Prediction of overall glucose yield in hydrolysis of pretreated sugarcane bagasse using a single artificial neural network: good insight for process development

BACKGROUND: In this work a single artificial neural network (ANN) was used to model the overall yield of glucose (Y-GLC) as a function of a wide range of operating conditions of both pretreatment and enzymatic hydrolysis. RESULTS: The model was validated experimentally and presented good predictions of Y-GLC. Sensitivity analysis using the ANN model indicated that most of the operating parameters, except for pretreatment time, were statistically significant (P-value<0.05). Experiments showed that the processing of sugarcane bagasse (in natura) results in a satisfactory glucose yield of 69.34% when pretreated for 60 min with low initial biomass concentration and acid concentration (10% and 1.0% w/v), and followed by enzymatic hydrolysis for 72 h with 3.0% w/v substrate loading and 60 FPU per g(WIS) enzyme concentration. CONCLUSION: This study demonstrated how pretreatment and enzymatic hydrolysis data can be used to parameterize a single ANN model. Acceptable predictions of Y-GLC are achieved in terms of RSD, MSE and R-2. Supported by the model, this study provided a good insight for process development. (C) 2017 Society of Chemical Industry.