Open AccessComparative analysis of numerical methods for determining parameters of chemical reactions from experimental data
Author(s) -
Aleksei Prikhodko,
Maxim Shishlenin,
Olga A. Stadnichenko
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2092/1/012011
Subject(s) - identifiability , inverse problem , mathematical optimization , ode , simulated annealing , gradient method , particle swarm optimization , mathematics , computer science , algorithm , mathematical analysis , machine learning
The aim of this paper is to select an optimal numerical method for determining the parameters of chemical reactions. The importance of the topic is due to the modern needs of industry, such as the improvement of chemical reactors and oil or gas processing. The paper deals with the problem of determining reaction rate constants using gradient methods and stochastic optimization algorithms. To solve an forward problem, implicit methods for solving stiff ODE systems are used. A correlation method of practical identifiability of the required parameters is used. The genetic algorithm, particle swarm method, and fast annealing method are implemented to solve an inverse problem. The gradient method for the solution of the inverse problem is implemented, and a formula for gradient of the functional is given with the corresponding adjoint problem. We apply an identifiability analysis of the unknown coefficients and arrange the coefficients in the order of their identifiability. We show that the best approach is to apply global optimization methods to find the interval of global solution and after that we refine inverse problem solution using gradient approach.