Researchers at Delft University of Technology (TU Delft) have developed a metamaterial that’s strong but supercompressible at the same time. Or rather: they had computers do it. Using machine learning, two different designs were made, one on a macro-scale for maximum compressibility and one on a micro-scale imparting strength and stiffness. Applied together on a brittle polymer, they transform into a lightweight, recoverable and supercompressible material.
Metamaterials derive special characteristics from repetitive structural patterns imparted on them. Finding the right pattern (or combination of patterns) for a particular set of characteristics is largely a trial-and-error process, requiring a lot of experimentation. “We argue in favor of inverting the process by using machine learning for exploring new design possibilities while reducing experimentation to an absolute minimum,” says Miguel Bessa, an assistant professor in materials science at TU Delft.
The essential requisite for successfully using AI to design metamaterials is, of course, data. There needs to be enough of it and it needs to be of sufficient quality.