Current Awardee

Design of Multiphysics Structures using Generative Adversarial Networks

Kai James
Kai James

College: Grainger College of Engineering
Award year: 2021-2022
NCSA collaborators: Diab Abueidda, Erman Guleryuz

Until recently, machine learning was used almost exclusively for analytical tasks such as classification and regression. However, with the advent of generative adversarial networks, machine learning tools can now autonomously perform open-ended creative tasks such as generating original artwork or writing poetry. More recently, researchers have begun to apply this technology to structural design optimization, however existing efforts in this domain have been limited to relatively simple single-discipline problems. This project will introduce a series of novel network architectures for generating complex structures while accounting for multiple performance criteria from different physics disciplines. The outcome of this effort will be the creation of a generative design algorithm that can rapidly produce multifunctional structures that more closely resemble the engineering structures used in real-world industry applications. By applying this design approach to real-word design problems, particularly in the area of aircraft and vehicle design, the team will contribute to the development of better performing systems that are more lightweight and that generate reduced emissions, thereby contributing to more sustainable vehicles.