Cross-Domain Fault Diagnosis Method for Rotating Machinery Based on Multi-Representation Adaptation Neural Network

The success of the fault diagnosis method based on deep learning depends on two conditions: 1) The training and test data sets must come from the same data distribution. 2) A large amount of labeled data with fault information is required. However, the working environment of rotating machinery is usually harsh and complex, the speed and load will change constantly, and it is very difficult to obtain plenty of labeled data with fault information. To solve these problems, a cross-domain rotating machinery fault diagnosis method is proposed. Firstly, the vibration signal of time series is transformed into a time-frequency image by continuous wavelet transform. Then the multi-representation adaptation network model is used for cross-domain fault diagnosis. Finally, the effectiveness of the proposed method is verified by six transfer fault diagnosis tasks.