A new prognostics approach for bearing based on entropy decrease and comparison with existing methods

Abstract

In this paper, a new method is proposed for the bearing prognosis based on the energy entropy, in which the normalized energy in the frequency spectrum is calculated over the cycles, frequency band is selected that shows greater decrease relative to the others, and entropy is computed as a trending feature. As opposed to the traditional features, which exhibit noisy fluctuation, non-monotonic change or only an abrupt increase near the end of life, the proposed energy entropy shows the smooth and constant decrease over the cycles which may represent the degree of fault progression. In order to illustrate the advantage, four traditional features - RMS, kurtosis, MAS kurtosis and envelope and the new feature - energy entropy are examined and compared using the three cases of bearing data named FEMTO, IMS and LOCAL, all from the bearing life test.