Cross-Domain Fault Diagnosis with One-Dimensional Convolutional Neural Network

Intelligent fault diagnosis methods based on deep learning have been widely used in intelligent manufacturing. Most of these methods focus on the diagnosis of fault data with the same distribution in a single domain, but pay poor attention to the diagnosis of cross-domain fault data with different distributions. To address this problem, this paper firstly integrates the fault datasets from eight universities into a cross-domain dataset. A new model named one-dimensional improved LeNet-5 (1D ILeNet-5) is proposed for cross-domain fault diagnosis. One-dimensional convolutional operation is used for feature extraction and batch normalization technique is introduced to accelerate the network convergence in this model. The effectiveness and generalization performance of this method are verified using the aforementioned cross-domain dataset. The results demonstrate that our method outperforms the state-of-the-art transfer learning model with fewer parameters and shorter training time.