault,which can ensure the stable operation of the motor and the safety of industrial production.Due to the complexity of the vibration signal increases when the motor bearing fails,this paper proposes a fault detection method for motor bearings based on permutation entropy.The reasonable fault threshold is set according to Chebyshev’s theorem,which can separate the fault signals and normal signals accurately,so as to detect whether there is a fault in the bearing in time.Considering the non-stationary and nonlinear characteristics of motor bearing vibration signal,a fault feature extraction method based on ensemble empirical mode decomposition(EEMD)and permutation entropy is proposed to adaptively analyze the time-frequency domain features of fault signal.Since the intrinsic mode function(IMF)components decomposed by EEMD have false components that cannot describe the feature information,the IMF components with high correlation are screened out by calculating the correlation coefficient between each IMF component and the original signal.Combined with the ability of permutation entropy algorithm to measure signal complexity,the permutation entropy values of highly correlated IMF components are extracted as the fault feature vector of motor bearing.Since it is difficult to obtain a large amount of typical fault data of motor bearings,a support vector machine(SVM)suitable for small sample problems is selected to realize fault identification of motor bearings.In order to ensure that SVM has the best classification performance,a hybrid gray wolf optimization(HGWO)algorithm based on three improved strategies is proposed to optimize the parameters of the SVM model,including the population initialization strategy based on the good point set theory,the nonlinear control parameter strategy based on Sigmoid function,and the position update strategy based on the teachinglearning-based optimization algorithm.The HGWO algorithm solves the problems of premature convergence and low convergence accuracy in the gray wolf optimization algorithm.Considering that the binary classifier SVM cannot identify various faults of motor bearings,a multi-class HGWO-SVM model based on the directed acyclic graph for fault diagnosis of motor bearings is proposed.The bearing data samples from Case Western Reserve University were selected for simulation experiments,and the permutation entropy values of the highly correlated IMF components acquired by the EEMD decomposed vibration signal were extracted as the fault feature vectors,which was input into HGWO-SVM for training and learning.Then a multi-classification HGWO-SVM model based on directed acyclic graph was constructed.The experimental results of diagnostic tests show that the proposed fault diagnosis method for motor bearings based on optimized SVM can effectively diagnose the fault type and damage degree of motor bearings with high accuracy.