关键词: 传动系统, 主轴承故障, 故障诊断, 辅助分类器生成对抗网络, 堆叠降噪自编码器

Abstract: In order to solve the problem of low accuracy of fault diagnosis for main bearing of transmission system, anACGAN-SDAE fault diagnosis model was proposed, which combined auxiliary classifier generating adversarial networks (ACGAN) and stackeddenoising auto encoder (SDAE). ACGAN was used to generate high-quality new samples to expand the sample size of the main bearing fault of the transmission system, and SDAE was used to extract robust features from noisy samples to improve the accuracy of fault diagnosis. The simulation results show that the ACGAN-SDAE fault diagnosis model can effectively diagnose the faults of the main bearing of the transmission system under different fault sample sizes, and has good domain self adaptability and anti-noise performance. The average fault diagnosis accuracy is more than 90%. Compared with the common fault diagnosis models of SDAE, SVM and MLP, ACGAN-SDAE has certain advantages.

Key words: drive system, main bearing fault, fault diagnosis, auxiliary classifier generative adversarial networks(ACGAN), stacked denoising auto encoder(SDAE)