Abstract—An artificial intelligence based on a genetic algorithm to build chemical reaction network (CRN) from chemical species concentration data from batch reaction is introduced. This is achieved through a two level optimization approach. The first level constructs the CRN through combinations of stoichiometric coefficients of all chemical species and optimized using genetic algorithm. Second level determines the best estimate for the reaction rate constants for each of the reactions using a standard non-linear optimization algorithm. The process is repeated through a number of generations where the genetic algorithm will successively reduce the number of possibilities through elimination of poor CRNs (based on how closely the CRN is able to predict concentration profiles) and retaining and re-optimizing better CRNs. This system’s capability is demonstrated on an experimental data for the reaction between trimethyl orthoacetate and allyl alcohol. The results show that the system is able to develop a CRN that when simulated provides an accurate model (model predictions matching experimental measurements) with little human intervention.
Index Terms—Reactor modeling, differential equations, system identification.
The authors are with the Newcastle University, Newcastle Upon Tyne, NE1 7RU United Kingdom (e-mail: charles.hii-jun-khiong@ncl.ac.uk, allen.wright@ncl.ac.uk, mark.willis@ncl.ac.uk).
[PDF]
Cite: Charles J. K. Hii, Allen R. Wright, and Mark J. Willis, "Utilizing a Genetic Algorithm to Elucidate Chemical Reaction Networks: An Experimental Case Study," International Journal of Chemical Engineering and Applications vol. 5, no. 6, pp. 516-520, 2014.
