Integrating complex process knowledge into structural optimization of casting parts enables design proposals to exploit manufacturing processes’ full potential. However, a significant bottleneck for integrating process knowledge is the computational effort necessary for process simulations. In this article, we focused on low-pressure die casting. We used the medial axis transform and the shortest path algorithm to describe geometry-related features that we used as input data for a neural network metamodel, which replaced the casting process simulation. This allowed us to reduce the time for process simulation from multiple hours to a few seconds and, thus, incorporate the metamodel into the topology optimization framework. To reconstruct the geometry, we used an implicit modeling approach in which the modified geometry was built from volume lattices filtered afterward to obtain solid volumes. The approach was tested on two application examples and proved that the metamodel-based results are equivalent to the results obtained using casting process simulations.     «

Integrating complex process knowledge into structural optimization of casting parts enables design proposals to exploit manufacturing processes’ full potential. However, a significant bottleneck for integrating process knowledge is the computational effort necessary for process simulations. In this article, we focused on low-pressure die casting. We used the medial axis transform and the shortest path algorithm to describe geometry-related features that we used as input data for a neural network...     »