Structure Preserving Neural Network Based Entropy Closures for the Boltzmann Moment

Steffen Schotthöfer

Abstract: In numerous applications, macroscopic simulations do not yield sufficient accuracy, thus physical modelling at a kinetic level is required. We consider the moment methods as a numerical solution method for the Boltzmann equation, a famous example for a kinetic equation. The minimal entropy closure is a closure for the Boltzmann moment system with desirable structural properties, that come at the expense of high computational cost. This talk presents neural network based acceleration strategies for the minimal entropy closure for the Boltzmann moment system.

Speaker’s Bio: Steffen Schotthöfer studied Mathematics and Computer Science at the Technical University of Kaiserslautern, Germany. He is currently enrolled in a PhD program at Karlsruhe Institute of Technology- Karlsruhe, Germany, under supervision of Prof. Martin Frank, where he works on embedding neural networks in numerical methods for kinetic equations.