COMPARISON OF RESPONSE SURFACE METHODOLOGY AND ARTIFICIAL NEURAL NETWORK FOR MODELLING

ABSTRACT

Citric acid was produced from Groundnut shell, Pineapple peel and Potato peel through solid state fermentation using Aspergillus niger. Mixture design and response surface methodology were used to determine the effect of a tri-substrate feedstock (groundnut shell, potato peel and potato peel) on the production of citric acid using A. niger. Response surface methodology (RSM) and Artificial neural network (ANN) were used to model the effect of chelating agents on citric acid production. To design the experiment, a Box-Behnken design of RSM was applied using EDTA, NaF and (NH4)2C2O4 as the independent variables and citric acid yield as the response. Statistical performance indicators showed RSM (R 2 = 0.9822, MAE = 0.002353, AAD = 0.2%) models describe the process with higher precision and accuracy compared to ANN (R 2 = 0.9802, MAE = -0.10235, AAD = 0.22976%). The results also showed that the best combination of the estimated process input variables (EDTA 0.3g/L, NaF 0.053 g/L and (NH4)2C2O4 2 g/L) with highest citric acid yield (30.99g/L) was given by RSM. A precision of 12.297 means the model is adequate and good enough to navigate the design space.