|
|  |
|
|
Wednesday, October 27, 202112:00 PM - 1:00 PMWebEx
Postdoc Seminar
Model order reduction for nonlinear radiative transfer based on moment equations and data-driven approximations of the Eddington tensor
Joseph CoaleNorth Carolina State University
||FW: Talk: Model order reduction for nonlinear radiative transfer based on moment equations and data-driven approximations of the Eddington tensorBrady, Peter TThu 10/21, 9:22 AMDuback, Diane MBrady, Peter T;Duback, Diane M;Livescu, Daniel;Garcia, Angel Enrique;mass-app2022@lanl.gov;ccs-2@lanl.gov;Joe Coale WebexHi Diane,This is for a shared postdoc position with CLNS. Please advertise on the CLNS distribution list. His company is North Carolina State University and is being hosted by CCS-2.Thanks,PeterFrom: on behalf of "Brady, Peter T" Date: Thursday, October 21, 2021 at 9:20 AMTo: "Livescu, Daniel" , "Garcia, Angel Enrique" , "mass-app2022@lanl.gov" , "ccs-2@lanl.gov" , Joe Coale Subject: Talk: Model order reduction for nonlinear radiative transfer based on moment equations and data-driven approximations of the Eddington tensorPlease join us for a seminar by Joseph Coale on Model order reduction for nonlinear radiative transfer based on moment equations and data-driven approximations of the Eddington tensorAbstract:In this talk a new group of reduced-order models (ROMs) for multigroup thermal radiative transfer (TRT) problems is presented. These ROMs are formulated in a way that preserves fundamental physical properties of the TRT solution. This is achieved with a combined approach using both nonlinear projective techniques and data-driven methods of approximation. The Boltzmann transport equation is projected onto a series of subspaces to derive a hierarchy of moment equations that are closed by nonlinear functionals with dependence on the Boltzmann solution. The Eddington tensor that provides exact closure for the moment equations is approximated by the proper orthogonal decomposition and dynamic mode decomposition in the entire phase space and time. The resulting model has the same dimensionality as the multigroup P_1 model for TRT. The performance of the developed ROMs is demonstrated on the Fleck-Cummings test problem in 2D Cartesian geometry. A significant reduction in dimensionality is achieved compared to the full-order model. The proposed approach to model reduction can be extended to a wide class of multi-physical problems of high-energy density physics (e.g. radiation hydrodynamics).
Host: Peter Brady
|
