What if we could discover entirely new material properties beyond what human intuition can design? Mechanical metamaterials derive their unusual behaviour from carefully engineered geometries rather than chemical composition alone.
Traditionally, researchers have designed these structures using their understanding of mechanical principles. We explore how computational methods and machine learning can uncover hidden relationships between structure and functionality, opening new possibilities for complex, multifunctional materials.
How we approach the challenge
We develop computational tools to understand how geometry and mechanical interactions determine material behaviour, helping us explore designs beyond conventional intuition.
By learning complex structure–property relationships from data, machine-learning models can identify design rules and guide the discovery of new mechanical functionalities.
Our research demonstrates how neural networks can learn intricate combinatorial rules governing mechanical metamaterials, even from limited training data, offering new routes to material design.
Explore the research
PHYSICAL REVIEW LETTERS · 2022 Machine Learning of Implicit Combinatorial Rules in Mechanical Metamaterials Ryan van Mastrigt, Marjolein Dijkstra, Martin van Hecke & Corentin Coulais Read publication ↗David Dykstra