Spindle thermal error robust modeling using LASSO and LS-SVM

Abstract

To improve the spindle thermal error prediction accuracy, the least absolute shrinkage and selection operator (LASSO) is used to directly select the temperature-sensitive point subset to guarantee the prediction performance of the thermal error model built by least squares support vector machines (LS-SVM). Taking a horizontal machining center as a test stand, the thermal error experiments with different spindle speed states are carried out. Then the temperature-sensitive points are selected using LASSO. The number of temperature-sensitive points is reduced from 20 to 7. Afterward, the thermal error model is designed by LS-SVM. The prediction performance and generalization performance of the thermal error model are compared with another two thermal error models using gray model (GM) and multiple linear regression (MLR), respectively. The comparison results indicate that the thermal error model derived from LS-SVM shows better prediction performance and generalization performance than those derived from GM and MLR with the highest prediction accuracy increasing about 74.6 and 54.3%, respectively. Thus, the feasibility and effectiveness of the proposed spindle thermal error robust modeling method are validated.