Fault Diagnosis of Variable Working Conditions Based on Transfer Learning and Multi-channel CNN-LSTM Network

As an indispensable component in the operation of machines, rolling bearings play an important role in modern industrial systems. In order to accurately diagnose the target domain fault data in this context, in this paper, the variable working condition dataset is firstly constructed. Further, the discrete Fourier transform (DFT) is applied to the original vibration signals, which can obtain the frequency domain signals. Then, on the basis of the domain adversarial neural networks(DANN) network, the multi-channel convolutional long short-term memory neural network(CNN-LSTM) network is utilized to perform feature fusion, the feature extraction module is reconstructed, and a cross-domain fault diagnosis model is established. Finally, the experimental results show that the proposed network model can extract the temporal characteristics of vibration signals more fully, improve the diagnostic performance of target domain datasets under different variable working conditions, and have good generalization performance under different working conditions.