Parameter estimation of tapioca starch hydrolysis process: Application of least squares and genetic algorithm

Abstract

The performance of genetic algorithm (GA) in nonlinear kinetic parameter estimation of tapioca starch hydrolysis was studied and compared with the Gauss-Newton method. Both methods were employed for determining the model parameters of the modified version of Gonzalez-Tello model. To estimate and validate the model parameters, experimental works involving hydrolysing tapioca starch were conducted. The model was then used to predict glucose concentration profile for a given initial condition of the tapioca hydrolysis process. In terms of error index values, both methods produced good results. This study showed that the impact of user defined parameters of the GA was insignificant as compared with the influence of initial parameters of the Gauss-Newton method on the predictive performance. Furthermore, the GA approach requires no guessing of the initial values and is able to produce reasonable solutions for the estimated parameters.