Berlin 2024 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 29: Poster II

MM 29.17: Poster

Dienstag, 19. März 2024, 17:00–19:00, Poster B

Multi-scale modelling and machine learning based simulation of the mechanical behaviour of graphite-resin composites — •Tobias Stegmüller — DLR, Institut für Test und Simulation von Gasturbinen, Am Technologiezentrum 5, 86159 Augsburg

To simulate the mechanical behaviour of a structural component it is important to link the microstructural features and properties of the construction material with its macroscopic shape and the acting forces. To achieve this we developed a multi-scale approach that constitutes of the following steps: First, the microstructure and its properties are collected by CT scans and mechanical tests, which are used to conduct FEM simulations of representative volume elements (RVE) that study the deformation behaviour. The response of the RVE is then homogenised over its volume and the homogenised properties are used for FEM simulations on the macroscopic length scale. Finally, the results of these simulations are used as training data for a machine learning algorithm, which is in the end capable of predicting the mechanical behaviour of structural components. The approach as well as its application to a graphite-resin composite are going to be presented.

Keywords: Machine learning; Finite element method; Multi-scale modelling; Graphite