Multi-Class Imbalanced Fault Diagnosis Method Based on IMWMOTE and MFO-Bayes-LS-SVM

Abstract

In actual industrial production, the historical data of fault diagnosis are often imbalanced. Therefore, this paper uses a fault diagnosis method based on the improved MWMOTE(Majority Weighted Minority Oversampling Technique) algorithm and LS-SVM (Least Squares Support Vector Machines) under the Moth Flame Optimization (MFO) -Bayesian framework. The IMWMOTE algorithm was used to over-sample to obtain the balanced data set. To verify the effectiveness of IMWMOTE algorithm and optimize the parameters of LS-SVM classifier, we used MFO-LS-SVM method to diagnose whether the fault occurred. Then, the Bayesian- LS-SVM method is used to diagnose the fault types. An example of bearing fault diagnosis shows that the proposed method has higher fault diagnosis recognition rate and algorithm robustness than the existing algorithms.