Feature Extraction of Rolling Bearing Fault Based on Ensemble Empirical Mode Decomposition and Correlation Dimension

Rolling bearing is the core element of a machine, especially used in rotary machine. Its working status and healthy condition directly affect the efficiency and life cycle of a machine. So it is very important to monitor and diagnose the faults of rolling bearings. In this paper, a novel method based on ensemble empirical mode decomposition (EEMD) and improved correlation dimension (CD) is presented to extract fault feature of rolling bearing fault. The conventional CD has two defects, one is sensitive to the noise, and another is difficult to calculate the slope over the linear region (scaling region). In order to reduce the effects of noise, EEMD is used to decompose the components with truly physical meaning from signals. And in order to identify the scaling region and calculate the slope, an improved CD algorithm is proposed to acquire the scaling area automatically and verified by the well-known analytic models such as Lorenz attractor. Finally, the method is applied to detect the fault features of rolling bearings based on vibration signals and the experimental results indicate its applicability and effectiveness in fault diagnosis of the rolling bearings.