基于迁移学习的跨域滚动轴承剩余寿命预测研究

打开文本图片集

中图分类号：TH133.33 文献标志码：A文章编号：1006-0316（2025）07-0017-08

doi： 10.3969/j.issn.1006-0316.2025.07.003

Abstract ：In the data-driven prediction ofremaining useful life ofrolling bearings，in actual working conditions， due to the limited amount of degradation data and changes in working conditions，there are generally discrepancies between training dataand testing data. It results in the reduction in the performance of remaining useful life （RUL）prediction.Aimingat solving the problem of insufficient data volume and inconsistent distribution of data characteristics under different operating conditions，this paper constructs atransfer learning life prediction model based on Convolutional Neural Network with Long Short Term Memory （CNNLSTM）. Domain adaptation is achieved between the source domain and the target domain by using the Maximum Mean Discrepancy （MMD）metric combined with adversarial transfer strategies.The distribution between the two domains is narrowed so as to extract domain invariant features more effectively.The transfer learning method is used to predict the lifespan of roling bearings under diffrent working conditions. Comparative experiments are conducted among CNNLSTM，Domain adaptive Neural Network （DaNN），and Domain Adversarial Training of Neural Networks （DANN）. The results indicate that the method proposed in this paper is superior to the above three methods and has better accuracy in prediction ofremaining useful life of cross-domain rolling bearings.

Key words rolling bearings iremaining useful life prediction iconvolutional neural network with long short term memory ; transfer learning

近年来，深度学习已经在图像处理、语音识别等领域取得了很多研究成果。（剩余9112字）