Artificial Intelligence Methodology for Run-to-Target Manufacturing

An efficient management of manufacturing operations is becoming increasingly important in Industry 4.0 (I4.0). Current research aims to show the benefits of incorporating tools from I4.0 into well-known lean manufacturing methods such as run-to-target (RTT) manufacturing. The RTT approach focuses on iterative adjustments of the production processes using data analysis. Herein, we introduce an innovative approach to optimize the data analysis in RTT by combining machine learning and feature selection (FS) with SHapley Additive Explanations (SHAP). Different prediction models were trained using a database collected from the industrial plant of a multinational consumer goods company. This communication demonstrates that the prediction models trained had lower complexity, which is in line with the parsimony principle. Model performance was better after applying the SHAP-assisted FS procedure, showing low errors with a reasonable number of inputs. Consequently, our proposal has advantages in optimizing the data analysis in RTT because it simplifies the whole process and identifies variables driving the quality of the manufacturing process.