Classical Potential Barrier of Liquid Phase Radical Reactions and its Simulation Based on the Experimental Kinetic and Thermochemical Data Using Fuzzy Neural Networks

Abstract

This paper discusses the technology of extracting the chemical knowledge from the structured electronic collections using fuzzy neural networks. It proposes an approach to the formation of the information space of features and the reduction of their dimensionality on the basis of empirical models of radical reactions, the logical (and physical) structure of electronic collections of experimental data is presented on the basis of data warehousing technology for sampling for the purpose of scientific data analysis. The procedure for approximating the classical potential barrier of a bimolecular radical reaction in the liquid phase from experimental data using fuzzy neurons networks is briefly described. An analysis of the use of empirical models for elementary radical abstraction reactions as constraints for the proposed approach is given. An example is shown of constructing a fragment of a fuzzy knowledge base. The structure of the data warehouse on the rate constants of radical reactions in the liquid phase and the kiosk of data on the bond dissociation energies is given for the first time. Using a fuzzy knowledge base built by experts, and Mamdani's fuzzy inference method using triangular membership functions produces a good approximation of the values of the classical potential barrier. The proposed technology can serve as a basis for developing the fuzzy multi-agent expert systems in the domain chosen by the authors.