The Thermal Error Modeling Approach of Feeding Axes for Horizontal CNC Lathes Based on Transfer Learning

Abstract Modeling the thermal error for each feeding axis of the CNC machine one by one will lead to an exponential increase in the thermal error compensation time. Th e feeding axes in the same CNC machine tool have similar thermal characteristics. The thermal error modeling approach based on transfer learning was proposed to improve the efficiency of thermal error modeling. The thermal error module was bulit for the X-axis of the horizontal lathe. Then, the thermal error module of the X-axis was transferred to the Z-axis of the same lathe by transfer learning based on domain adaptation. After verification, the prediction residual range of the Z-axis thermal error module based on transfer learning was only -3.2μm~1.4μm, which has proved that the method proposed in this paper is effective in thermal error modeling of the feeding axis. It will provide an effective means to effectively improve the efficiency of thermal error modeling for the feeding axes of CNC machine tools.