Premium
Comparative Study of Hard‐ and Soft‐Modeling Algorithms for Kinetic Data Processing
Author(s) -
Pérez Pla Francisco F.,
Baeza J.J. Baeza,
Llopis Elisa,
Baeza Mireia Pérez,
Fernández Lorenzo
Publication year - 2016
Publication title -
international journal of chemical kinetics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.341
H-Index - 68
eISSN - 1097-4601
pISSN - 0538-8066
DOI - 10.1002/kin.21004
Subject(s) - algorithm , least squares function approximation , chemistry , kinetic energy , reduction (mathematics) , experimental data , mathematics , computer science , statistics , physics , geometry , quantum mechanics , estimator
This article critically compares the efficacy of three algorithms, namely Alternating Least‐squares Multi Curve Resolution (ALS‐MCR), Hard Modeling Alternating Least‐squares (HM‐ALS), and classical Hard Modeling Multi Curve Resolution (HM‐MCR) in finding the true values of rate constants associated with a kinetic model. Simulated experiments on the simple system ( A 1 ⟶ A 2 ⟶ A 3 ) indicate that soft‐modeling ALS‐MRC methodology, which is subject only to linear constraints, does not ensure that experimental responses are correctly deconvolved, thus preventing further calculations to determine the true rate constants. Inclusion of the kinetic model in the ALS scheme, which gives rise to the HM‐ALS methodology, was found to yield a correct assessment of the rate coefficients but had a large computational cost. Numerical experiments employing a more complex model ( A 1 ⇌ A 2 ⇌ A 3 ) were also carried out, mainly to evaluate strategies for performing efficient searches on multidimensional multimodal least‐squares surfaces using HM‐ALS and HM‐MCR. This study again revealed the efficiency and reliability of classical HM‐MCR methods. Results from simulations were corroborated by analysis of data from an experimental study of chromate reduction by hydrogen peroxide; the mechanism of which is similar in complexity to those considered in simulations. The present work suggests that HM‐MCR algorithms implementing a multiminimum search strategy are the method of choice for analyzing two‐dimensional kinetic data.