Remaining Useful Life Prediction Approach for Aviation Bearings Based on Multigenerator Generative Adversarial Network and CBAM

The remaining useful life (RUL) prediction of aviation bearings is of paramount importance in reducing additional maintenance costs and enhancing the monitoring and reliability of aviation equipment, while is still a challenge especially when data samples are few and hard to analyze. Although scholars have done a lot of research on it, the data processing approach with the efficiency of prediction is still not up to expectations. A novel approach based on multigenerator generative adversarial network (GAN) and convolutional block attention module (CBAM) is proposed, which can effectively solve the above two problems. In this article, we first build a multigenerator GAN, utilizing the Wasserstein GAN with gradient penalty (WGAN-GP) framework to augment data. The efficiency and quality of data generated by the generators are improved through weight sharing and feature fusion strategies. Then, the augmented data are fed into the constructed convolutional neural network (CNN) to extract unhealthy data with degenerate features, while reducing the learning burden of the prediction network. Finally, the fully processed data are fed into the network based on CBAM and long short-term memory (LSTM) to get the RUL prediction of bearings. The proposed approach is evaluated using the public FEMTO dataset. Experimental results show that the proposed approach is effective in improving the accuracy and stability of RUL prediction for aviation bearings.