关键词: 轴承, 连续小波变换, 时频图, 半监督学习, GAN(生成对抗网络), 故障诊断

Abstract: A fault diagnosis method was proposed for rolling bearing based on improved generation adversarial networks(GAN). Taking the vibration signal as the main basis， combined with the ability of continuous wavelet transform to process non-stationary signals and the functions of semi-supervised generation adversarial networks (SSGAN) processing and image recognition for fault diagnosis， the condition model was introduced based on semi-supervised generation adversarial networks， and the loss function was optimized to guide the generator and discriminator. For the first time， the improved GAN algorithm was applied to the field of fault diagnosis and its generation model and semi-supervised learning ability were used to solve the problem of sample data shortage and sample labeling. Experimental results showed that the fault diagnosis method combining continuous wavelet transform and improved GAN can achieve higher accuracy than that of the other mainstream diagnostic methods.

Key words: bearing, continuous wavelet transform, time-frequency representations, semi-supervised learning, generation adversarial networks, fault diagnosis