Domain Adaptive Fault Diagnosis Method for Rotating Machinery Based on MMD and CORAL Metrics
In industrial production,due to the difference in the distribution of source domain data and target domain data and the small amount of labeled fault data,traditional rotating machinery bearing fault diagnosis methods are difficult to achieve effective cross-domain fault diagnosis.In order to solve the problem of distribution alignment and knowledge transfer between target domain and source domain,many domain adaptive methods are proposed.However,they mostly focus on marginal distribution alignment(MDA)and ignore conditional distribution alignment(CDA).In view of this,this paper proposes a migration domain adaptive fault diagnosis model based on joint distribution.In the domain adaptive module,in order to enhance the distribution alignment of two domains,the edge distribution and conditional distribution are matched.In order to eliminate domain confusion,the maximum mean difference(MMD)and correlation alignment(CORAL)are combined as a new distribution difference measure.In this paper,the bearing open data set JNU of Jiangnan University is used for validation.The experimental results show that the proposed method has higher diagnostic accuracy than the common domain adaptation methods,indicating that the method can effectively learn the transferable features.
transfer learningdomain adaptationjoint distributionfault diagnosis
万方数据
维普
