Lab on AI-based Metamaterials & Metastructures Design

Objectives
The design of high-performance metamaterials and metastructures is a highly challenging problem from both the methodological and computational viewpoints due to the intrinsic complexity inherited from their multi-scale nature. Indeed, it often involves thousands of degrees-of-freedoms even for moderate-size devices composed by rather “simple” unit cells (UCs). Artificial Intelligence (AI) is a powerful tool to develop accurate and efficient surrogate models for
predicting the response of a system as a function of both its material/geometric descriptors and the external excitation. This forms the foundation for developing efficient “digital twins” (DTs). The laboratory aims at providing the methodological skills and knowledge on efficient AI-driven design of metamaterials and metastructures through computer-guided examples.

Topics
Fundamentals of AI and Machine Learning (ML) algorithms
ML as a "three-steps" process for building accurate DTs
Dimensionality reduction and single-shot/adaptive sampling techniques
Surrogate model for computational expensive metastructures and metamaterials
System-by-Design (SbD) framework for the computationally-efficient design of complex multi-scale structures
Applicative examples of AI-based metamaterials and metastructures designs

Required Skills
Course of “Foundations of Electromagnetic Waves” and of “Foundations of Mechanics and Acoustic”.

Course Modality and Verification of Learning
The classes will be given in hybrid (i.e., onsite and online) format. The teaching activity includes SW/HW emulator exercises.
The exam consists in the development and presentation of a project assignment.