Study on optimal independent variables for the thermal error model of CNC machine tools

In the technology of thermal error compensation for CNC machine tools, it is particularly important to select modeling variables which can stably reflect the relationship between temperature field and thermal expansion in terms of modeling. This paper analyzes the theories and experiments on the thermal properties of the temperature-sensitive points distributed on one-dimension pole. It is found that the prediction model performs better in prediction accuracy and robustness when established with linear points as independent variables than with nonlinear ones. However, because of the complicated structure of machine tools, it is rather hard to fix the positions of linear points, which consequently lead to the proposal of a comprehensive temperature-feature extraction method that uses feature extraction algorithm and weight optimization to construct linear temperature-sensitive points. Experimental facilities verified the feasibility of its proposal. What’s more, based on the effectiveness of building linear measuring points, it is proposed to arrange the temperature sensors along the deforming direction. With the feeding system of a gantry machine tool as the testing platform, the thermal error model established according to the proposed method is actually tested under different working conditions. The result shows this proposed method has higher prediction precision and robustness.