| The study of mixtures is central to the science of chemistry, and this work examines the application of mathematical modeling techniques to two mixture problems of practical interest. In both cases, nonlinear optimization methods are used to address the difficult task of developing mathematical descriptions of the systems under study.;In the first application, artificial neural networks (ANNs) are used to predict some simple physical properties of binary, ternary and quaternary solvent mixtures. In this work, ANNs have been used to model the non-ideal behavior of the density, refractive index, and viscosity in several solvent mixtures, including water/methanol/acetonitrile/tetrahydrofuran, mixtures of hydrocarbons and mixtures of alcohols.;In a second application, the problem of environmental receptor modeling was investigated. The objective of receptor modeling is to identify the sources of pollutants at a receptor site and assess the magnitude of their individual contributions. One way to do this is to look at the distribution of chemical constituents (e.g. elements) in potential sources, known as emission profiles, and compare those to the distribution at the receptor, known as the receptor profile. (Abstract shortened by UMI.). |
