Abstract:In order to accurately predict the thermal error of the feed system of precision CNC machine tools and improve its quasi-static accuracy, a method based on SFLA optimized least squares support vector machine (SFLA–SVM) is proposed to optimize the key parameters of the thermal error model of feed system. Firstly, the sensitive measuring points are selected by using the spectrum clustering method and relative entropy to extract the measuring points which have a great effect on the thermal error of the feed axis. Secondly, the key parameters of SVM based on SFLA are optimized accurately and quickly, and the thermal error prediction model of SFLA–SVM is established. Finally, the performance comparison of the prediction model is carried out on a precision three-axis CNC machine tool. The experimental results show that the root mean square error of the thermal error prediction value of SFLA–SVM is reduced by 58.53% and 66.0% compared with the GA–SVM and GA–BP in the steady state.