A Simulation Study of Non-Linear Structure Identification of a Fed-Batch Bakers' Yeast Process

Abstract

The fed-batch bakers' yeast fermentation is a non-linear and time-varying process. A simulation model of the process, based on ordinary differential equations combined with empirical rule and algebric expressions, has been made and implemented in the Matlab software environment. Two non-linear structure identification techniques have been implemented in Matlab and tested on the simulated process: an extended Wiener (-Hammerstein) cascade model described by Volterra kernels [4] and a non-linear model, which is linear in the parameters [3]. The objective of the work was to identify the structure and parameters of a SISO-model of the glucose flow (input) → respiratory quotient (output) dynamical relation of the process. This model will be used as a reference model for control purpose. The results obtained show that both techniques are capable of modelling the process. The technique of Billings is found to be more flexible with respect to the choice of model structure and the estimation of the parameters. Note that both techniques can be applied to any other non-linear process.