However, the practical abnormal data are insufficient for model training, resulting in limited diagnostic performance. To address this problem, we propose an abnormal audio generation method based on ...