Look-ahead prediction of spindle thermal errors with on-machine measurement and the cubic exponential smoothing-unscented Kalman filtering-based temperature prediction model of the machine tools

Abstract

Reducing the thermal equilibrium time is one effective way to improving the efficiency and utilization and ensuring the accuracy of machine tools. This paper proposed the look-ahead prediction of spindle thermal errors based on CES-UKF temperature prediction model using on-machine measurement. It can predict the thermal errors during the warm-up stage just with the measured temperature data in the required on-machine measurement time (ROMT). With the thermal error compensation function, the machine tool only needs warm-up for ROMT to improve the efficiency and ensuring the accuracy of machine tools. The look-ahead temperature prediction is realized combining UKF and CES including real-time adjustment of exponential smoothing coefficients and ROMT calculation. Temperature-sensitive point selection based on self-attention mechanism is established. Gate recurrent unit (GRU)-based thermal error model is proposed considering the thermal hysteresis effects. Verifications and cutting experiments are carried out to testify the effectiveness and robustness of the method.