Optimizing progress variables for ammonia/hydrogen combustion using encoding-decoding networks

We demonstrate a strategy to optimize low-dimensional parameterizations of turbulent flames using an encoding-decoding artificial neural network architecture. A gradient descent optimizer is informed by the reconstruction quality of important quantities of interest (QoIs) that enter the optimization as the decoder outputs. Our focus is on the combustion of ammonia/hydrogen blends. The literature on ammonia combustion to date lacks an efficient definition of a reaction progress variable (PV) to parameterize the thermo-chemical state-space. A quality parameterization should be able to represent the thermo-chemical state variables accurately, as well as any functions of those, e.g., the source terms of the non-conserved PVs. Our approach incorporates information about the reaction source term of a PV and about important combustion products into the PV optimization. This approach naturally promotes parameterizations where a QoI is uniquely and smoothly represented over the manifold. We show that with an adequate definition of a PV, we can steer the model's accuracy towards improved representation of selected products and pollutants. The definition of a PV automatically adapts to best complement the remaining physics-based parameters, such as the mixture fraction or the enthalpy defect.