Page 353 - Design of Simple and Robust Process Plants

8.4 Control Design 339
From the RGA matrix the control loops can be selected by pairing the controlled
outputs with the manipulated inputs in such a way that the relative gains are posi-
tive and as close as possible to unity. The RGA must be complemented with
dynamic interactive control analysis.

8.4.9.
Evaluate Dynamic Interaction of the Reduced Set of Selected Pairings

Dynamic models are the bases for the development and testing of robust controllers.
These controllers might also include models based on algebraic equations to de-
coupled loops or support feed-forward and feed-back actions. Dynamic simulations
have some specific requirements. Dynamic model building is nevertheless the avail-
ability of unit library models, still time consuming. In general reduced versions are
made of the static simulations. As control studies do not require the same level of
detail, while the studies concentrate around a steady-state operational point where
discrete event are avoided. An exception is that of the models for batch processes,
which have to simulate discrete events.

Model reduction The number of components is often reduced in as much as they
do not play a major role regarding dynamics of the system. They are kept in the
simulation when they:
. have a significant effect (negative as well as positive) on the reaction;
. are key components for separations; and
. are accumulated in the system and need to be specifically removed.
Reactor models are quite commonly reduced to conversion models (Verwijs, 1992,
2001, although when product properties must be controlled (as in polymer pro-
cesses), components cannot be excluded.
Distillation models can be reduced by lumping trays of towers. A large effect on
the size of the model (much less equations) for a distillation tower are achievable
when the number of trays are reduced by lumping and the number of components
reduced.
Heat exchangers are normally not dynamically simulated, but in the case of heat
integration this might be a requirement, particularly so when there is no back situa-
tion (as provided in Figure 4.29 in Chapter 4). Due to their fast dynamics in relation
to other process dynamics, pressure control is in general not dynamically simulated;
neither are level controllers.
Some designers work on black box models derived by mathematical model reduc-
tion techniques from fundamental models these are less transparent and should, to
authors opinion, not be used for control strategy design. With these models the pro-
cess insight is lost ± you could call this designing in the dark. Although these mod-
els find useful application in model-based controllers designed to save run time in a
real-time environment.
The objective of the dynamic simulation is to design both control strategies and
controllers, and this requires a reasonable accuracy of the dynamics of the models.