MANU-ML: Methodology for the application of machine learning in manufacturing processes

Abstract

The advances in interconnectivity and digitalization offer the potential of data-driven approaches in the manufacturing industry. Thereby, the application of machine learning (ML) has gained attention for optimizing manufacturing processes and has helped to understand complex parameter causations. However, companies still struggle to implement ML approaches for manufacturing due to the vast requirements of interdisciplinary knowledge and skills. This work presents MANU-ML, a methodology for applying machine learning in manufacturing processes. After assessing established data mining (DM) and ML methodologies, we modified and extended these methodologies to provide companies with a detailed four-layer model to address the integration from operational technology (OT) to information technology (IT). The model covers hardware inventory and implementation, data transmission, the ML pipeline, and the company’s expertise and goals. The layered design enables identifying upcoming challenges during implementation on a higher level and solving them using multiple working blocks within each layer. Finally, we evaluate MANU-ML by applying it to a manufacturing process for baked goods.