IGWO algorithm based WSVM classification of different faults in roller bearing using vibration signals

Abstract

The changes in electrical and mechanical components creates many problems and complications in roller bearing element which results in production of vibration signals and cause the whole system damaged. In this paper, the vibration signals are obtained from the roller bearings in four different conditions namely normal,inner race fault,outer race fault and roller rub fault.Then the signals are decomposed using improved Empirical mode decomposition (EMD)denoising technique.By applying shifting, the Intrinsic Mode Functions(IMF) components are extracted from the measured signalafter performing three iterations. To classify different faults, the statistical features likemean, standard deviation, kurtosis, skewness, energy and entropyare extracted. The improved grey wolf optimization (IGWO) is used to improve the performance of WSVM. The results reveal that, IGWO implemented WSVM classifier shows better performance in terms of fitness strength and confusion matrix due to flexibility in fitness strength and good out-of-sample generalization.