A novel method for fast identification of a machine tool selected point temperature rise based on an adaptive unscented Kalman filter

Abstract

A novel method is presented for fast identification of a machine tool selected point temperature rise, based on an adaptive unscented Kalman filter. The major advantage of the method is its ability to predict the selected point temperature rise in a short period of measuring time, like 30 min, instead of 3 to 6 h in conventional temperature rise tests. A fast identification algorithm is proposed to predict the selected point temperature rise and the steady-state temperature. An adaptive law is applied to adjust parameters dynamically by the actual measured temperature, which can effectively avoid the failure of prediction. A vertical machining center was used to validate the effectiveness of the presented method. Taking any selected point, we could identify the temperature rise at that point in 28 min. However, if the method was not used, it took 394 min to obtain the temperature rise curve from the start-up of the machine tool to the time when it reached a steady-state temperature. The root mean square error (RMSE) between the estimated and measured temperatures in the period of 394 min was 0.1291 °C, and the error between the estimated and measured steady-state temperatures was 0.097 °C. Therefore, this method can effectively and quickly identify a machine tool selected point temperature rise.