Ultrasound assisted phytochemical extraction of persimmon fruit peel: Integrating ANN modeling and genetic algorithm optimization

Abstract

In the present study, ultrasound assisted extraction (UAE) of phytochemicals from persimmon fruit peel (PFP) was modeled using an artificial neural network (ANN) and optimized by integrating with genetic algorithm (GA). The range of process parameters selected for conducting the experiments was ultrasonication power (XU) 150–––350 W, extraction temperatures (XT) 30–––70 °C, solid to solvent ratio (XS) 1:15–––1:35 g/ml, and ethanol concentration (XC) 40–––80 %. The range of responses total phenolic content (YP), antioxidant activity (YA), total beta carotenoid (YB) and total flavonoid content (YF) at various independent variables combinations were found to be 7.72–––24.62 mg GAE/g d.w., 51.44–––85.58 %DPPH inhibition, 24.78–––56.56 µg/g d.w. and 0.29–––1.97 mg QE/g d.w. respectively. The modelling utilised an ANN architecture with a configuration of 4–12-4. The training process employed the Levenberg–Marquardt method, whereas the activation function chosen for the layers was the log sigmoid. The optimum condition predicted by the hybrid ANN-GA model for the independent variables, XU, XT, XS and XC was found to be 230.18 W, 50.66 °C, 28.27 g/ml, and 62.75 % respectively. The extraction process was carried out for 25 min, with 5-minute intervals, at various temperatures between 30 and 60 °C, to investigate the kinetic and thermodynamic characteristics of the process, under the optimal conditions of XU, XS and XC. The UAE of phytochemicals from persimmon peel followed pseudo second order kinetic model and the extraction process was endothermic in nature.