Detrended fluctuation analysis of vibration signals for bearing fault detection

Abstract

Rolling element bearings are widely used in various rotary machines. Accordingly, a reliable bearing fault detection technique is critically needed in industries to prevent these machines' performance degradation, malfunction, or even catastrophic failures. Although a number of approaches have been reported in the literature, bearing fault detection, however, still remains a very challenging task because most of the bearing fault related signatures are nonstationary. This paper presents the detrended fluctuation analysis (DFA) of vibration signals to tackle this technical challenge in bearing fault detection. The DFA offers the advantage over the traditional spectral analysis methods in that it can deal with nonstationary signals, and also its application does not rely on the selection of mother functions as the wavelet transform does. The effectiveness of the proposed technique is examined through a series of experimental tests, and the investigation results demonstrate that the faulty bearing conditions can be well detected by analyzing the power-law characteristics of the DFA.