Generation of clean parametric CAD models from topology optimization results
Topology optimization is a method for the automatic generation of complex components. Typical applications are mechanically stressed components, designed to be as light as possible without failing under the applied loads. For this procedure automatic solvers are available, which try to find the best possible solution. Frequently, truss-like structural designs or other types of substructures are generated, although the model itself does not provide any more detailed information about these. If the component is to be manufactured as a differential design from individual parts, these substructures currently still have to be identified and modeled manually.
The aim of the project is to develop a method for the automated identification of manufacturing-relevant substructures within a topology-optimized structure.
To identify structural regions, the FE-elements of the topology-optimized structure are first classified based on their stress tensor, grouped into substructures and subdivided into strut and node regions. The structures are then transformed from the FE mesh representation into a parametric form. In the process, similar substructures occurring at several points in the overall component shall be identified, which can later be manufactured as identical parts if desired. To ensure data consistency throughout the process, a central data model is used, which is implemented in a graph-based design language.
The automated subdivision process eliminates the need for time-consuming manual geometry reconstruction. This advantage increases with each generated component, making the generation of different product variants in particular significantly more efficient.
Research Coordinator "Mobility Technologies"