Milling deformation prediction for thin-walled components based on fusion model

Milling is the primary machining method for manufacturing thin-walled workpieces. Due to the low rigidity, these workpieces are easy to deform during the milling process, which significantly affects machining accuracy. This paper proposes a fusion model to predict the milling deformation based on the U-NET neural network and the physical model. Firstly, a deformation measurement experimental platform is designed to capture the deformation distribution on the surface of the thin-walled workpiece. Secondly, an approximate solution for the theoretical deformation caused by axial cutting force is obtained based on the elastic deformation behavior of the thin-walled workpiece. At last, the captured original deformation and the theoretical deformation are then used as inputs for the fusion model to predict the relative deformation distribution of the machined thin-walled workpiece. The end milling experiment of the thin wall cavity is conducted to validate the accuracy of the prediction results and the effectiveness of the fusion model, the deformation in the milling process is predicted by the U-NET model and fusion model, and then compared with the measured deformation. The result shows that the proposed fusion model achieved an average root mean square error of less than 2.65% on the normalized dataset, providing accurate predictions close to measurements and with the speed of the physical model. Furthermore, compared to the data-driven U-NET model, the fusion model exhibited significant improvement in prediction performance on the test set, demonstrating good physical interpretability and generalization ability. The fusion model is used to realize the high-precision and high-speed milling deformation prediction, to guide the milling process planning of thin-walled parts, provide a basis for realizing the online machining compensation of thin-walled parts, and help to improve the machining quality and processing efficiency of thin-walled parts.