An Improved Robust Thermal Error Prediction Approach for CNC Machine Tools

Thermal errors significantly affect the accurate performance of computer numerical control (CNC) machine tools. In this paper, an improved robust thermal error prediction approach is proposed for CNC tools based on adaptive Least Absolute Shrinkage and Selection Operator (LASSO) eXtreme Gradient Boosting (XGBoost) algorithms. Specifically, LASSO method enjoys oracle property selecting temperature-sensitive variables. After variable selection, XGBoost algorithm further adopted to model predict errors. Since decision tree based, it has natural advantages address multicollinearity provide interpretable results. Furthermore, experimental data from Vcenter-55 type 3-axis vertical machining center, compared with benchmark methods demonstrate its superior accuracy 7.05 ?m (over 14.5% improvement), robustness 5.61 12.9% worst-case scenario predictions 16.49 25.0% percentage 13.33% 10.7% improvement). Finally, real-world applicability verified through compensation experiments.