Laboratory reactor design and the precision of parameter estimates.

Abstract

This study is concerned with investigating the dependence of the precision of estimated kinetic parameters on the type of reactor used for performing the kinetic measurements. Two ideal reactors, the plug-flow reactor (PFR) and the continuous-stirred-tank reactor (CSTR), were simulated using a Monte-Carlo computer simulation. Parameters were estimated using nonlinear multiresponse estimation techniques, and the distributional characteristics of the parameter estimates were calculated. Comparison between the reactors involved the study of overall measures of precision such as the size, shape and orientation of the 95% joint confidence region, and the determinant of the covariance matrix of the parameter estimates. Five variables were identified as having a possible affect on the precision: the nature of the reaction network, the kinetic model, the magnitudes of the rate parameters, the covariance structure of the responses, and the experimental design. The dependence of parameter precision on these variables is presented along with recommendations for determining the reactor type to give the most precise kinetic parameters.