Gerrit Felsch (University of Freiburg) Generative design of curved beam metamaterials

Termín: 9. 5. 2023
Odkaz: Odkaz na web
Materials whose properties are determined by their internal architecture in addition to composition — so-called metamaterials — have emerged as a growing field of study over the past decades [1]. These materials are usually assembled from periodically arranged unit cells. While the mechanical properties of these materials can be predicted though finite element simulations, many applications also require to identify architectures with specific target properties [2]. To solve this inverse problem, we introduce a deep-learning framework for generating metamaterials with desired properties. By supplying the generative model with a guide structure in addition to the target properties, we are not only able to generate a large number of alternative architectures with the same properties, but also to express preference for a specific shape. To demonstrate the capabilities of this approach we applied it to generate unit cells for a new class of reentrant-hexagonal metamaterials based on curved beams. Reentrant-hexagonal metamaterials are well known to be able to exhibit a wide range of different material properties based on their architecture. This includes properties tied to unusual behavior such as a negative Poisson’s ratio [3], which can be tuned by adjusting the angles between beams. However, changing the angles also influences the overall dimensions of the unit cells. By replacing straight beams with curved ones, it is possible to control the Poisson’s ratio of reentrant-hexagonal metamaterials without affecting the overall dimensions. We show that our deep-learning framework is able to accurately generate unit cells fitting specific properties for curved beam metamaterials.

09.05.2023, 15.30 - 16.00, Room B-366 @ Thákurova 7, 166 29 Prague 6

Za stránku zodpovídá: Ing. Mgr. Radovan Suk