Energy Prediction for CNC Machines Using G-Code Evaluation, Machine Learning and a Real-World Training Part

Abstract

Conference Title: 2025 11th International Conference on Mechatronics and Robotics Engineering (ICMRE)

Conference Start Date: 2025, Feb. 24

Conference End Date: 2025, Feb. 26

Conference Location: Lille, France

Rising energy prices have led to increased interest in energy demand forecasts. In manufacturing environments, this creates the need to identify energy-intensive processes. This study aims to assess the energy demand of CNC-manufactured components to identify opportunities for optimization. CNC machines use a programming language (G-code) to control movements, speeds, and tool operations. Standardized G- and M-commands are used to predict the energy requirement before a part is produced. The energy measurements include both the total energy and the individual energies of the axes, spindle, and tool change. Various Machine Learning (ML) models are deployed to forecast the energy consumption of a 5-axis machine tool to determine the most effective algorithm. Cross-validation techniques are used to identify the optimal performance among the considered models. Additionally, to reduce training efforts, real-world components are utilized for both training and validation of the proposed methodology.