Comparison between developed models using response surface methodology (RSM) and artificial neural networks (ANNs) with the purpose to optimize oligosaccharide mixtures production from sugar beet pulp

Abstract

This work aimed the assessment of the use of artificial neural networks (ANNs) as alternative tool for modelling and predicting the suitability of sugar beet pulp (SBP) to produce oligosaccharides in comparison with the response surface methodology (RSM). The variables polygalacturonase to solid ratio (PGaseSR), cellulase activity to polygalacturonase activity ratio (CPGaseR), and reaction time (t) were selected as independent variables and their effects on the recovered liquors mass, the conversion of different polysaccharide into monosaccharides, and the conversion of each polysaccharide into oligomers were investigated. ANN models improved the RSM models between a 5.58% and a 61.78% for the solid yield (%) and Galactan conversion into galactooligosaccharides (%), respectively. However, RSM models presented better accuracy to predict the polysaccharides conversion into monosaccharides. The ANNs implemented in this study showed that are suitable to optimize and predict the oligosaccharides production using direct enzymatic hydrolysis from SBP.