Energy consumption optimization in Computer Numerical Control milling operations represents a critical challenge in modern manufacturing, where energy efficiency directly impacts both economic sustainability and environmental footprint. This study presents a comprehensive deep learning framework for energy consumption pattern recognition and prediction in milling processes. An experimental dataset of 5000 samples was collected, encompassing 13 process parameters, including spindle speed, feed rate, cutting forces, tool wear, and vibration signatures. We developed a deep neural network (DNN) architecture consisting of four dense layers (256–32 neurons) with batch normalization and dropout regularization. The proposed DNN model achieved performance with R2 = 0.5902, RMSE = 104.77 W, and MAPE = 10.76%, outperforming traditional machine learning methods, including ridge regression (R2 = 0.5869), random forest (R2 = 0.5580), and gradient boosting (R2 = 0.4931). Feature importance analysis revealed that tool wear (25.12%), cutting forces (42.3% combined), and spindle speed (17.29%) are the most significant contributors to power consumption. Energy efficiency classification achieved 89.33% accuracy with 96.24% recall for high-efficiency operations, enabling real-time monitoring and optimization. This study demonstrates that deep learning approaches can effectively capture complex nonlinear relationships in machining dynamics, providing a foundation for intelligent energy management systems in smart manufacturing.
Sun et al. (Thu,) studied this question.