Abstract:  For the fault diagnosis of rotating machinery in the actual industrial scene, the variable working conditions of rotating machinery could lead to the distribution difference between the testing samples and the training samples of the fault diagnosis model, which could affect the accuracy of the fault diagnosis model. For this issue, a transfer fault diagnosis method for rotating machinery based on deep belief network（DBN） was proposed. Firstly, the maximum overlap discrete wavelet packet transform was used to process the original vibration signal, extract the statistical features and construct the original feature set; Secondly, based on the labeled feature data of the source domain and the feature data in the normal state of the target domain, the proposed feature selection method based on the Fisher score and maximum mean difference between domains was performed, the features with good discrimination performance and domain invariance were selected for the subsequent model training; Then, by using the pretraining fine-tuning transfer learning method, a transfer deep belief network, which was applied for target domain data fault identification and classification, was constructed. Finally, the bearing and motor fault data collected from SQI-MFS mechanical fault simulation test-bed were used for fault diagnosis experiments under different working conditions. The research results show that selecting features with good discrimination performance and domain invariance for training diagnosis model can significantly improve fault diagnosis accuracy. The fault diagnosis accuracy of bearing and motor under different working conditions can reach 90.83% and 86.83% respectively, which verifies the effectiveness of the proposed migration fault diagnosis framework. In addition, a series of comparative experiments show that the diagnosis performance of the proposed framework is significantly better than the comparison model, which verifies that the proposed method has the potential to be applied to fault diagnosis in practical industrial scenarios.

Key words:  rotating machinery; fault diagnosis; transfer learning; deep belief network(DBN);features selection

LIAO Yu-bo, YU Xiao, LI Wei-sheng, et al. Transfer fault diagnosis for rotating machinery based on deep belief network［J］.Journal of Mechanical & Electrical Engineering, 2022,39(2):193-201.