Identifying Chemical Reaction Network Models

In this work, an automated chemical reaction network identification procedure using a genetic algorithm (GA) is introduced. The GA uses chemical species concentration data obtained from batch reactors during process experimentation to build ordinary differential equation (ODE) models that represent the chemical reactions occurring. This is achieved using a two-tiered optimization approach. The main tier is an integer optimization problem using GA where the stoichiometric coefficients of the network of chemical reactions are determined. Using these stoichiometric coefficients, the specific rate constants for each reaction are obtained by solving a non-linear optimization problem in the second tier of the proposed approach. The prediction accuracy of any potential reaction network is determined by comparing the results obtained from the ODE model generated by the GA and the measured values obtained from experimentation. More promising models are retained by the GA and are used to construct even better networks in subsequent steps (or generations) of the GA through its evolution process. After a number of generations, the GA is terminated and the best network (in terms of prediction capability) is extracted. Using simulated data, the proposed optimization procedure is demonstrated to be capable of accurately determining a chemical reaction network.