Method for rolling bearing fault diagnosis under variable working conditions based on mixed noise dictionary and transfer subspace learning

Rolling bearings are affected by complex working conditions in actual operation, which makes the vibration signals cannot always satisfy the independent and identical distribution of training and test data. At the same time,the vibration signal is usually mixed with a large amount of noise and irrelevant information, which directly affects the bearing fault diagnosis ability. Therefore, a rolling bearing fault diagnosis method based on mixed noise dictionary and transfer subspace learning under variable working conditions was presented. First, a mixed noise dictionary model was constructed to remove the interference of irrelevant information components on dictionary learning. Then, a transfer subspace model was constructed to transfer the sparse signals into a common subspace. The distribution difference between the two domains was reduced by introducing the method of joint distribution adaptation and reducing the classification error of the source domain. Finally, an alternating direction method of multipliers was used to optimize the solution. The experimental results show that the proposed method can accurately identify rolling bearing fault types under complex variable working conditions. © 2022 Chinese Vibration Engineering Society. All rights reserved.