Adaptive Multiscale Weighted Permutation Entropy for Rolling Bearing Fault Diagnosis

Bearing vibration signals contain non-linear and non-stationary features due to instantaneous variations in the operation of rotating machinery. It is important to characterize and analyze the complexity change of the bearing vibration signals so that bearing health conditions can be accurately identified. Entropy measures are non-linear indicators that are applicable to the time series complexity analysis for machine fault diagnosis. In this paper, an improved entropy measure, termed Adaptive Multiscale Weighted Permutation Entropy (AMWPE), is proposed. Then, a new rolling bearing fault diagnosis method is developed based on the AMWPE and multi-class SVM. For comparison, experimental bearing data are analyzed using the AMWPE, compared with the conventional entropy measures, where a multi-class SVM is adopted for fault type classification. Moreover, the robustness of different entropy measures is further studied for the analysis of noisy signals with various Signal-to-Noise Ratios (SNRs). The experimental results have demonstrated the effectiveness of the proposed method in fault diagnosis of rolling bearing under different fault types, severity degrees, and SNR levels.