The control problems that arise in a fed- batch process are caused by the poorly understood nature of the process, its nonlinearity, the wide range of operating states passed through during a batch and the unmeasurability of key process variables.
In this paper alternative approaches to modelling and state estimation of the process will be outlined and a summary given of the comparative performance of numerically based and neural net derived models. A Self organising system based on fuzzy logic controller has been adapted to control the state variables of the process. The feasibility of using pattern recognition for modelling and state estimation of the process will be illustrated. Finally a brief treatment on the software structure, including expert system shells, that will allow these emerging AI techniques to be applied in real time to the fermentation process will be described.