Abstract
Sensitivity analysis and gradient-based optimization are useful augmentative tools for high-fidelity predictive simulations. However, their use in chaotic systems is challenging because of the ill-defined sensitivity associated with chaos. A trivial example of Lorenz system will be used to introduce this. Then we will consider the challenges of particle-in-cell (PIC) simulation models for plasma kinetics and turbulence. Alternative routes to use sensitivity for chaotic systems are presented. For PIC, a particle-pdf approach is developed to flexibly commute between Lagrangian and Eulerian frameworks. It is illustrated for the Debye shielding response of plasmas, and an adaptive extension is demonstrated for a transient in a plasma sheath. For turbulent flows, the consequence of chaos on gradient-based optimization is quantified and analyzed for the far-field sound control of a turbulent jet. A penalty-based multiple-shooting optimization method is proposed and then demonstrated on canonical chaotic dynamics. Progress toward application to turbulence is discussed.
Bio
SeungWhan Chung is a Ph.D. candidate in Theoretical and Applied Mechanics at the University of Illinois at Urbana-Champaign. He received his Bachelor of Science in Mechanical and Aerospace Engineering from Seoul National University, Korea. He has been working with Prof. Freund since 2014 at the Center for Exascale Simulation of Plasma-Coupled Combustion (XPACC) and collaborating with Dr. Eric C. Cyr and Dr. Stephen D. Bond from Sandia National Laboratory. His research topics include adjoint-based sensitivity analysis and gradient-based optimization for chaotic dynamical systems, particle-in-cell plasma-kinetics simulations, far-field sound control of turbulent jets, and exact-coherent structures of turbulent channel flows.
Host: Professor Jon Freund
