Rolling bearing fault diagnosis based on LCD–TEO and multifractal detrended fluctuation analysis

Abstract

A rolling bearing vibration signal is nonlinear and non-stationary and has multiple components and multifractal properties. A rolling-bearing fault-diagnosis method based on Local Characteristic-scale Decomposition–Teager Energy Operator (LCD–TEO) and multifractal detrended fluctuation analysis (MF-DFA) is first proposed in this paper. First, the vibration signal was decomposed into several intrinsic scale components (ISCs) by using LCD, which is a newly developed signal decomposition method. Second, the instantaneous amplitude was obtained by applying the TEO to each major ISC for demodulation. Third, the intrinsic multifractality features hidden in each major ISC were extracted by using MF-DFA, among which the generalized Hurst exponents are selected as the multifractal feature in this paper. Finally, the feature vectors were obtained by applying principal components analysis (PCA) to the extracted multifractality features, thus reducing the dimension of the multifractal features and obtaining the fault feature insensitive to variation in working conditions, further enhancing the accuracy of diagnosis. According to the extracted feature vector, rolling bearing faults can be diagnosed under variable working conditions. The experimental results demonstrate its desirable diagnostic performance under both different working conditions and different fault severities. Simultaneously, the results of comparison show that the performance of the proposed diagnostic method outperforms that of Hilbert–Huang transform (HHT) combined with MF-DFA or LCD–TEO combined with mono-fractal analysis.