MODELING AND OPTIMIZATION OF ETHANOL FERMENTATION USING Saccharomyces cerevisiae: RESPONSE SURFACE METHODOLOGY AND ARTIFICIAL NEURAL NETWORK

In this study, the capabilities of response surface methodology (RSM) and artificial neural networks (ANN) for modeling and optimization of ethanol production from glucose using Saccharomyces cerevisiae in batch fermentation process were investigated. The effects of three independent variables in a defined range of pH (4.2-5.8), temperature (20-40 degrees C) and glucose concentration (20-60 g/l) on the cell growth and ethanol production were evaluated. The results showed that the prediction accuracy of ANN was apparently similar to RSM. At optimum conditions of temperature (32 degrees C), pH (5.2) and glucose concentration (50 g/l), suggested by the statistical methods, the maximum cell dry weight and ethanol concentration obtained from RSNI were 12.06 and 16.2 g/l, whereas experimental values were 12.09 and 16.53 g/l, respectively. The present study showed that using ANN as a fitness function, the maximum cell dry weight and ethanol concentration were 12.05 and 16.16 g/l, respectively. Also, the coefficients of determination for biomass and ethanol concentration obtained from RSNI were 0.9965 and 0.9853 and from ANN were 0.9975 and 0.9936, respectively. The process parameters optimization was successfully conducted using RSNI and ANN; however, prediction by ANN was slightly more precise than RSM. Based on experimental data, the maximum yield of ethanol production of 0.5 g ethanol/g substrate (97% of theoretical yield) was obtained.