Presented By: Aerospace Engineering
Jerry Faeth Memorial Lecture | Integrated Simulation of Composite Scramjets, Jonathan Freund, UIUC
Jonathan Freund
Willett Professor and Head
Aerospace Engineering
University of Illinois Urbana-Champaign
The talk will overview the integrated predictive science efforts in the Illinois PSAAP3 Center for Exascale-enabled Scramjet Design, which targets the analysis and design of composite scramjet engines for hypersonic propulsion. This will include the experimental prediction target, the physical models and their corresponding physics-targeted experiments, the numerical techniques, software strategies, the computer science methodology in which the simulation tools are being developed, and the center management approach. This high-level overview will include deeper dives into the technical details of two subprojects. In the first, the adjoint governing equations are used in conjunction with a neural network back-propagation algorithm to train sub-grid-scale models for turbulence. Models trained with such a "full" physical constraint are confirmed to be both accurate and robust in simulations of canonical turbulent flows. The other deep dive will present a multi-point adjoint-based optimization method that can skirt the effects of chaos that typically impedes gradient-based search in order to find effective controls for the relatively deterministic components of the turbulence. To do this discontinuity degrees of freedom are introduced into the optimization problem, with spacing in time comparably to the chaotic time scale, and then iteratively penalized. Demonstrations on model problems, including a three-dimensional Kolmogorov turbulence control configuration, show that the method finds minima that are far better than those found by standard gradient search.
About the speaker...
Jonathan Freund is the Donald Biggar Willett Professor and Head of Aerospace Engineering at the University of Illinois at Urbana--Champaign. He is a Fellow of the American Physical Society, and a winner of the 2008 Frenkiel Prize from its Division of Fluid Dynamics and has served in division leadership in several capacities. He is on the editorial board of Annual Review of Fluid Mechanics. Computational science has been central to his research across and beyond fluid mechanics, which has included simulations of turbulent jet noise and its control, the dynamics of molecularly thin liquid films, nanostructure formation by ion-bombardment of semiconductor materials, bubble dynamics in confinement, and the dynamics of red blood cells flowing in the narrow confines of the microcirculation. He is the PI and co-director of the Center for Exascale-enabled Scramjet Design (CEESD).
https://umich.zoom.us/j/93096785925; Passcode: AE585
Willett Professor and Head
Aerospace Engineering
University of Illinois Urbana-Champaign
The talk will overview the integrated predictive science efforts in the Illinois PSAAP3 Center for Exascale-enabled Scramjet Design, which targets the analysis and design of composite scramjet engines for hypersonic propulsion. This will include the experimental prediction target, the physical models and their corresponding physics-targeted experiments, the numerical techniques, software strategies, the computer science methodology in which the simulation tools are being developed, and the center management approach. This high-level overview will include deeper dives into the technical details of two subprojects. In the first, the adjoint governing equations are used in conjunction with a neural network back-propagation algorithm to train sub-grid-scale models for turbulence. Models trained with such a "full" physical constraint are confirmed to be both accurate and robust in simulations of canonical turbulent flows. The other deep dive will present a multi-point adjoint-based optimization method that can skirt the effects of chaos that typically impedes gradient-based search in order to find effective controls for the relatively deterministic components of the turbulence. To do this discontinuity degrees of freedom are introduced into the optimization problem, with spacing in time comparably to the chaotic time scale, and then iteratively penalized. Demonstrations on model problems, including a three-dimensional Kolmogorov turbulence control configuration, show that the method finds minima that are far better than those found by standard gradient search.
About the speaker...
Jonathan Freund is the Donald Biggar Willett Professor and Head of Aerospace Engineering at the University of Illinois at Urbana--Champaign. He is a Fellow of the American Physical Society, and a winner of the 2008 Frenkiel Prize from its Division of Fluid Dynamics and has served in division leadership in several capacities. He is on the editorial board of Annual Review of Fluid Mechanics. Computational science has been central to his research across and beyond fluid mechanics, which has included simulations of turbulent jet noise and its control, the dynamics of molecularly thin liquid films, nanostructure formation by ion-bombardment of semiconductor materials, bubble dynamics in confinement, and the dynamics of red blood cells flowing in the narrow confines of the microcirculation. He is the PI and co-director of the Center for Exascale-enabled Scramjet Design (CEESD).
https://umich.zoom.us/j/93096785925; Passcode: AE585
Explore Similar Events
-
Loading Similar Events...
