UKFNN-based cutting energy consumption model of numerical control machine tools

Abstract

The energy consumption of numerical control (NC) machine tools is mainly from the cutting process. Since the cutting process has multiple characteristics, such as nonlinear, multi-variable and time-varying characteristics, it is difficult to establish an accurate model of the cutting energy consumption by both the static modeling and the traditional mechanism modeling. In this paper, the unscented Kalman filter neural network (UKFNN) is proposed to model the cutting energy consumption of NC machine tools. In the resulting model, the unscented Kalman filter (UKF) is adopted to update weights and thresholds of the neural network in real time, improving the generalization ability and the accuracy of the model. Experimental results show that the proposed method is effective and reliable to model the cutting energy consumption of NC machine tools.