ANN-based power consumption modelling for feed drive systems in CNC machining using systematic data acquisition

The growing focus on sustainable manufacturing has increased the need to reduce the energy consumption of machine tools. For energy-aware optimisation of CNC machining processes, accurate modelling of feed drive power demand is essential. This paper introduces a systematic data-collection procedure and a predictive modelling framework based on artificial neural networks (ANNs) for estimating feed-drive current consumption without additional sensors, relying solely on internal CNC control signals. Experimental results demonstrated that instantaneous axis velocity and acceleration are necessary and sufficient input parameters for the precise prediction of feed drive current consumption. To efficiently generate a wide range of velocity–acceleration combinations, a specially designed motion cycle was developed, enabling dataset acquisition in approximately one minute per axis. The analysis demonstrated that a simple fully connected feed-forward neural network with three hidden layers of 64 neurons each effectively captures the nonlinear relationships between motion parameters and current consumption. The proposed ANN model significantly outperformed traditional second- and third-order regression models, achieving a high coefficient of determination without overfitting. Case studies conducted on a three-axis machining centre confirmed rapid adaptability and strong generalisation capability, with relative prediction errors below 5% for complex trajectories. The proposed framework provides a solid basis for energy-aware tool path optimisation, digital twin applications, and process monitoring. The approach can be extended to spindle power modelling and integrated with cutting force models to enable comprehensive energy optimisation of CNC machining processes.