Fault Diagnosis Of Motor Bearing Based On Optimal Generalized S-transform And IWOA-SVM

Motors are the main driving equipment in the manufacturing and play an important role in industrial production.Bearings are the core components of motors,and their health is closely related to the operational reliability of motors.In order to solve the problem that the motor bearing faults are difficult to diagnose,this paper carries out research from two aspects of feature extraction and fault identification,respectively.The optimization algorithm is adopted for adaptive selection of parameters to construct fault diagnosis scheme of the motor bearing.The main research contents are as follows:Aiming at the problem that the Whale Optimization Algorithm is easy to fall into local optimality,a dual strategy of initializing the population by elite opposition-based learning and perturbing the optimal individual by t-distribution mutation is proposed to construct the Improved Whale Optimization Algorithm(IWOA).The function optimization test shows that the optimization accuracy and stability of IWOA are better than other swarm intelligence algorithms.Aiming at the problem that the adjustment parameter selection of generalized S-transform was too dependent on experience in the past,the time-frequency aggregation degree is introduced as the objective function,and the IWOA algorithm is used for optimization to construct the optimal generalized S-transform.Simulation experiment shows that the optimal generalized S-transform has better time-frequency aggregation performance than generalized S-transform.Aiming at the problem that the previous optimization algorithms have low accuracy in determining the penalty factor and core parameter of Support Vector Machine(SVM),the IWOA-SVM classification algorithm is constructed by using IWOA to optimize the parameters.The classification experiment shows that the classification accuracy of IWOA-SVM is higher than other common SVM algorithms.Taking the general data set of motor bearing as a diagnosis example,a feature extraction method based on the optimal generalized S-transform singular entropy is proposed,which is the combination of optimal generalized S-transform,singular value decomposition and information entropy theory.The fault diagnosis scheme for the motor bearing is constructed by combing the feature extraction method with IWOA-SVM classification algorithm.In the case that the dimension of the feature sample is only three and the number of training sample is small,the average diagnosis accuracy of the proposed scheme can still be as high as 99.72%.A fault diagnosis system for motor bearings based on LabVIEW and MATLAB is established,and the on-site diagnosis is carried out through this system,verifying the validity of the proposed diagnosis scheme again.The results show that the average diagnosis accuracy of the proposed scheme is 97.58% and 88.69% when no noise is added and 5 d B Gaussian white noise is added,respectively,which means the scheme contains good anti-noise ability.Comparative experiment shows that the feature extraction method based on optimal generalized S-transform singular entropy performs better than wavelet packet energy and ensemble empirical mode decomposition,the fault recognition of IWOA-SVM can reach a higher accuracy than BP neural network.