Temperature-sensitivity points selection for positioning error modeling of CNC machine tools based on analytic hierarchy process

Abstract

Selecting the appropriate temperature-sensitivity points (TSPS) is the key to robustly predict the thermal error, which is an important part of the thermal error data-driven model (DDM). At present, grouping search is a comprehensive and scientific method to select the TSPS. However, the temperature measurement points with relatively low correlation to thermal error will be selected as the TSPS by this method, which can affect the performance and the stability of the error model's long-term prediction. Therefore, analytic hierarchy process (AHP) is introduced to alleviate this problem. In this paper, AHP is firstly applied to reassign the weights of the TSPS, and then the TSPS are reselected again according to the result. To prove the effectiveness of AHP, extreme learning machine with AHP (ELM-AHP) and other three classic models are taken as examples to construct the thermal error model. The results show that ELM-AHP model has high predictive accuracy and strong robustness. It is proved that AHP can solve the problem of TSPS with low correlation to a certain extent, which can reduce the number of TSPS, and simplify the error model.