Multivariable adaptive control of baker's yeast fermentation

Abstract

The paper describes an ongoing investigation into the application of a multivariable adaptive control technique to a fed-batch baker's yeast fermentation. The application of conventional 3-term controllers provides poor control due to problems in tuning individual loops, owing to the time-variant and nonlinear properties of the process. Furthermore, the main process variable, cell mass, cannot be measured directly and hence control must be effected by either estimating the cell mass online or, alternatively, by controlling the environment for optimal cell growth. The latter approach requires the simultaneous optimisation of two or more process variables, requiring multivariable techniques. Results are presented describing the practical problems of establishing a suitable structure for the adaptive technique and the experience gained in using recursive least-squares estimation of process variables. The paper also includes results of the online application of multivariable adaptive control on a pilot-plant fermenter.

Modelling and control of fed-batch fermentation processes: towards improved quality and productivity

“…The convergence time for parameters in the adaptive control can be reduced by using a priori knowledge of the initial parameters (Williams and Montgomery, 1986). Reported applications of adaptive control schemes include controlling pH and temperature of the fermenter (Melin et al, 1982), waste water treatment (Bastin et al, 1982), baker's yeast (Williams et al, 1984;Williams and Montgomery, 1986;Ramseier et al, 1992), penicillin (Montague et al, 1985), alcohol (Samaan et ~rl, 1990;Roux et al, 1992) and so on. In these approaches, the key process variables are usually required to be measurable on-line.…”