Evaluation and prediction of synergistic antioxidant effects of green walnut hulls, potato peel, and date pulp extracts on the stability of sunflower oil by deep neural networks

Abstract

This study investigated the synergistic effect of the combination of three plant extracts including green walnut hulls, potato peel, and date pulp on the oxidation of sunflower oil over 15 days of storage at 70 °C. The total polyphenol and flavonoid compounds of three extracts were measured. Also, the antioxidant efficiency was studied by evaluating the DPPH scavenging assay and IC₅₀. Furthermore, the effectiveness of plant extracts on the oxidation of sunflower oil was determined by measuring p-anisidine (AV), peroxide (PV), thiobarbituric acid (TBA-V), and total oxidation (TOTOX). The total polyphenols and flavonoids ranged from (TP: 1084.36, 1076.59, and 414.71mg GAE/100 g extract; and TF: 549. 9mg CE/100 g extract, 475.28, 304.18mg CE/100 g extract) for green walnut hulls, potato peel, and date pulp extracts, respectively. The DPPH assay indicated that TBHQ is almost 2, 3, and 24 times more effective than those of green walnut hulls, potato peel, and date pulp extracts, respectively. According to the obtained results the combination of plant extracts indicated high antioxidant effects which was competitive with the synthetic antioxidant of TBHQ. The best synergistic effects were obtained for GP100 in terms of PV and TOTOX values, and SP200 and GP500 for TBA-V, and AV, respectively. Furthermore, a regression model based on a deep neural network to predict oil oxidation was proposed for the first time. First, a dataset using data collected from the laboratory experiments was organized. After developing several regression models based on deep neural networks, the models were trained and tested to choose the better model. Finally, cross-validation techniques to test the model for prediction of four parameters of AV, PV, TBA-V, and TOTOX values were conducted. The accuracy of predictions for AV, PV, TOTOX, and TBA-V was more than %99, %99, %94 and %90, respectively.