ANN-MATOPT hybrid algorithm: determination of kinetic and non-kinetic parameters in different reaction mechanisms

Abstract

AbstractIn this work we have applied the computational methodology based on Artificial Neural Networks (ANN) to the kinetic study of distinct reaction mechanisms to determine different types of parameters. Moreover, the problems of ambiguity or equivalence are analyzed in the set of parameters to determine in different kinetic systems when these parameters are from different natures. The ambiguity in the set of parameters show the possibility of existence of two possible set of parameter values that fit the experimental data. The deterministic analysis is applied to know beforehand if this problem occurs when rate constants of the different stages of the mechanism and the molar absorption coefficients of the species participating in the reaction are obtained together. Through the deterministic analysis we will analyze if a system is identifiable (unique solution or finite number of solutions) or if it is non-identifiable if it possesses infinite solutions. The determination of parameters of different nature can also present problems due to the different magnitude order, so we must analyze in each case the necessity to apply a second method to improve the values obtained through ANN. If necessary, an optimization mathematical method for improving the values of the parameters obtained with ANN will be used. The complete process, ANN and mathematical optimizations constitutes a hybrid algorithm ANN-MATOPT. The procedure will be applied first for the treatment of synthetic data with the purpose of checking the applicability of the method and after, it will be used in the case of experimental kinetic data.