Using Model Selection and Reduction to Develop an Empirical Model to Predict Energy Consumption of a CNC Machine

With an ever growing need to reduce energy consumption in the manufacturing industry, process users need to become more aware on how production impacts energy consumption. Computer numerically controlled (CNC) machining tools are a common manufacturing apparatus, and they are known to be energy inefficient. This paper describes the development of an empirical energy consumption model of a CNC with the aim of predicting energy consumption based on the number of parts processed by the machine. The model can then be deployed as part of a decision support (DS) platform, aiding process users to reduce consumption and minimise waste. In using the Calibrated Model Method, the data undergoes initial preparation followed by exploratory data analysis and subsequent model development via iteration. During this analysis, relationships between parameters are explored to find which have the most significant on energy consumption. A training set of 191 datapoints yielded a linear correlation coefficient of 0.95, between the power consumption and total units produced. RMSE, MAPE and MBE validation test yielded results of 0.198, 6.4% and 2.66% respectively.