Adversarial adaptive 1-D convolutional neural networks for bearing fault diagnosis under varying working condition
share
In recent years, intelligent fault diagnosis algorithms using machine learning technique have achieved much success. However, due to the fact that in real world industrial applications, the working load is changing all the time and the data distributions are different under different working loads, degradation of the performance of intelligent fault diagnosis methods is very serious. In this paper, a new model called A2CNN is proposed to address the problem. Main contributions are concluded: a new domain adaptation method based on adversarial network is proposed in this paper and the method is suitable for processing 1-D Fourier amplitude in fault diagnosis. The transfer learning based on transferring knowledge of parameters is integrated into the train of the proposed model. It can achieve high accuracy when working load is changed. We also visualize the learned features and the networks to try to analyze the reasons behind the high performance of the model.
READ FULL TEXT