The reconstruction of thermo-chemical scalars in combustion from a reduced set of their principal components

Abstract

We compare two reconstruction approaches for thermo-chemical scalars (TCSs) in turbulent combustion using principal component analysis. The first approach is based on the inversion of the linear relation between the TCSs and their principal components (PCs). The second is based on a regression of TCSs with a reduced set of the PCs using artificial neural networks. The study is based on one-dimensional turbulence simulation data of Sandia Flame F. We find that regression potentially offers superior reconstruction to the inversion expression when a truncated set of the original PCs is used.