Time: | January 15, 2025, 4:00 p.m. – 5:00 p.m. |
---|---|
Venue: | Hörsaal V7.01 Pfaffenwaldring 7 70569 Stuttgart |
Download as iCal: |
|
Multiscale material modeling has two primary goals: the forward problem (i) aims at understanding and predicting a material’s properties based on its small-scale architecture, while the inverse problem (ii) seeks to identify those small-scale structural features that enable us to control and optimize a material’s properties and performance. Owing to the rise of additive manufacturing, architected materials (or metamaterials) have emerged as a special class of man-made materials with exciting, peculiar, or tunable properties, and as a new playground for computational modeling. While the forward problem (i) has been tackled by many successful modeling techniques across scales, the inverse problem (ii) has remained a challenge: how do we design (meta-)materials with target properties? At the core of this challenge are the abundant design and property spaces, as well as the fact that the map from structure to properties is not invertible. In this seminar, we will discuss how machine learning has recently offered new opportunities for the inverse design of architected materials through generative models that predict metamaterial building blocks with extreme, peculiar, or general target mechanical properties.