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

•The look-ahead prediction of spindle thermal errors is realized with temperature data in ROMT to reduce the warm-up time.•The look-ahead prediction temperature model is established by combing the cubic exponential smoothing and UKF.•The temperature-sensitive point selection is proposed based on self-attention mechanism with the attention value.•Thermal error modeling with input of temperature-sensitive points is proposed by GRU to represent the thermal hysteresis.•The effectiveness is testified and errors of parts with look-ahead prediction and compensation are reduced by>68 %.