On the integration of model identification and process optimization

Abstract

Reliable models for rate-based phenomena are the backbone of model-based process design. These models are often unknown in the early design phase and need to be determined from laboratory experiments. Although model-based experimental analysis and process design are often executed sequentially, the kinetic models might not be suitable to reliably design a process. In this paper, we address this problem and present a first step on the integration of model identification and process optimization. Rather than decoupling model identification and process optimization, we use information from process optimization to design optimal experiments for improving the quality of the kinetic model given the intended use of the model. Sensitivities, which describe the influence of parametric uncertainties on the economic objective used in process optimization, are used as weights for optimal experimental design. This way, the confidence in the parameter values is maximized to reduce their influence on the process optimization objective. This first step on the integration of model identification and process optimization improves the predictive quality of a reaction kinetic model for process design without any further experimental effort.