Intelligent rolling bearing imbalanced fault diagnosis based on Mel-Frequency Cepstrum Coefficient and Convolutional Neural Networks

• Published in: Measurement : journal of the International Measurement Confederation 2022-12, Vol.205, p.112143, Article 112143
• Main Authors: Yao, Peng, Wang, Jinxi, Zhang, Faye, Li, Wei, Lv, Shanshan, Jiang, Mingshun, Jia, Lei
• Format: Article
• Language: English
• Subjects: Bearing fault diagnosis; Convolutional Neural Networks; Efficient Channel Attention; Mel-Frequency Cepstrum Coefficient; Mode normalization
• ISSN: 0263-2241; 1873-412X
• DOI: 10.1016/j.measurement.2022.112143

•Mel-Frequency Cepstrum Coefficient (MFCC) is adopted to better extract low and medium frequency feature, and use cepstrum lifting technique for feature enhancement.•To improve the domain adaptability of the MECNN proposed, use Mode Normalization to reduce the internal covariant shift caused by data distribution discrepancy, and Effective Channel Attention is adopted to enhance the feature to improve the anti-interference ability.•To evaluate the performance of the MFCC-MECNN method proposed, set 2 types of data distribution shift experiments (data imbalance and operating condition change). To improve the bearing fault diagnosis performance under the condition of data distribution shift, an intelligent diagnosis method based on MFCC (Mel-Frequency Cepstrum Coefficient) and MECNN (Convolutional Neural Networks optimized by Mode Normalization (MN) and Efficient Channel Attention (ECA)) is proposed. Firstly, Mel filters are adopted to extract the feature of different frequency bands of vibration signal, and by the feature enhancement of Cepstrum Lifting Technique, the final 2D MFCC is obtained. Secondly, MN is applied to reduce the internal covariant shift caused by the data distribution discrepancy, and improve the generalization ability. ECA is adopted to enhance the fault feature and improve anti-interference ability. Finally, experiments under data distribution shift have been carried out, and an average accuracy of 99.72% was obtained under the data imbalance, and 99.50% was obtained under the operating condition change. Compared with the existing methods, the proposed has higher accuracy and better domain adaptability.