A comparative study for the selection of an entropy technique to classify bearing faults

Abstract

The diagnosis of faults in ball bearings has been a major challenge especially in the non-stationary environment. This paper presents a comparative study to select a method among the approximate entropy, Sample entropy and Multi-Scale entropy that can successfully be employed as a feature vector for fault classification of ball bearings. The acquired vibration signals from the normal and faulty condition of ball bearing are firstly de-noised and then disintegrated by Ensemble empirical mode decomposition method to various intrinsic mode decompositions. From selected intrinsic mode functions the signals are reconstructed and entropies are computed by the proposed methods. The results demonstrate the effectiveness of the Multi-scale entropy technique been able to effectively differentiate the condition of bearing and can be considered to act as an important fault feature vector. © 2019 IEEE.