Inverse Design of Metamaterials

The inverse design of metamaterials aims to systematically develop microstructures that achieve prescribed macroscopic material properties. In this project, deep learning methods are employed to efficiently capture complex structure–property relationships and make them usable for data-driven material development.

Using porous media as an example application, generative neural networks, in particular property-variational autoencoders (pVAE), are used to represent 3D microstructures in a latent space, modify them in a targeted manner, and optimize them with respect to desired properties. This enables the efficient generation of new material designs without the need to fully simulate every candidate structure.

Ongoing developments extend this approach to further classes of materials, particularly the inverse design of metal alloys. The overall goal is the development of tailored metamaterials for various engineering applications.

The project is carried out in cooperation with Prof. Aldakheel, LU Hannover.

Back