Machine Learning for Chemical Reaction Dynamics in High Energy, Rarefied Gas Flow.
Abstract
The present proposal employs state-of-the-art machine learning, quantum chemistry and molecular dynamics simulations to calculate the cross sections and reaction rate coefficients for atom+diatom and diatom+diatom gas phase reactions relevant to hypersonics, applies the knowledge to modeling non-equilibrium chemical processes for a typical composition of reactive air flow consisting of N, O, N2, O2, and NO species, and provides direct input for more coarse grained approaches such as direct simulation Monte Carlo, computational fluid dynamics and chemical kinetics codes for reactive, high-energy, rarefied flow. To this end, new neural network-based state-to-distribution (STD) models for mapping initial to final states based on machine learning techniques will be developed and applied and existing techniques, such as reproducing kernel Hilbert space representations, will be extended to larger molecular systems.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 04, 2023
- Source ID
- FA86552117048
Entities
People
- Markus Meuwly
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Basel