determine the algorithm to be used for each controller (P, PI, or PID)
and to tune each controller. We strongly recommend the use of P-only
controllers for liquid levels (even in some liquid reactor applications).
Tnning of a P controller is usually trivial: set the controller gain equal
to 1.67. This will have the valve wide open when the level is at 80
percent and the valve shut when the level is at 20 percent (assuming
the stream flowing out of the vessel is manipulated to control liquid
level; if the level is controlled by the inflowing stream the action of the
controller is reverse instead of direct).
For other control loops, we suggest the use of PI controllers. The
relay-feedback test is a simple and fast way to obtain the ultimate gain
(K,) and ultimate period (P,). Then either the Ziegler-Nichols settings
(for very tight control with a closed-loop damping coefficient of about
0.1) or the Tyreus-Luyben (1992) settings (for more conservative loops
where a closed-loop damping coefficient of 0.4 is more appropriate) can
be used:
3.3 Steps of Plantwide Process Control
Design Procedure
In this section we discuss each step of the design procedure in detail.
The use of PID controllers should be restricted to those loops where
two criteria are both satisfied: the controlled variable should have
a very large signal-to-noise ratio and tight dynamic control is really
essential from a feedback control stability perspective. The classical
example of the latter is temperature control in an irreversible exothermic
chemical reactor (see Chap. 4).