Cross-domain fault diagnosis of rolling bearings based on deep multi-source sub-domain adaptation networks
In order to overcome the low classification accuracy under small number of samples or that the subclasses of the source domain data set are too close,a deep multi-source subdomain adaptation network(DMSAN)method for bearing fault diagnosis is presented.Firstly,for the problem of small samples in the target domain,the deep convolutional generative adversarial network(DCGAN)is used to expand it.Secondly,the shared features of multiple source domains are obtained through the network branching structure.Then,the local maximum mean discrepancy(LMMD)is used to align the subdomains of each source and target domains.Finally,the weighting module is used to minimize the global loss and realize the joint diagnosis of multiple source domains.Experiments are conducted using a dataset of bearing failures measured at Case Western Reserve University and by building a troubleshooting platform.The experimental results show that the cross-domain fault diagnosis accuracy of the proposed model is higher than that of other domain adaptation comparison models,especially for the target domain with less data.
万方数据
维普
