Leveraging Peripheral Systems Data in the Design of Data-Driven Services to Increase Resource Efficiency

Production and sustainability represent a challenge that still exists today. The demand for more efficient use of resources and operating materials is clear, and possible through the pragmatic integration of digital technologies and the approach of the circular economy along the entire process chain. However, when leaving individual processes, the complexity of data increases since causal effects between the process steps and their impact on the resulting KPIs must be considered simultaneously. This is where data-driven analysis unfolds its full potential. For this purpose, in addition to the manufacturing process itself, it is imperative to consider the often-neglected peripheral systems, including the provision of raw materials, consumables and supplies. In addition to the necessary consistent and cross-process-step data, manufacturing companies and especially small and medium-sized companies lack usable digital services for the demand-oriented control of process and periphery and for event-based instead of time-controlled recommendations for action for staff, maintenance, and management to achieve an increase in resource efficiency. This work provides an approach that addresses prediction, classification, and anomaly detection using modular machine learning models trained on heterogeneous data in the electroplating industry and gives a conceptual outlook for transforming these models into robust edge services for control systems in manufacturing.