(DMC) strategy is rather simple. In a DMC strategy (Figure 3.9) the non-linear controller
may be derived analytically from the process model and from a reference model. The
reference model imposes desirable dynamics for the convergence of control variable to
the corresponding setpoint profile, e.g. first order dynamics of convergence. This strategy
was applied with good results for a penicillin production process (Oliveira et al., 2000b).
Trends in Biorector Control: On-line Optimising ControlThe follow up of an open-loop control implementation is often the conclusion that the
process is operated at sub-optimal conditions not coincident with the expected process
performance evaluated by the off-line optimisation procedure. The natural conclusion to
be made is that in order to improve further productivity and/or product yield and/or
product quality, on-line optimising control should be implemented. There are two
obvious explanations for this. First, open-loop control is not robust to disturbances in the
process since no process data acquired on-line is used to take corrective actions. In
practice, open-loop control is complemented by on-line corrective actions performed
manually by process operators. When for instance viscosity increases above a threshold
level the operator supersedes the open-loop set-point decreasing manually the sugar feed
rate in order to prevent the process runing into oxygen limitation. Second, open-loop
control leads invariably to sub-optimal process trajectories due to inaccuracy of models,
also commonly referred to in the literature as parametric uncertainty. The straight-
forward way to improve robustness in open-loop control is investing in on-line
optimising control.
Basically, on-line optimisation control consists of the optimisation of a cost function
on-line and in real-time using successively measurements acquired on-line and/or off-line
for the current process run. A schematic representation for the tasks accomplished in on-
line optimising control is depicted in Figure 3.10.
Figure 3.10 General structure of an
on-line optimisation control system.
New methodologies for multiphase bioreactors 3 83