Feature Extraction Using Refined Composite Hierarchical Fuzzy Dispersion Entropy and its Applications to Bearing Fault Diagnosis

Fault diagnosis of rotating machinery is a key technology to ensure safe and economic operation. However, because of the variable service conditions, machine systems behave nonlinearly and nonsmoothly, resulting in unbalanced bearing vibration signals accompanied by strong noise, and the degradation state of the bearing is hardly accurately represented. Hence, a novel method of effective feature extraction based on refined composite hierarchical fuzzy dispersion entropy (RCHFDE) is proposed in this work. First, RCHFDE is employed to extract the multilevel complexity features of the bearing signals to construct the raw feature space. Second, the original feature space is processed by multilevel noise reduction using the multilevel principal trend denoising (MPTD) method, which removes noise and redundant information to increase the feature space signal-to-noise ratio (SNR) and stability. Third, the low-noise feature space is input into the Bayesian optimization support vector machine (BO-SVM)-based diagnostic framework for training, recognition, and diagnosis. Finally, the advantages of the proposed feature extraction method and fault diagnosis framework are demonstrated through two case studies. The result indicates an effective diagnosis of the working condition and fault location of the bearing with a high recognition rate, high stability, and robustness of the proposed method.