On the Use of Low-Dimensional Simulations for Database Generation in PCA-Based Modeling

Abstract The use of principal component analysis (PCA) for turbulent combustion modeling has been the focus of multiple recent research efforts. PCA has been shown to be effective as a dimension reduction technique. PCA examines a database to determine directions of maximum variance. The compositions in the full state space can then be projected onto a reduced state space spanned by a subset of these directions leading to dimension reduction. This significantly reduces the number of transport equations that need to be solved at runtime leading to significant computational savings. One of the crucial steps when using PCA in reacting flow modeling is database generation. Specifically, the database needs to be generated in a computationally-efficient manner and needs to be comprised of compositions that are representative of those encountered at runtime. Multiple low-dimensional configurations have been explored in the recent literature, with one of the most popular choices being one-dimensional counterflow flames. In the current work, we perform a-priori comparisons between compositions extracted from a variety of different turbulent combustion configurations and databases generated using corresponding one-dimensional counterflow flame (CF) configurations. These a-priori comparisons are used to assess the suitability of counterflow flame configuration for database generation in the context of PCA-based modeling.