Optimisation of fed-batch fermentation processes usually employs the calculus of variations to determine optimal feed-rate profiles that will maximise a given objective function. This results in a two-point boundary-value problem and because of the nonlinear nature of the processes, the optimal solution usually falls out as an openloop control algorithm. One advantage of this approach is that it does not need measurements of state variables which are often difficult to obtain on-line. Instead it assumes that the state variables are proceeding along known paths a-priori determined by models. However, the disadvantage of such an approach is that the performance will severely deteriorate in the presence of process disturbances or plant-model mismatch. Recently, on-line estimation of state variables has been successfully developed and used in the industry and therefore a method which operates the fedbatch fermentation in a closed-loop control scheme using state feedback is proposed. This is achieved by a two-step method. First, the optimal substrate concentration profile which governs the biochemical reactions in the fermentation process is determined. Then a controller is designed in closed-loop form to track this desired profile. Simulation studies for both primary and secondary metabolite production processes show that better performance is obtained by this closed-loop aptifaaal control method than by the openloop optimal feed rate profile method. This is due to the self-correcting property of the proposed method, which proves to be advantageous when there are disturbances or plant-model mismatch.

No abstract

A predictive time-sequence iterative learning control method is proposed in this paper and its potential application to some industrial processes that are difficult to describe mathematically but often with input/output data measured from the processes available, such as fed-batch fermentation processes, has been studied. By the proposed method, the control performance can be improved in the first trial instead of waiting until the following trials. The idea of the proposed method is to use predictive learning for the future control in the same batch operation, and to use penalty/reward to achieve improvement on past control experiences. A simulation study in applying the proposed learning control method to a fed-batch fermentation process is illustrated.

No abstract

This paper describes research into modelling, estimation and control of fed batch fermentation and an application to a large fermentation plant producing secondary metabolites. The problems are examined in a general sense from the viewpoints of both biological science and control engineering. Very extensive modelling and estimation work has been carried out. Only the most significant outcomes are reported here but references to the detailed research are given. Some of the research has resulted in the development of a novel advanced supervisory system -the Intelligent Process Management System (IPMS). Experience of applying the IPMS to a large-scale production process is described. The software has been designed on a modular basis and although implemented initially at one production plant, the approach is generic and fundamentally plantindependent so that further application to a range of other plants is envisaged.