A ResUNet diffusion adversarial generative network for small-sample bearing fault diagnosis

Rolling bearing fault diagnosis under limited-sample conditions often suffers from insufficient training data. To address this issue, a ResUNet Diffusion Adversarial Generative Network (RUDAGN) is proposed. First, vibration signals are transformed into time–frequency representations using Continuous Wavelet Transform (CWT). Then, a ResUNet architecture is embedded into the Diffusion Model (DM), and a collaborative optimization strategy combining diffusion loss with a lightweight PatchGAN discriminator is introduced to enhance the structural consistency and texture fidelity of generated CWT images. Subsequently, limited-sample datasets are constructed based on the Case Western Reserve University (CWRU) bearing dataset, and RUDAGN is employed for sample augmentation. Finally, a DenseNet classifier is utilized for fault identification. Experimental results demonstrate that the proposed method can generate high-quality samples and effectively improve fault diagnosis accuracy under limited-sample conditions.