Evaluation of the response surface and hybrid artificial neural network-genetic algorithm methodologies to determine extraction yield of Ferulago angulata through supercritical fluid

Abstract

In this research, the influence of operational parameters such as pressure, temperature, particle size and dynamic time on the extraction of Ferulago angulata via supercritical carbon dioxide was investigated. Two models, response surface method (RSM) and artificial neural network (ANN), were applied for modeling and predicting of the extraction yield. The capability and sensitivity analysis of two methodologies was evaluated by some statistical parameters including, correlation coefficients (R2), mean square error (MSE), and absolute average relative deviation (AARD). The error values of RSM model (R2 = 0.9645, MSE = 0.0032 and AARD = 3.39) and ANN model (R2 = 0.9971, MSE = 0.0014 and AARD = 1.32) were calculated. Both models showed good agreement with the experimental data, but ANN model was more accurate than the RSM model. RSM and ANN-GA models were subsequently utilized to determine optimal operating conditions in order to gain the maximum yield of Ferulago angulate extraction. The essential oil extraction yield in the optimal conditions was estimated to be 0.851 w/w% and 0.863 w/w% by the RSM and ANN-GA models, respectively. Assessment and evaluation of the above mentioned models were carried out through performing triplicate experiments and finally the average experimental extraction yield was achieved to be 0.867 ± 0.004 w/w% at the best operating conditions 20 MPa, 35 °C, 0.742 mm and 101.3 min.