Modeling of Microwave-Assisted Extraction of Lycopene from Tomato Peels

Abstract

Dry tomato peels were investigated as a natural source of lycopene in this study. Microwave-assisted extraction (MAE) was used for the isolation of bioactive compound of interest. The MAE was performed at a microwave power of 120, 230, and 385 W for a duration of 2, 5, 8, and 10 min. The concentration of lycopene in the separated extract with n-hexane was quantified by UV/Vis spectrophotometer. The absorbance at 501 nm was selected for the determination of lycopene content and the calculation of lycopene yield (mg/100 g). An adequate feedforward artificial neural network (ANN) model with architecture 2-10-1, trained by the Levenberg-Marquardt (LM) backpropagation algorithm was developed for modeling the MAE process and for the prediction of the lycopene yield. The model predicted lycopene yield with a high coefficient of determination R2=0.9957 and low mean squared error MSE=0.475. The response surface methodology (RSM) was successfully applied and combined with the ANN model to optimize the MAE process with R2=0.9477 and a low mean absolute error MAE=1.84507. The analysis of variance showed that extraction time and time-squared have statistically significant effects on the lycopene yield. The created 3D response surface showed that the optimum lycopene yield was 27.25 mg/100g of tomato peels at the microwave power of 324 W and extraction time of 8.6 min.