Rolling Bearing Fault Diagnosis Based on CEEMDAN and Refined Composite Multiscale Fuzzy Entropy

Considering the nonlinear and nonstationary characteristics of rolling bearing vibration signals, we propose a rolling bearing fault diagnosis method based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), refined composite multiscale fuzzy entropy (RCMFE), Laplace score (LS), and the particle swarm optimization-probabilistic neural network (PSO-PNN). First, the method employs CEEMDAN to decompose the vibration signal and select the intrinsic mode functions (IMFs) containing the primary fault information via the frequency-domain correlation coefficient method. Then, it uses RCMFE to extract the characteristic information from the selected IMF. In addition, it uses LS to select and construct low-dimensional sensitive feature vectors, which are incorporated into the PSO-PNN model for diagnostic analysis to realize the state recognition of rolling bearing. Finally, the effectiveness of the method is verified by the analysis of the experimental data.