Physics-Aware Neural Network Flame Closure for Combustion Instability Modeling in a Single-Injector Engine

Abstract: Neural networks (NN) are implemented as sub-grid flame models in a large-eddy simulation of a single-injector liquid-propellant rocket engine with the aim to replace a look-up table approach. The NN training process presents an extraordinary challenge. The multi-dimensional combustion instability problem involves multi-scale lengths and characteristic times in an unsteady flow problem with nonlinear acoustics, addressing both transient and dynamic-equilibrium behaviors, superimposed on a turbulent reacting flow with very narrow, moving flame regions. Accurate interpolation between the points of the training data becomes vital. A major novel aspect of the proposed NNs is that they are trained to reproduce relevant portions of the information stored in a flamelet table by using only limited data from a few CFD simulations of a single-injector liquid-propellant rocket engine under different dynamical configurations. This is made possible by enriching the training set with contrived data resulting from the physical characteristics of the combustion model and also by including the flame temperature as an extra input to the NNs that are trained to model other flame variables of interest. These physics-aware NN-based closure models are first tested offline by comparing them directly with the flamelet table and then are successfully implemented into CFD simulations in place of the flamelet table and verified on various dynamical configurations. The results from those tests compare favorably with counterpart table-based CFD simulations. Computational advantages of the approach are discussed.