Biomedical Engineering Reference
In-Depth Information
substitute a predetermined feeding strategy that is employed to guarantee no
limitation of oxygen. The dissolved oxygen consumption rate r DO X t is measured
via a soft sensor using the approximation of Eq. ( 9 ), whereas b is a proportionality
factor and CER is the carbon dioxide evolution rate, which is assumed to be
approximately the carbon dioxide transfer rate.
r DO X t ¼ b OUR þ CER
V
ð 9 Þ
The master-loop controls the specific consumption rate r DO . The slave loop
controller is applied to control the feed rate based on OUR and CPR measure-
ments. It is possible to efficiently control the specific consumption rate of oxygen.
The main advantage is that it is not limited to a specific strain of microorganisms
and applicable for a wide variety of fungal fermentations.
Another application for MIMO-PID bioprocess control is a multiloop PID
feedback controller for HCDC control applied by Chung et al. [ 25 ], coupling the
OTR and the CPR. They compared the method with a model predictive controller
and presented better results for their MIMO-PID using simulation studies. The
controller can compensate for disturbances in the exhaust gas measurement data.
Ranjan and Gomes [ 26 ] also applied a cascade MIMO, showing a performance
enhancement compared with a normal PI controller.
For all these previously described applications, the parameters of the PID
controller
must
be
determined.
Different
approaches
to
determine
the
PID
parameters are discussed in the next section.
3.2.3 PID Tuning
The obviously crucial part for all PID control-based approaches is the determi-
nation of the corresponding PID parameter values. Changes in the process dynamic
will most likely lead to suboptimal control actions. Ideally, a tuned controller
should show a minimum of oscillation and lead the system quickly and reliably to
the set-point.
Tuning methods can be divided into two groups: parametric model and non-
parametric. Parametric methods use either model or experimental data to deter-
mine the controller parameters and are mostly described as offline tuning methods,
though online approaches have also been tested. Nonparametric methods only
partially use models such as critical states and are suitable for online use as well as
for implementation without previous extensive plant studies. Wahab et al. [ 20 ]
compared four nonparametric methods for multivariable PID tuning, introducing
one on their own and comparing it with established methods from Davison [ 27 ],
Penttinen-Koivo [ 28 ], and Maciejowski [ 29 ]. Soons et al. [ 23 ] used a parametric
tuning algorithm proposed by Bastin and Dochain [ 30 ], guaranteeing stable
behavior and fast convergence towards the set-point. Other parametric approaches
Search WWH ::




Custom Search