The project aims to develop an assistance system that facilitates a more resource-efficient target configuration of process parameters by mapping optical quality features of the product and its process variables in an AI model during manufacturing. For the machine learning process, optical image features, such as surface texture, weld pool, droplets and meniscus, as well as associated process data for the addressed additive manufacturing processes are provided by the application partners via a cloud interface. The AI model is demonstrated for two application scenarios: additive metal deposition welding and drop-on-demand processes for personalized drug printing, following closely to production conditions.
The approach will allow systematic recognition of required cognitive human capabilities in reaction to deviations in additive manufacturing, correlation from trained experiential knowledge of actors with process characteristics and initiation of adequate measures in real time via machine control with transfer to the machine capabilities of AI. The innovative aspect of the presented solution approach that goes beyond the state of the art therefore consists of machine recognition of new quality characteristics, correlation to relevant process characteristics with corresponding process manipulated variables, and the appropriate adjustment of these manipulated variables via the process signals.
Specific project goals of PSI