@inproceedings{DannenSchindelePruemmeretal.2022,
  author    = {Dannen, Tammo and Schindele, Benedikt and Pr{\"u}mmer, Marcel and Arntz, Kristian and Bergs, Thomas},
  title     = {Methodology for the self-optimizing determination of additive manufacturing process eligibility and optimization potentials in toolmaking},
  series = {Procedia CIRP},
  volume    = {107},
  booktitle = {Procedia CIRP},
  publisher = {Elsevier},
  address   = {Amsterdam},
  issn      = {2212-8271},
  doi       = {10.1016/j.procir.2022.05.188},
  pages     = {1539 -- 1544},
  year      = {2022},
  abstract  = {Additive Manufacturing (AM) of metallic workpieces faces a continuously rising technological relevance and market size. Producing complex or highly strained unique workpieces is a significant field of application, making AM highly relevant for tool components. Its successful economic application requires systematic workpiece based decisions and optimizations. Considering geometric and technological requirements as well as the necessary post-processing makes deciding effortful and requires in-depth knowledge. As design is usually adjusted to established manufacturing, associated technological and strategic potentials are often neglected. To embed AM in a future proof industrial environment, software-based self-learning tools are necessary. Integrated into production planning, they enable companies to unlock the potentials of AM efficiently. This paper presents an appropriate methodology for the analysis of process-specific AM-eligibility and optimization potential, added up by concrete optimization proposals. For an integrated workpiece characterization, proven methods are enlarged by tooling-specific figures. The first stage of the approach specifies the model's initialization. A learning set of tooling components is described using the developed key figure system. Based on this, a set of applicable rules for workpiece-specific result determination is generated through clustering and expert evaluation. Within the following application stage, strategic orientation is quantified and workpieces of interest are described using the developed key figures. Subsequently, the retrieved information is used for automatically generating specific recommendations relying on the generated ruleset of stage one. Finally, actual experiences regarding the recommendations are gathered within stage three. Statistic learning transfers those to the generated ruleset leading to a continuously deepening knowledge base. This process enables a steady improvement in output quality.},
  language  = {en}
}