MAE 298 Seminar: Machine Learning Acceleration of Turbulent Combustion and Nonequilibrium Flow Predictions

Abstract: Accurately predicting complex flows in engineering systems remains a significant challenge. Computationally affordable predictions generally require Reynolds-Averaged Navier-Stokes (RANS) simulations or Large-Eddy Simulation (LES), which can be insufficiently accurate in certain regimes due to the nonlinear coupling of unresolved physical processes. For example, in turbulent combustion, flame–turbulence interactions can lead to inverse-cascade energy transfer, which violates the assumptions of many RANS and LES closures. In hypersonic flows, the standard viscous stress, heat flux and no-slip boundary condition may be called into question at subcontinuum Knudsen numbers. We present an adjoint-based, solver-embedded data assimilation method to augment the Navier-Stokes equations with neural network terms. These are optimized in-situ using high-performance, Python-native flow solvers that leverage automatic differentiation programming techniques to construct the adjoint equations needed to optimize the networks. We present applications to RANS and LES of turbulent premixed jet flames, continuum predictions of nonequilibrium hypersonic flows, and chemical-kinetic model reduction for ignition delay and detonation predictions. We also discuss potential future applications of adjoint-based machine learning to the prediction and control of complex flows.

Bio: Jonathan MacArt is an assistant professor of aerospace and mechanical engineering at the University of Notre Dame. His research centers on computational simulation and modeling of turbulent, reacting and hypersonic flows, with emphasis on solver-embedded optimization and novel HPC techniques. He leads several NSF and DOD sponsored research projects including an NSF CAREER award. Prior to joining Notre Dame, he held a postdoctoral appointment in the Center for Exascale Simulation of Plasma-coupled Combustion at the University of Illinois Urbana-Champaign. He earned his Ph.D. in mechanical and aerospace engineering from Princeton University in 2018.