Optimization of supercritical extraction of galegine from Galega officinalis L.: Neural network modeling and experimental optimization via response surface methodology

Davoodi, Pooya; Ghoreishi, Seyyed Mohammad; Hedayati, Ali

Abstract

Supercritical CO2 extraction of galegine from Galega officinalis L. was carried out under different operating conditions of temperature (35-55 °C), pressure (10-30MPa), dynamic extraction time (30-150min), CO2 flow rate (0.5-2.5 mL/min) and constant static extraction time of 20 min. Design of experiment was by response surface methodology (RSM) using Minitab software 17. The response surface analysis accuracy was verified by the coefficient of determination (R2=93.4%) along with modified coefficient of determination (mod-R2=87.7%). The optimum operating conditions were found by using RSM modeling to be 42.8 °C, 22.7MPa, 141.5min and 2.15 mL/min, in which the maximum galegine extraction yield of 3.3932mg/g was obtained. Artificial neural network (ANN) using Levenberg-Marquardt backpropagation training function with six neurons in the hidden layer was implemented for the modeling of galegine extraction such that the coefficient of determination (R2) was 96.6%.

Citation

Davoodi, P., Ghoreishi, S. M., & Hedayati, A. (2017). Optimization of supercritical extraction of galegine from Galega officinalis L.: Neural network modeling and experimental optimization via response surface methodology. Korean Journal of Chemical Engineering, 34(3), 854-865. https://doi.org/10.1007/s11814-016-0304-2

• Journal Article Type: Article
• Acceptance Date: Oct 24, 2016
• Online Publication Date: Dec 19, 2016
• Publication Date: 2017-03
• Deposit Date: Jun 12, 2023
• Journal: Korean Journal of Chemical Engineering
• Print ISSN: 0256-1115
• Electronic ISSN: 1975-7220
• Publisher: Springer Verlag
• Peer Reviewed: Peer Reviewed
• Volume: 34
• Issue: 3
• Pages: 854-865
• DOI: https://doi.org/10.1007/s11814-016-0304-2
• Keywords: General Chemical Engineering; General Chemistry; Supercritical Extraction; Galega officinalis L.; Artificial Neural Network (ANN); Response Surface Methodology (RSM); Galegine