Prediction of Simple Physical Properties of Mixed Solvent Systems by Artificial Neural Networks

Artificial neural networks (ANNs) are used to predict the density, viscosity and refractive index of several ternary and quaternary solvent systems based on training data from binary systems. These networks employed a relatively simple topology consisting of one hidden layer with three nodes and sin...

Full description

Saved in:
Bibliographic Details
Published in:Analytica chimica acta 1998-10, Vol.371 (2), p.117-130
Main Authors: Mehlman, S.A., Wentzell, P.D., McGuffin, V.L.
Format: Article
Language:English
Subjects:
Artificial neural networks
Chemistry
Density
Exact sciences and technology
General and physical chemistry
Mixed solvent properties
Refractive index
Solution properties
Solutions
Viscosity
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Artificial neural networks (ANNs) are used to predict the density, viscosity and refractive index of several ternary and quaternary solvent systems based on training data from binary systems. These networks employed a relatively simple topology consisting of one hidden layer with three nodes and single linear output node. The topology was optimized empirically using the water–methanol–acetonitrile–tetrahydrofuran system and applied to data for four other solvent systems obtained from the literature. The Bertrand–Acree–Burchfield (BAB) equation is used to predict the viscosity and refractive index for the same systems and the results are compared. The BAB equation and the ANNs performed comparably for most of the mixtures, but the BAB equation provided somewhat better predictions in a number of cases. The relative standard error of prediction using the ANNs was generally less than 1% for density and refractive index for all of the systems examined but ranges from 1% to 15% for the viscosity.
ISSN:0003-2670
1873-4324
DOI:10.1016/S0003-2670(98)00359-6