Thermal error modeling of a coordinate boring machine based on fuzzy clustering and SVM

To investigate the effect of the thermal characteristics of a motorized spindle system on the precision of a machine tool, two thermal error modeling of a CNC boring machine spindle were proposed, and a five-point method was used to measure the thermal errors of the spindle. The relationships between the spindle speed and temperature field, and thermal errors were analyzed. Then the method combining fuzzy clustering and correlation analysis was presented to optimize temperature variables and select the variables sensitive to thermal error. Subsequently, the least square support vector machine (LS-SVM) and multivariable linear regression analysis (MLRA) models were established for axial elongation and radial declinations. The results indicate that the fuzzy cluster can reduce the multicollinearity among temperature variables and improve the stability of the model. Moreover, the LS-SVM has better generalization than MLRA under different cutting conditions, and the prediction accuracy could reach up to 90%, which could be used to compensate thermal errors of the machine.