Data Augmentation Integrated with Feature-Enhanced Convolutional Neural Network for Imbalanced Fault Diagnosis in Rolling Bearings

Imbalance in fault samples is a key problem with most rolling bearing datasets in real industrial processes, limiting the performance of fault detection. In addition, existing fault diagnosis methods for rolling bearings do not adopt a modular division between the initial layer and the classification layer. This lack of modularization inhibits the full activation of the network's feature extraction capabilities and impedes subsequent model improvements. In this paper, a novel fault diagnosis method based on data augmentation integrated with a feature-enhanced convolutional neural network (DA-FECNN) is proposed. First, Variational Autoencoder-Generative Adversarial Network (VAE-GAN) is utilized to expand the number of fault samples and improve the diversity of few-shot. Second, a uniform feature-enhanced structure is designed with a feature extraction module, a multi-scale classification module, and a reconstruction module. The network's ability on feature extraction is improved by focusing on high quality features. Thereby, accurate fault classification is realized simultaneously. Experiments on real world data show the effectiveness of the proposed method on fault diagnosis task.