A rolling bearing fault diagnosis method based on multi-scale fuzzy entropy and variable predictive model-based class discrimination

• Multi-scale fuzzy entropy (MFE) is developed to measure the complexity of time series.
• MFE is contrasted with MSE and is utilized to analyze rolling bearing vibration signal.
• Laplacian Score is utilized for feature selection according to their importance.
• VPMCD is introduced and employed to achieve fault diagnosis automatically.

A new rolling bearing fault diagnosis method based on multi-scale fuzzy entropy (MFE), Laplacian Score (LS) and variable predictive model-based class discrimination (VPMCD) is proposed in this paper. Compared with previous approximate entropy (ApEn) and sample entropy (SampEn), MFE has taken into account the dynamic nonlinearity, interaction and coupling effects among mechanical components and thus it provides much more hidden information in different scales of vibration signal. Hence, MFE is employed to characterize the complexity and irregularity of rolling bearing vibration signals. Besides, to fulfill an automatical fault diagnosis, the VPMCD, as a new classification approach, is employed to construct a multi-fault classifier for making decision. Also, Laplacian Score (LS) for feature selection is utilized to refine the feature vector by sorting the features according to their importance and correlations with the fault information to eschew a high dimension of feature vector. Finally, the proposed method is implemented to rolling bearing experimental data and the results indicate that the proposed method is able to discriminate the different fault categories and degrees effectively.