Nonlinear Adaptive Control of Fermentation Processes Utilizing a Priori Knowledge

This paper describes both SISO and MIMO adaptive versions of nonliner Generic Model Control (GMC) applied to a baker´s yeast fermentation. An a priori nonlinear representation of process knowledge is combined with a simple and effective adaptation scheme to yield optimal flexibility of the model. Only a few parameters which appear linearly in the model need to be estimated on-line, resulting in very fast and potentially offset-free tracking of the process by the model-based controller. Simulations demonstrate that an adaptive MIMO version of GMC is superior to the corresponding nonadaptive version when faced with modeling errors. Experiments with a bench-scale yeast fermentation system demonstrate for the SISO case the applicability of adaptive nonlinear control methods to an actual process and compare adaptive GMC performance with that of a conventional PI controller.