Solubility Parameters of Nonelectrolyte Organic Compounds: Determination Using Quantitative Structure-Property Relationship Strategy
| Autor: | Ali Eslamimanesh, Farhad Farjood, Farhad Gharagheizi, Amir H. Mohammadi, Dominique Richon |
|---|---|
| Přispěvatelé: | Saman Energy Giti Co., CEP/Fontainebleau, Centre Énergétique et Procédés (CEP), MINES ParisTech - École nationale supérieure des mines de Paris, Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL)-MINES ParisTech - École nationale supérieure des mines de Paris, Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL), School of chemical engineering, University of Birmingham [Birmingham] |
| Jazyk: | angličtina |
| Rok vydání: | 2011 |
| Předmět: |
Quantitative structure–activity relationship
Chromatography Artificial neural network Chemistry General Chemical Engineering 02 engineering and technology General Chemistry 021001 nanoscience & nanotechnology Industrial and Manufacturing Engineering Support vector machine Nonlinear system Hildebrand solubility parameter 020401 chemical engineering Molecular descriptor Genetic algorithm [SPI.GPROC]Engineering Sciences [physics]/Chemical and Process Engineering 0204 chemical engineering Solubility 0210 nano-technology Biological system |
| Zdroj: | Industrial and engineering chemistry research Industrial and engineering chemistry research, American Chemical Society, 2011, 50 (19), pp.11382-11395. ⟨10.1021/ie200962w⟩ |
| ISSN: | 0888-5885 1520-5045 |
| Popis: | International audience; The solubility parameter is considered to be a significant parameter for the chemical industry. In this study, the quantitative structure -property relationship (QSPR) method is applied to develop three models for determination of the solubility parameters of pure nonelectrolyte organic compounds at 298.15 K and atmospheric pressure. To propose comprehensive, reliable, and predictive models, about 1400 data belonging to experimental solubility parameter values of various nonelectrolyte organic compounds are studied. The genetic function approximation (GFA) mathematical approach is applied for selection of proper model parameters (molecular descriptors) and to develop a linear QSPR model. To study the nonlinear relations between the selected molecular descriptors and the solubility parameter, two approaches are pursued: the three-layer feed forward artificial neural networks (3FFANN) and the least square support vector machine (LSSVM). Furthermore, the Levenberg -Marquardt (LM) and genetic algorithm (GA) optimization methods are respectively implemented to optimize the 3FFANN and LSSVM models. Consequently, we obtain three predictive models with satisfactory results quantified by the following statistical parameters: absolute average relative deviation (AARD) of the represented/predicted properties from existing experimental values by the GFA linear equation of 4.6% and squared correlation coefficient of 0.896; AARD of the QSPR-ANN model of 3.4% and squared correlation coefficient of 0.941; and AARD of 3.1% and squared correlation coefficient of 0.947 evaluated by the QSPR-LSSVM model. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|