Machine Tools Thermal Error Modeling with Imbalanced Data Based on Transfer Learning

Thermal error caused by the thermal deformation is one of the most significant factors influencing the accuracy of the machine tool. Complicated working condition lead to difficulties in data collection. Compare to the processing state, it is relatively easier to collect the data of the rest and idling states. The difficulty of data collection results in a large amount of data collected in resting and idling conditions, and a small amount of data in processing conditions. To overcome the limitation of data imbalance during the thermal error modeling of CNC machine tools, the paper proposed a transfer learning thermal error modeling method and performs similarity analysis on imbalanced experimental data by calculating the maximum mean difference. Through experiments, we conclude that the proposed method with insufficient data volume is superior to traditional machine learning methods in terms of prediction accuracy and model training efficiency.