Rolling bearing is a key part of turbomachinery. The performance and reliability of the bearing is vital to the safe operation of turbomachinery. Therefore, degradation feature extraction of rolling bearing is important to prevent it from failure. During rolling bearing degradation, machine vibration can increase, and this may be used to predict the degradation. The vibration signals are however complicated and nonlinear, making it difficult to extract degradation features effectively. Here, a novel degradation feature extraction method based on optimal ensemble empirical mode decomposition (EEMD) and improved composite spectrum (CS) analysis is proposed. Firstly, because only a few IMFs are expected to contain the information related to bearing fault, EEMD is utilized to pre-process the vibration signals. An optimization method is designed for adaptively determining the appropriate EEMD parameters for the signal, so that the significant feature components of the faulty bearing can be extracted from the signal and separated from background noise and other irrelevant components to bearing faults. Then, Bayesian information criterion (BIC) and correlation kurtosis (CK) are employed to select the sensitive intrinsic mode function (IMF) components and obtain fault information effectively. Finally, an improved CS analysis algorithm is used to fuse the selected sensitive IMF components, and the CS entropy (CSE) is extracted as degradation feature. Experimental data on the test bearings with single point faults separately at the inner race and rolling element were studied to demonstrate the capabilities of the proposed method. The results show that it can assess the bearing degradation status and has good sensitivity and good consistency to the process of bearing degradation.
Partial Discharge Feature Extraction Based on Ensemble Empirical Mode Decomposition and Sample Entropy
