Principal component analysis of kinetic models

Abstract

An eigenvalue–eigenvector analysis is used to extract meaningful kinetic information from linear sensitivity coefficients computed for several species of a reacting system at several time points. The main advantage of this method lies in its ability to reveal those parts of the mechanism which consist of strongly interacting reactions, and to indicate their importance within the mechanism. Results can be used to solve three general kinetic problems. Firstly, an objective condition for constructing a minimal reaction set is presented. Secondly, the uncovered dependencies among the parameters are shown to confirm or deny validity of quasi‐steady‐state assumptions under the considered experimental conditions. Thirdly, taking into account only sensitivities of observed species, the analysis is used to yield error estimates on unknown parameters determined from the experimental observations, and to suggest the parameters that should be kept fixed in the estimation procedure. To illustrate we chose the well‐known hydrogen–bromine reaction and the kinetics of formaldehyde oxidation in the presence of CO.