AI for Energy Consumption Optimization in CNC Milling

Abstract

Improving energy efficiency in CNC milling is essential for reducing manufacturing costs and environmental impacts. This paper presents a deep learning-assisted framework for predicting and minimizing energy consumption through cutting parameter optimization. A convolutional neural network (CNN) implemented in MATLAB is trained using machining condition descriptors (e.g., cutting speed, feed rate, and cutting depths) to model nonlinear relationships with power and energy. The trained model is then coupled with a constrained optimization procedure to identify the parameters that reduce energy consumption while meeting surface integrity requirements. On the experimental/simulation dataset (N = 216 operating points at 8 parameters, equivalent to 1728 cells), the proposed model achieves very robust prediction accuracy, with the correlation coefficient $\mathbf{R}$ reaching values of over 99% for the training, validation, and test datasets, respectively, and enables energy optimization compared to the selection of reference parameters, indicating that the prediction quality is nearly perfect. The results indicate that CNNbased surrogate modeling can support rapid energy-aware parameter selection and provide a practical path toward datadriven energy management in Industry 4.0 machining environments.