Thermal Errors Classification Compensation without Sensor for CNC Machine Tools

Abstract

Both multicollinearity and utilization deficiency of temperature sensors affect the robustness and the prediction precision of traditional thermal error prediction models. To address the problem, a thermal error prediction model without temperature sensors is proposed. Firstly, the paper analyzes the temperature field and thermal deformation mechanisms of the spindle of a CNC gear grinding machine in accordance with the parameters, efficiencies, and structures of the machine spindle and bearing. A preliminary theoretical model is established on the basis of the mechanism analysis. Secondly, the theoretical model is corrected according to the actual operation parameters of the machine. Thirdly, verification experiments are carried out on machine tools of the same type. It is found that the corrected model has higher precision in predicting thermal errors at the same rotational velocity. The standard deviation and the maximum residual error are reduced by at least 39% and 48% separately. The prediction precision decreases with the increase in prediction range when predicting thermal errors at different rotational velocities. The model has high prediction precision and strong robustness in the case of reasonable prediction range and classified prediction. In a word, prediction precision and robustness of the model without temperature sensors can be effectively ensured by reasonably determining the prediction range and practicing classified prediction and compensation for thermal errors at different rotational velocities. The model established can be applied to machine tools that have difficulties in arranging sensors or those whose sensors are significantly disturbed.