A Novel Fault Diagnosis Method for Planetary Gearboxes under Imbalanced Data

To address the issue of fault diagnosis of planetary gearboxes under imbalanced data, a novel fault diagnosis method based on the improved energy-based generative adversarial network (IEBGAN) is proposed. Firstly, convolutional layers are added to the energy-based generative adversarial network (EBGAN) discriminator, thereby improving the feature extraction ability. Then, the classification loss is introduced into the loss function of EBGAN with the purpose of expanding the classification function of the discriminator. Finally, a planetary gearbox fault diagnosis model with sample generation capability is established to achieve the Nash equilibrium by the confrontation between the generator and the discriminator. Experimental results illustrate that the proposed method can improve the accuracy of fault diagnosis for planetary gearboxes even under imbalanced data.