This article presents an approach for a self-optimizing compensation of tool load induced surface deviations in 5-axis ball-end milling. In order to predict the surface deviation independently from the workpiece geometry, the tool deflection is modelled as a function of the tool engagement using a machine learning approach. For that purpose, a novel description of the cutting conditions in ball-end milling is introduced. The selected features are derived from a process-parallel simulation. Subsequently, the learning behavior, the transferability of process knowledge to other shapes and the feasible compensation are investigated experimentally. It is shown that the developed approach can reduce the shape error by over 70%.

本文提出了一种在五轴球头立铣刀中自动优化补偿刀具载荷引起的表面偏差的方法。为了独立于工件几何形状预测表面偏差，使用机器学习方法将工具挠度建模为工具啮合的函数。为此，介绍了球头铣削中切削条件的新颖描述。所选特征是从过程并行仿真中得出的。随后，通过实验研究了学习行为，过程知识向其他形状的可传递性以及可行的补偿方法。结果表明，所开发的方法可以将形状误差降低70％以上。