Development of a Reaction Structure Identification Algorithm

Abstract

The most important step in the design procedure of a chemical reactor is the understanding the chemical reaction network (CRN), which will take place in that reactor. The structure of a CRN as representation of the reaction mechanism contains all the elementary reaction steps that are required to convert the reagents into products. The aim of the reaction mechanism analysis is the identification of the route how the system goes from its initial to the end state. In order to do this, a lot of knowledge is required about chemistry supplemented with some analytical measurements. In this work, we focus on the development of an algorithm, which requires a few data inputs to reveal all the reaction steps that are important in the reactor design point of view. It is trivial that the structure of a dynamic system cannot be determined without the identification of the model parameters that belong to that structure. Hence, the algorithm reported here can be used to obtain the parameters of the reaction rate equations for each identified chemical reaction. First, the structure is identified followed by its parameters. In this study the processed data are obtained by a CRN generator, which is applied to generate random CRNs based on some specified parameters to reach maximal reaction order. Concentration profiles, which characterize each CRN at a specific reaction rate constants combination, are obtained as a result of simulations based on the randomly generated CRNs. The developed algorithm for reaction mechanism identification is based on a modified type of linear least-squares method (LLSM) in which the searching variables must be higher than zero, since the reaction rate constants cannot have negative values. The developed algorithm is tested in different cases to check the applicability of LLSM in reaction structure identification procedure and the obtained results show that with some further improvements the proposed algorithm can be applied solving more complex identification tasks.