Une commande nonlineaire de bioprocedes fedbatch

Abstract

The need to control bioprocesses, that by their very nature are nonlinear and with time varying parameters, leads to different approaches based to linear and/or nonlinear technics. We present here a nonlinear adaptive one, of L/A type. Its basic idea is to operate on a nonlinear elementary relation, such as the ratio set point/output, and to use the power of the ratio as a (self)tuning parameter of the algorithm.

Applying this technique on an elementary fermentation process of fedbatch type, nonlinear algorithms, able to control biomass or/and substrate concentrations of the bioprocess, are obtained. They take into account constraints, such as the positivity of process variables, imposed by this one, and compensate with good dynamics some disturbances effects; by using selftuning powers one may improve their impact. They have a very simple structure and present other application facilities: computer implementation of L/A algorithms is very easy: there is no need to introduce initial and/or test conditions and the memory used is a minimal one (all L/A algorithm consider only current values of involved variables).

Firstly we present the idea of L/A control laws and then apply them to control a fedbatch process. Time responses and phase plane trjectories, obtained by simulations, illustrate their performances. It is worthy to note that real life experiences are under way and the results obtained up to now seem to confirm our ideas.