Abstract: At present, transfer learning of machine fault is a relatively popular research, its main problem is the imbalance of training data caused by the lack of actual fault data. The existing ...