中国农机化学报2026，Vol.47Issue(2)：188-195,8.DOI:10.13733/j.jcam.issn.2095-5553.2026.02.025
基于两阶段动态对抗迁移建模的轴承故障诊断
Fault diagnosis of bearings based on two-stage dynamic adversarial discriminant migration modeling
摘要
Abstract
Aiming at the problems that the use of gradient inversion layer in traditional fault diagnosis methods is easy to cause gradient disappearance,and the similarity between source and target domain samples cannot be dynamically judged due to the difference in distribution between the two distributions,resulting in very low generalization ability of fault diagnosis model,a two-stage dynamic adversarial migration network(TSDAMN)is put forward to realize the migration of bearing fault diagnosis.This network introduces a new two-step antagonism strategy and dynamic discriminant idea,by using a distributed adversarial strategy,the classifier and discriminator are interactively trained step by step,effectively solving the problem of gradient disappearing.Then,a dynamic adaptive learning algorithm is proposed,which can effectively improve the fault feature clustering ability and improve the accuracy of fault diagnosis by dynamically adjusting the weight of edge distribution and condition distribution in the whole migration process.Finally,θ control is added to improve the cross entropy loss function to further refine the accuracy of fault classification.The bearing data set of Case Western Reserve University is used for experimental verification and compared with the classical algorithm,the average accuracy of fault classification of the proposed method reaches 98.67%.The results show that the proposed method can classify faults more accurately and solve the problem of lack of annotation.关键词
故障诊断/迁移学习/梯度消失/两步式对抗/动态判别/滚动轴承Key words
fault diagnosis/transfer learning/gradient vanishing/two-step adversarial/dynamic antagonistic/rolling bearings分类
机械制造引用本文复制引用
Liu Yuewen,Xu Fan,Li Yongting,Qi Yongsheng,Wang Shunli..基于两阶段动态对抗迁移建模的轴承故障诊断[J].中国农机化学报,2026,47(2):188-195,8.基金项目
国家自然科学基金(62363029) （62363029）
内蒙古自然科学基金(2025MS06038) （2025MS06038）
内蒙古自治区高等学校科学研究项目(NJZY22365) （NJZY22365）
呼和浩特市科技创新领域人才项目(2023RC—联合体—3) （2023RC—联合体—3）
