Aiming at the small sample problem of cutting saw blade wear state recognition

a recognition model based on data enhancement and attention network is constructed. Firstly

a wear experiment platform was built

and vibration signals during the wear fault were collected

and the vibration signals were denoised with wavelet packet decomposition. Then

the K-nearest neighbor model was used as the scoring standard to optimize the Generative Adversarial Network (GAN)

and the data set was expanded based on the K-GAN (K-Nearest Neighbor and Generative Adversarial Network) model. The time and frequency domain similarity between the generated data and the real data was analyzed

which showed that the generated data was highly similar to the real data. Finally

the attention network is used to identify the wear state

and the recognition accuracy reaches 97.5%. Compared with the model before optimization

the results show that the performance of this model is better than that before optimization.