On the use of Bayesian approach, experimental data, and artificially designed data, for the identification of missing reactions

Abstract

We propose and analyze the use of a Bayesian approach for the investigation of missing reactions, considering both modeling and experimental uncertainties. The main idea is the use of two calibration data sets: one is experimental and the other is artificially designed in such a way that features observed in the experimental data set are magnified. We apply our proposed methodology on the investigation of a combustion kinetics phenomena that is modeled with a highly nonlinear system of several ordinary differential equations. More specifically, we quantify uncertainties in the reduced chemistry (6 reactions) model of the HCN/O2/Ar mixture kinetics proposed by Thielen and Roth,1 and try to identify which reaction(s) should be added to the reduced chemistry in order to improve its predictions of species concentration profiles at high temperature scenarios. Our data analysis methodology using both experimental and artificially designed data indeed helps us identify some critical reactions out of the pool of 27 extra reactions. Due to its simplicity, the approach can be potentially applied to a wide variety of engineering problems. Copyright © 2011 by Kenji Miki.