MAE Seminar: "Multifidelity Methods for Efficient Computational Analysis and Design of Aerospace Systems"

Dr. Leifur Thor Leifsson
Associate Professor
Purdue University

Thursday, December 2, 2021 | 2:00 – 3:00 PM

Virtual via Zoom

Abstract

Physics-based numerical simulations support many facets of engineering decision-making. Applications, such as optimal design and global sensitivity analysis, often exceed available resources when using only accurate high-fidelity models. Multifidelity methods, approaches to fuse information from models of varying degrees of fidelity, have been recognized as being important in alleviating the computational burden. This talk will present recent multifidelity methods for reduced-cost modeling, analysis, and optimal design. In particular, several approaches for constructing fast multifidelity models based on low-fidelity models, such as coarse-discretization simulations and simplified physics models, and novel adaptation techniques are presented. The methods will be demonstrated through several numerical examples, including aerodynamic shape optimization and global sensitivity analysis of nondestructive testing systems. Future research directions will be discussed.

Bio

Leifur Thor Leifsson is an Associate Professor with the School of Aeronautics and Astronautics at Purdue University, West Lafayette, Indiana. Prior to that, he was a faculty at Iowa State University, Ames, Iowa, and at Reykjavik University, Reykjavik, Iceland. Leifsson received his undergraduate and master’s degrees in Mechanical Engineering from the University of Iceland in 1999 and 2000, respectively, and a PhD degree in Aerospace Engineering from Virginia Tech in 2006. Leifsson’s research focuses on advancing fundamental computational methods for surrogate-based modeling and optimal design of engineered systems. Current applications include aerodynamic shape optimization, transonic flutter, dynamic stall, model-assisted nondestructive evaluation, and microwave antenna and filter design.