Production of bioethanol from pumpkin peel wastes: Comparison between response surface methodology (RSM) and artificial neural networks (ANN)

Abstract

Pumpkin peel wastes are an unutilized source of starch, whose bioethanol production has not been reported by any research work so far. Such bioethanol is believed to show the potential uses of these wastes. Therefore, the present work was conducted to optimize the reducing sugar concentration and bioethanol production from pumpkin peel wastes using two modeling approaches (artificial neural networks and response surface methodology). Actually, a central composite rotatable design was used to optimize bioethanol production to obtain maximum reducing sugar and bioethanol concentrations. ANN proved to be superior to RSM in terms of its estimation and prediction capabilities. Therefore, the optimum conditions were obtained based on predicted ANN-model as follows. Concerning the hydrolysis process, hydrolysis time was 120 min, loading substrate was 17.5 g/L, α-amylase concentration was 7.5 U/g and amyloglucosidase concentration was 56.40 U/mL. As for the fermentation process, the optimal conditions were: fermentation temperature 45 °C, pH 5.06, shaking speed 188.5 rpm, and yeast concentration of 1.95 g/L. Under these conditions, the experimental concentration values of reducing sugar and bioethanol were 50.60 and 84.36 g/L, respectively, which are in good agreement with those predicted by the ANN-model (84.27 and 50.69 g/L, respectively). Besides, the results revealed that substrate loading and fermentation temperature were the most significant factors affecting the reducing sugar and bioethanol concentration, respectively (p < 0.01). Subsequently, the kinetics of yeast growth and bioethanol formation under the optimized conditions were estimated using the Monod, logistic and modified Gompertz models, respectively. Subsequently, the pumpkin peel wastes would offer an energy-saving alternative for fuel-ethanol production.