Identifying kinetic models from reactor measurements

Abstract

This paper investigates parameter estimation of rate constants appearing in chemical mechanisms. As most chemical mechanisms are more complex than can be supported by available laboratory measurements, model reduction is a required first step. The quasisteady-state assumption and the reaction equilibrium assumption are presented as the two main model reduction methods. Reliable quantification of the approximate confidence intervals of the estimated parameters is a second key step. A brief overview of current numerical software for this purpose is provided. Parameter estimation with a starting mechanism and typically available simulated laboratory measurements is then applied to three illustrative example systems: (i) an electrochemical oxygen reduction reaction, (ii) butene isomerization by a metathesis mechanism, and (iii) enzymatic kinetics taking place in bacterial microcompartments. The reliability of the model reduction techniques and the current computational software is assessed based on the outcomes of these three example chemistries. In all three examples, the quasisteady-state assumption was required to remove some large rate constants governing low concentration, highly reactive species that could not be measured. After the model reduction, the parameter confidence intervals were then used to determine what extra measurements were required to identify the model, or the model was reparameterized to obtain an identifiable reduced set of parameters for the given measurements.