BIOREACTOR MODELLING: A PROCESS CONTROL
PERSPECTIVE
Several closed-loop control strategies are commonly implemented in fermentation
systems regarding physico-chemical and environmental properties (temperature, pH,
pressure, oxygen partial pressure). Classical control policies such as PID or on/off control
are sufficient to solve successfully such control tasks. However, to progress into
biological mechanisms control, models are required to establish the link between
extracellular physico-chemical quantities (the quantities that we can manipulate) with
intracelullar phenomena. In several ways and variants, dynamic bioprocess models for
control aim at establishing this link.
In spite of the undeniable potential that models may have in bioprocess control it
should be stressed that modelling is not a current practice in the biotechnological
industry. For such industrial applications we should probably ask ourselves why model-
based developments are not sufficiently attractive rather than engaging on the task of
developing all kinds of more and more complex models which at the end are not of
practical utilisation.
Several reviews on bioprocess modelling have been published (Kneinstreuer, 1987;
Thornhill and Royce, 1991; Nielsen and Villadsen, 1992; Bellgardt, 1993) and it is not
intended to add much to these here. In addition to the traditional mathematical modelling
in bioprocesses other modelling techniques based on artificial intelligence have been
catching attention in recent years. These techniques will be briefly presented in relation to
the potential application for process control.
Bioprocess models address two main systems: 1) the bioreactor system and 2) the cells
system. Bioreactor models deal with transport processes, rheology, mass/heat/momentum
transfer aspects and flow patterns. Cell models deal with the kinetics at the individual cell
level and at the whole cell population level. The bioreactor system and the cell system
have very complex interactions and cannot be analysed separately. Cells are continuously
transforming the liquid phase by consuming several nutrients, which are metabolised into
several products, some of them being excreted into the surrounding media.
Classical Mathematical ModellingBasic modelling conceptsHaving in mind model-based applications for bioprocess monitoring and control, when
we develop a model we are mainly interested in the description of the dynamics of
macroscopic quantities that influence the behaviour of the cell system, recognising that
cells are ultimately the bioreaction promoters. These quantities are mainly concentrations
in the biosuspension of several components such as biomass, substrates, and products,
and also several other quantities such as temperature and pH.
Those variables which define the process state are termed state variables. A
mathematical (mechanistic) dynamical description of the state variables, irrespective of
New methodologies for multiphase bioreactors 3 67