Bearing Fault Diagnosis Based on Coordinated Attention ACGAN
Xutong Shu, Lin Zhou, Shilong Wang, Miao Wu, ZiJun Lu, Qiufang Lian · 2024
In this paper, the traditional discriminant bearing fault diagnosis algorithm relies on artificial feature extraction under complex working conditions, and the diagnosis effect is not good. The Auxiliary Classifier Generative Adversarial Network (ACGAN) is applied to the bearing fault diagnosis. Firstly, the bearing vibration signal is converted into 6000 two-dimensional time-frequency images by continuous wavelet transform, an improved Coordinate Attention mechanism is introduced into the producer. The model can pay more attention to the feature relationship between different channels in the data, so as to establish the CA-ACGAN model, Finally, the classification accuracy of CAACGAN model is compared with other mainstream algorithms, and the anti-noise ability of other algorithms is tested in different signal-to-noise ratio. and the classification results are displayed by t-SNE reduction. The results show that the fault diagnosis accuracy of this model is as high as 98.2%, and it has good feature extraction ability, The test shows that the model also has good anti-noise performance