Bearing fault diagnosis model based on class domain adaptation

Rolling bearings as a key part of rotating machinery, seriously restrict the power generation efficiency in wind turbines due to frequent failures. However, traditional rolling bearing fault diagnosis methods require the same distribution of training data and test data, which leads to their insufficient generalization ability and cannot effectively solve the problem of unlabeled cross-domain fault diagnosis in practical industry. Therefore, a domain adaptation bearing fault diagnosis method based on class was proposed, which uses labeled source domain data to achieve fault classification of unlabeled target domain. It uses the one-dimensional convolutional neural network as a feature extractor to extract the depth features of original vibration signals, and according to the class of the source domain to design a group of anchor layers to narrow the cross-domain distance between the same class faults and expand the cross-domain distance between different class faults. Moreover, the comparative experimental results on two bearing fault data sets show the effectiveness of the proposed method, which achieves the goal of high-precision cross-domain bearing fault diagnosis, and can be used as an effective tool for cross-domain fault diagnosis. © 2023 Chinese Vibration Engineering Society. All rights reserved.