Using sensitivity entropy in experimental design for uncertainty minimization of combustion kinetic models

Due to the inherent uncertainties in combustion kinetic model parameters, especially the rate coefficients of elementary reactions, the uncertainties of model predictions can be quite large. Uncertainty minimization using experimental measurements can reduce the uncertainty space of the rate coefficients of elementary reactions, and further reduce the uncertainties of model predictions. Many mathematical methods have been developed for this purpose, while little research has been done to guide us in designing experiments which are relatively efficient for uncertainty minimization. In this work, "sensitivity entropy" is proposed as a measure of the degree of dispersion of uncertainty sources of a model output. The smaller the sensitivity entropy is, the lower degree of dispersion of uncertainty sources will be. The experimental measurement of a target which has smaller sensitivity entropy will be more efficient for the uncertainty minimization. To illustrate the practicability of using sensitivity entropy to guide the selection of relatively efficient experimental systems for specific targets for model uncertainty minimization, the methanol/O₂/Ar laminar premixed flame system is investigated. Based on the analysis of sensitivity entropy, two experiments in which many targets have small sensitivity entropies are designed. Results show that these well designed experiments can provide strong constraints on the uncertainty space of some rate coefficients, which are expected to be useful in further kinetic model development.