Research on asymmetrical edge tool wear prediction in milling TC4 titanium alloy using deep learning

TC4 being a difficult material to machine, rapid tool wear during milling leads to surface deterioration and increased manufacturing costs. It has been shown that edge geometry has a significant impact on tool wear. Failure to understand the wear of these complex edges during the cutting process can result in significant economic losses. This paper presents a tool wear prediction method for predicting asymmetrical edged tools with different shape factors. The model takes the cutting force signal as input and tool wear as output. Feature extraction of cutting force signals is performed by stacked sparse autoencoder(SSAE) networks, and the relationship between depth features and tool wear is established by BP neural networks. In order to improve the prediction accuracy and generalization ability of the model, an improved loss function with sparse and weight penalty terms is used as the loss function of the SSAE model. Compared with traditional machine learning methods, this deep learning feature extraction method can effectively avoid relying on a priori knowledge. To verify the superiority of the model, it is compared with the traditional neural network model based on manual feature extraction and a support vector regression model. The root mean square error (RMSE) of the proposed model reaches a minimum of 3.41 and the coefficients of determination (R2) are above 0.8, and these performance indicators are better than the other two models. This indicates that the proposed model has higher prediction accuracy.