The manufacturing process linear flow splitting enables resource-efficient production of branched profiles from flat sheets. Due to their geometry and manufacturing-related properties, these sheets are particularly well suited for applications in the automotive industry. The tool system used in this process offers high flexibility, which makes precise adjustment and stable operation a complex task.?A large number of input and output variables must be continuously monitored and optimized.

The aim of this thesis is to develop a model-based prediction system for feed curves using reinforcement learning. To achieve this, extensive process data from an automated production line will be analyzed and incorporated into a data-driven model. In addition to predicting and optimizing the feed curves, the early detection of deviations plays a crucial role in ensuring product quality and increasing the efficiency of this highly flexible process. By correlating process parameters with quality characteristics, the work will identify opportunities for continuous process improvement.

Research method

Theoretical, experimental