๐๐๐ฒ๐ฟ ๐๐ผ๐ป๐ฑ๐ฒ๐ฟ๐ฒ๐ฑ ๐ต๐ผ๐ ๐๐ผ ๐บ๐ฎ๐ธ๐ฒ ๐ฎ ๐บ๐ฎ๐๐ฒ๐ฟ๐ถ๐ฎ๐น ๐๐ต๐ฎ๐ ๐ฐ๐ฎ๐ป ๐๐บ๐ถ๐น๐ฒ ๐ผ๐ฟ ๐ณ๐ฟ๐ผ๐๐ป?
Together with my PhD advisors, Marjolein Dijkstra, Martin van Hecke & Corentin Coulais, we set out to answer this seemingly silly question and, in the process, developed a two-step approach to design mechanical metamaterials with multiple, desired deformation pathways.
๐ ๐ง๐ต๐ฒ ๐ธ๐ฒ๐ ๐ถ๐ป๐๐ถ๐ด๐ต๐
Instead of directly searching the vast combinatorial design space for the perfect material, we split the design process in two: first, generate designs with rich, general mechanical responses; second, combine and fine-tune to achieve the desired response. Specifically, our method combines neural networks and a genetic algorithm to prospect for designs with a high ๐ฑ๐ญ๐ถ๐ณ๐ช๐ฑ๐ฐ๐ต๐ฆ๐ฏ๐ค๐บโthe potential to exhibit multiple, spatially extended deformationsโand refine those into a final material with the desired deformations: like a smiley and frowny.
I am happy to share that we published this work in American Physical Society’s Physical Review Research, where it is featured as an editors’ suggestion!
๐ You can check out the paper here.

Leave a comment