Neural Identification of Supercritical Extraction Process with Few Experimental Data
| Authors: | Soares Rosana, Universidade Federal do Para, Brazil
de Oliveira Roberto Limão, Universidade Federal do Para, Brazil
Miranda Vladimiro, Faculty of Engineering of the University of Porto/Institute of Engineering in Systems and Computers of Porto, Portugal
Barreiros José Augusto, Universidade Federal do Para, Brazil |
|---|
| Topic: | 1.1 Modelling, Identification & Signal Processing |
|---|
| Session: | Applications of Nonlinear Modeling Methods |
|---|
| Keywords: | Training, Neural Network, Identification, Process Models |
|---|
Abstract
In this paper, an Artificial Neural Network (ANN) is designed to model an extraction process that uses a supercritical fluid as solvent. Two approaches are used in the ANN training, which they differ in the strategy that was used to complement the experimental data that were collected during extraction procedures of useful compositions for the pharmaceutical industry. While the first methodology involves the interpolation of the data during an operation of extraction, the second one uses pseudo experiments generated from the real data and of the incorporation of qualitative characteristics of the process.
