화학공학소재연구정보센터
Computers & Chemical Engineering, Vol.121, 618-632, 2019
Computational approaches to kinetic model selection
This paper demonstrates how the stoichiometry and kinetic model of a chemical synthesis involving multiple reactions can be selected via a computational approach which uses consecutive optimisation steps. First, a list of all feasible stoichiometric relations consistent with the molecular weights or the elemental makeup of participating species is developed using integer linear programming ( ILP). A second ILP is then used to construct all plausible stoichiometric schemata ( combinations of the stoichiometric equations) which are used to instantiate kinetic model structures. Using a numerical integration routine, the models are simulated and unknown parameters estimated using an iterative optimisation algorithm. Produced model structures are then numerically scored, ranked and compared. This allows selection between competing models using both physical and the statistical evidence the data provides. The methods are demonstrated using synthetic and experimental data sets assuming liquid-phase reactions occurring in a well-mixed isothermally operated batch reactor. (C) 2018 Elsevier Ltd. All rights reserved.