End milling has been widely adopted to machine the thin-plate parts that play increasingly important role in the aerospace industry, due to the advantages of high machining accuracy and fine machined surface quality. In this paper, a systematic method is proposed to predict and compensate the wall thickness errors in end milling of thin-plate parts. The errors are caused by the static deflections induced by the varying cutting force imposed on the weakly rigid part. To improve the efficiency of computing the part deformation, a novel FE model is firstly developed by combing the methods of substructure analysis, special mesh generation and structural static stiffness modification. Then, the time- and position-dependent deformations of the part are calculated based on the proposed FE model to predict the wall thickness errors left on the finished part. It reveals for the first time that the surface topography of the finished thin-plate part is formed by the repeated cutting with the bottom edge of the cutter (BEC) in end milling. Owing to the coupling between the axial cutting depth (ACD) and the force-induced deflection, the modified ACDs for compensation of the static wall thickness errors are finally determined by an iterative adjustment method. The proposed method is verified by three-axis end milling experiments. The experiment results show that the predicted wall thickness errors match well with the really measured ones, and the errors are reduced by 77.18% with the help of the proposed compensation method. Moreover, the proposed FE model reduces the computational time elapsed for error prediction by 67.44% as compared with the benchmark FE model.

由于具有高加工精度和良好的加工表面质量的优势，端铣已广泛用于加工在航空航天中起越来越重要作用的薄板零件。本文提出了一种系统的方法来预测和补偿薄板零件立铣中的壁厚误差。误差是由施加在弱刚性零件上的切削力变化引起的静态挠度引起的。为了提高零件变形的计算效率，首先通过结合子结构分析，特殊网格生成和结构静刚度修改的方法，开发了一种新型的有限元模型。然后，根据所提出的有限元模型计算零件随时间和位置的变形，以预测成品零件上留下的壁厚误差。它首次揭示了成品薄板零件的表面形貌是通过在端铣削中用铣刀（BEC）的底边反复切割而形成的。由于轴向切削深度（ACD）和力引起的挠度之间的耦合，最终通过迭代调整方法确定了用于补偿静态壁厚误差的改良ACD。通过三轴立铣实验验证了该方法的有效性。实验结果表明，预测的壁厚误差与实测值吻合得很好，借助所提出的补偿方法使壁厚误差降低了77.18％。此外，