Rolling Bearing Fault Diagnosis Based on Multi-scale Entropy Feature and Ensemble Learning

Aiming at the problem of feature extraction and fault recognition for rolling bearings, a fault diagnosis mthod based on multi-scale entropy and ensemble learning is proposed in this paper. Firstly, the variable mode decomposition algorithm is used to decompose the vibration signal, and then the cross-correlation number method is used to reconstruct the signal to realize the signal denoising. Subsequently, in order to improve the effectiveness of feature extraction for rolling bearings, a feature extraction method based on Refined Composite Multiscale Reverse Permutation Entropy (RCMRPE) is proposed. Then, in order to improve the accuracy of rolling bearing fault identification, this paper proposes a fault diagnosis model based on Stacking- CatBoost ensemble learning. Finally, relevant experiments were conducted on signal denoising, feature extraction, and fault recognition. The RCMRPE entropy extraction method was compared with the common entropy extraction methods, and the proposed fault diagnosis model was compared with the common machine learning models. The experimental results show that the feature extraction error based on RCMRPE is small and can comprehensively reflect the actual fault information of bearings; the accuracy and recall of the fault diagnosis model based on Stacking- CatBoost ensemble learning are both above 99%, and the diagnostic effect is significantly better than other models.