3 / 2021-05-12 15:07:58

Multisource Domain Transfer Learning for Bearing Fault Diagnosis

Deep Learning-based fault diagnostic methods assume that training and testing data share the same distribution. This assumption will not hold in practical scenarios due to the variable working conditions of rotating machineries. Transfer learning (TL) overcomes this problem by utilizing knowledge learned from the source domain to help target tasks. However, most TL-based fault diagnostic studies have focused only on single-source TL, while useful multisource domains with sufficient labeled samples are available. In this work, a novel multisource TL model called the moment matching-based intraclass multisource domain adaptation network is proposed. This model uses a feature learner to generate features of each source and target domain data to enable the joint weight classifier to predict target labels. The model also introduces a moment matching-based distance metric to reduce distance among all source and target domains. During the training of the model, an intraclass alignment training strategy is applied to simultaneously match the marginal and conditional distributions of each domain. Experiments under four load conditions are performed, whose results validate the proposed model’s reliability and generalizability.