Page 355 - Design of Simple and Robust Process Plants

8.4 Control Design 341
The evaluation of the controllability performance is valuable input for the final
flowsheet selection during the process synthesis (Perkins, 1989).

8.4.10.
Establish the Final Pairing and Design the Controllers

This is the working platform for the control specialist. When a control strategy for a
process has to be developed, the first step is to compute the RGA matrix as a func-
tion of frequency for the pre-selected pairings. The selection of the adequate set of
input±output pairs is the challenge in reference to the following aspects:

. Prefer yielding the RGA matrix close to identity at frequencies around cross-
over to ensure that instability is not caused by other loops.
. Avoid pairing with a negative steady-state relative gain.
. Prefer a pairing with minimal restrictions on the achievable bandwidth to
obtain acceptable disturbance rejection while realizing stability.
The next steps are to:
± compute PRGA and CLDG, and plot these as functions of frequency;
± analyze potential individual loops for stability, and feasible performance
bandwidth;
± avoid input constraints at frequencies where control is required; and
± select the pairing for controller design.

8.4.11.
Develop and Test the Performance of the Controller in a Dynamic Simulation

The controllability analysis objective was to select pairing of CVs and MVs based on
simulations to achieve robust control at the basic control layer. The controller design
can start, based on the selection method. The control design can effectively be veri-
fied and tuned by implementation in the dynamic simulation. The above selection
procedure does not necessarily result in a set of pairings, but still might be subject
to unacceptable interaction The solution can be found in de-coupling of the loops by
defining the relation of the interaction derived from the fundamental models. The
interaction is neutralized in a multi-variable controller at basic control level as pre-
sented in the next section.

8.4.12.
Model-based Control at the Basic Control Level

Model-based control has a reflection of a complex multi-variable controllers which
have predictive, optimizing, and constraint control properties. These are quite com-
monly designed based on input and output modeling, but are less reliable and
require ongoing maintenance. These impressions are valid, and mainly caused by
the limited validity of the models. The modeling errors can be found in the process,