Speaker
Alessandro Gabbana
Description
In this work we explore the possibility of learning a collisional operator
for the Lattice Boltzmann Method from data using a deep learning approach.
We present results where a Neural Network is successfully trained as a surrogate of the single relaxation time BGK operator.
We show that only by embedding in the Neural Network physical properties such as
conservation laws and symmetries, it is possible to correctly reproduce
the short and long time dynamics of standard fluid flows.
Primary authors
Alessandro Gabbana
Alessandro Corbetta
(Eindhoven University of Technology)
Vitaliy Gyrya
(Los Alamos National Laboratory)
Daniel Livescu
(Los Alamos National Laboratory)
Joost Prins
(Eindhoven University of Technology)
Federico Toschi
(Eindhoven University of Technology)
