Research on Bearing Variable Condition Fault Diagnosis Based on RDADNN

Due to the influence of working conditions, the data distribution of bearings is challenging to maintain consistency in practical engineering, which leads to the problem of low fault diagnosis accuracy of bearings under variable working conditions. Therefore, this paper proposes a bearing fault diag...

Full description

Saved in:
Bibliographic Details
Published inJournal of failure analysis and prevention Vol. 23; no. 4; pp. 1663 - 1674
Main Authors Jin, Zhenzhen, Sun, Yingqian
Format Journal Article
LanguageEnglish
Published Materials Park Springer Nature B.V 01.08.2023
Subjects
Artificial neural networks
Bearings
Fault diagnosis
Feature extraction
Neural networks
Regularization
Working conditions
Online AccessGet full text

Cover

More Information
Summary:Due to the influence of working conditions, the data distribution of bearings is challenging to maintain consistency in practical engineering, which leads to the problem of low fault diagnosis accuracy of bearings under variable working conditions. Therefore, this paper proposes a bearing fault diagnosis method based on regularized domain adaptive deep neural network (RDADNN). Firstly, a wide convolutional neural network with an embedded squeeze and excitation block module is proposed to improve the source and target domain’s feature extraction effect. Then, the coral criterion is used to match the difference in data distribution between the source domain and target domain, and label regularization is used to improve the model’s generalization ability. Finally, the feasibility of RDADNN is verified by bearing a data set. The experimental results show that the proposed method can effectively realize the cross-domain fault diagnosis of bearings. It performs superior in six cross-domain scenarios in two sets of experiments and has good robustness and generalization.
ISSN:1547-7029
1864-1245
DOI:10.1007/s11668-023-01713-9