DFG Research Group - AI-based Methodology for the Fast Maturation of Immature Manufacturing Processes

  • Contact:

    M.Sc. Johannes Mitsch

  • Funding:


    Deutsche Forschungsgemeinschaft (DFG)

  • Partner:


  • Startdate:


  • Enddate:

    31.12.2026 (additional second funding period planned)

AI-based Methodology for the Fast Maturation of Immature Manufacturing Processes

New processing technologies allow processing of complex materials and yield high-performance components. But even if the underlying working principles are well understood and the technology has proven itself on a laboratory scale, there is still a long way to go before it can be used in industrial series production. This is because each technology has to be adjusted to the individual product, which in practice requires extensive trial campaigns and is accompanied by high reject rates - an expensive bottleneck.

Artificial intelligence (AI) techniques may help bring such processes into active production more quickly. Within the DFG research group " AI-based Methodology for the Fast Maturation of Immature Manufacturing Processes", eight institutes from the engineering and computer science are looking for fundamentally new solutions.

The approach is to combine data from experiments, physics-based simulation and their evaluation with machine learning (ML) methods. For this purpose, the example process - the forming of thermoplastic tapes ("stamp forming") - is sensor- and actor-wise overinstrumented along with da digital twin for simulation. In this way, maximum process insight and forceability is achieved. AI‑methods then enable the analysis of the considerably more complex state and action spaces and provide recommendations for process control.

KIT-FAST-LB takes over the simulative mapping of the process on two levels: Firstly, with detailed and computationally intensive finite element simulations, and secondly, with specifically simplified and therefore significantly faster approaches. The accurate simulations support the understanding of the process and complement experimental process data, while the simplified models support the AI methods with fast process responses.

Work package at FAST-LB:Simulation of the manufacturing process
Project workflow and collaboration