Speaker: Prof. Andrew Christlieb, Michigan State University
Abstract: In this talk, I am going to start out very broad and define the multi-scale challenges associated with the goal of creating the mathematical tools that would enable optimal design of fusion energy systems. The twin challenges of the cursive dimensionality and the need for structure preserving representations will play a central theme. I will highlight work going on across the Center for Hierarchical and Robust Modeling of Non-Equilibrium Transport (CHaRMNET), a DoE MMICC center, targeted at addressing these issues. In the latter half of my talk I will introduce the development of blended computing. The goal in blended computing is the development of an augmented low fidelity model that produces high fidelity results at the cost of the low fidelity model. Here we are working in 1D with the BGK model of kinetic theory. In this context, we are developing structure preserving machine learning surrogates to close the Grad moment expansion with high fidelity kinetic data. Here, structure preserving means that the model maintains the necessary hyperbolic structure for long-time stability of the model, among other structure preserving properties. This framework was developed as part of CHaRMNET and is being transitioned to Los Alamos National Lab for reduced modeling of Fokker Planck descriptions of capsule implosion of inertial confinement fusion energy (IFE) systems.