Generative Adversarial Network based Data Augmentation Method for Bearing Fault diagnosis

As a critical component of rotating machinery, the fault diagnosis of bearings holds significant importance. The deep learning-based approach to bearing fault diagnosis necessitates a substantial amount of fault data, which is often challenging to obtain in practical engineering scenarios. This data imbalance issue severely impacts the accuracy of fault diagnosis. To address this challenge, we propose a data augmentation method based on Multi-Source Generative Adversarial Network with Quality Assessment and Screening Module (MGAN-QASM). Firstly, the generator and discriminator are restructured using three loss functions, enhancing the stability of training and diversifying the generated data. Secondly, QASM screens for both similar and diverse samples to balance the training dataset. Finally, Deep Convolutional Neural Networks with Wide first-layer kernels (WDCNN) are employed for fault diagnosis. The results show that the proposed method has higher fault diagnosis accuracy than other methods under different imbalance ratio conditions.