Real-time compensation for geometric, thermal, and cutting force induced errors in machine tools.

Abstract

Compensation for machine tool errors has been studied for at least two decades, but with most research limited to geometric and thermal errors. However, geometric and thermal error compensation has some barriers such as robustness of the error component model to sensor placement, speed of learning, etc. In almost all the geometric and thermal error compensation work, cutting force-induced errors were ignored because it was felt that in finish machining, the cutting force is small and the resulting deflection can be neglected. The goal of this research is to improve real-time compensation for geometric and thermal errors and to develop a new error compensation scheme which can compensate for cutting force induced errors in a machine tool. Two kinds of empirical models were investigated for thermal error prediction: multiple regression analysis and neural networks. Both approaches correlated the thermal errors in terms of a temperature field. The modeling performance of both approaches has been compared on the two following issues: robustness to sensor placement, and speed of learning. Both issues are very critical in terms of industry implementation to reduce machine downtime. One particular neural network proved superior to other methods studied. A unique form of the planar error model is proposed to combine all the cutting force induced errors in a turning center. There are ten cutting force induced error components for a two-axis turning center. Six error components related to the tool post and the spindle are calibrated using a capacitance sensor system. Two error components related to the cutting tool are calculated theoretically based on the cantilever beam model. And the other two error components related to the workpiece are estimated using the finite element method in order to consider all possible part geometries. The error compensation control is implemented on the turning center based on a software approach, which is referred to as compensation by origin shift. Three different types of cutting tests were conducted to verify the effectiveness of this proposed method. In test I, the maximum diameter error was reduced from 33.0 $\mu$m to 5.1 $\mu$m, in test II, from 22.9 $\mu$m to 5.1 $\mu$m, and in test III, from 27.9 $\mu$m to 10.1 $\mu$m. Overall the maximum diameter error in the workpiece was reduced by 67-85% using this compensation system. Also, a generalized 5-D error model, which included 2 rotary axes, was derived to synthesize both geometric errors and thermal errors of a 5 axis machining center.