Intelligent bearing fault signature extraction via iterative oscillatory behavior based signal decomposition (IOBSD)

A typical vibration signal from a defective bearing is composed of fault-induced transients (i.e., fault signature), resonance, multiple vibration interferences and background noise. Extracting the fault-induced transients from such signals is the primary goal for bearing fault diagnosis. This paper proposes an intelligent oscillatory behavior based signal decomposition (OBSD) method for this purpose. The OBSD technique exploits both the signal separation capability of morphological component analysis (MCA) and the basis creation potential of tunable Q-factor wavelet transform (TQWT). To generate proper bases for each individual signal component, the Q-factor determination and tuning strategies for the TQWT are developed in this study. In the presence of multiple interferences, one-time OBSD may not be sufficient to extract fault feature successfully. As such, an iterative OBSD (IOBSD) procedure is proposed. With the proposed method, a vibration signal can be decomposed into three signal components, i.e., low-oscillation component (fault signature presenting non-oscillation), high-oscillation component (interferences manifesting sustained oscillation) and residual (noise), according to their oscillatory behaviors. Bearing faults can then be reliably detected. The main features of the proposed method include: (1) it can remove in-band interference and noise that cannot be removed by the widely used frequency-based approaches, as it is frequency-independent in nature, (2) it can extract fault signatures from heavily interference-obscured signals without filtering the signals, and (3) it does not require the prior information of the signals. With these features, the IOBSD method can substantially reduce human involvement and facilitate its implementation in a fault detection expert system, particularly in an environment with various interferences. The IOBSD method has been favorably compared with one of the popular filtering techniques, i.e., SK method. The effectiveness of the proposed method has also been tested by simulation and experiments. (C) 2015 Elsevier Ltd. All rights reserved.