Investigations Into the Application of State Estimation and Parameter Adaptive Control in FED Batch Fermentation for Penicillin Production

Abstract

Abstract This paper investigates the feasibility of state estimation and parameter adaptive control applied to a fed-batch penicillin fermentation process. The highly non-linear pseudo-batch process is mechanistically modelled using the approach of Bajpai (1980) with an extension to include terms for the production of carbon dioxide. Penicillin mould biomass is controlled to a reference trajectory by applying adaptive control to the on-line estimate of biomass obtained from a measurement of carbon dioxide production, use of the fermentation model and an extended Kalman filter. Various control methods have been considered; conventional PI, sub-optimal Riccati solution and minimum variance and pole-placement self-tuners, the parameter estimation being carried out by one of three possible prediction error-correction methods. Results are presented from the application of the adaptive control algorithms in a real time environment to the ‘control’ of a digital simulation of the penicillin fermenter running on a seperate desk top computer. The study forms part of a university/industrial collaborative project aimed at the optimising control of large fed batch fermenters for penicillin production.