1. 基于 S-MCLSTM 和 DANN 的滚动轴承 剩余寿命预测方法.
- Author
-
董志民 and 董洁超
- Subjects
- *
ROLLER bearings , *WORK design , *GENERALIZATION - Abstract
Aiming at the problems of poor generalization ability and low accuracy of remain useful life prediction of rolling bearings under different working conditions and different faults, this paper proposed a remain useful life prediction method based on Siamese multi-convolutional long short-term memory (S-MCISTM) and domain adversarial neural network (DANN). Firstly, to reduce the influence of different working conditions on the degradation process, it input two samples with a certain time interval to the S-MCLSTM differentiation feature extractor to extract the differentiated features. At the same time, it designed and trained the work condition discriminator adversarially with the feature extractor, which could avoid to extract redundant features due to different working conditions. Then, to reduce the influence of different faults on the degradation process, it designed and trained a fault diagnoser adversarially with the differentiation feature extractor. Finally, considering the differences in the mapping between degradation process and features in different degradation stages, it proposed a stage discriminator and applied different remain useful life predictors in different stages. In the end, experiments on the XJTU-SY bearing dataset show that the method can accurately predict the remain useful life under various working conditions and faults, and has a wide range of application scenarios. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF