ETH trains artificial intelligence to design metamaterials

ETH researchers are working on the development of metamaterials using artificial intelligence, according to a statement from the university. These materials have optical, electrical or magnetic properties that do not occur in nature. They can be used in a wide range of applications that require special properties, from protective equipment such as bicycle helmets to medical technology and soft robotics.

The team led by Dennis Kochmann, Professor of Mechanics and Materials Research at the ETH Department of Mechanical and Process Engineering, has developed innovative tools based on artificial intelligence (AI) that bypass the time-consuming and intuitive design process for metamaterials. Instead, they predict structures with unusual properties quickly and automatically. In their latest work, the researchers are using AI to systematically explore the numerous designs and mechanical properties of two types of metamaterials.

One PhD student in the group trained an AI model based on a dataset of one million structures and their simulated deformation. In a study published recently, the team also showed that the AI model can go beyond what it was trained for and predict structures that are more powerful than anything previously generated.

Another doctoral student is using the method of video diffusion for his study. He trained his AI system with 50,000 video sequences of deforming 3D-printable metamaterial structures. “I can insert the trajectory of how I want the structures to deform, and the AI produces a video of the optimal structure and the complete deformation response,” he explained. ce/mm