One of the challenges in rolling bearing fault diagnosis is the limited availability of fault sample data. This necessitates the use of data-driven fault diagnosis models, which require significant time and computational resources during the training process. To enhance the accuracy of fault identification, a transfer learning intelligent fault diagnosis method for rolling bearings based on TTS-GAN (Transformer Time-Series GAN Model Architecture) model was proposed. This employs data augmentation-based mechanical fault generation adversarial network to obtain sufficient expanded fault samples for fault classification. Firstly, the vibration signals of rolling bearings in different health categories are generated into corresponding training set and test set; secondly, the pre-trained TTS-GAN model is transferred to the training set through parameter sharing and fine-tuning to achieve the optimal parameters of the model; finally, the fault diagnosis capability of the model is verified by the test set. The method is validated on the Case Western Reserve University (CWRU) rolling bearing dataset.