Page 348 - Design of Simple and Robust Process Plants

334 Chapter 8 Instrumentation, Automation of Operation and Control

Cooling
Height duty

Mass balance L T
control P
Q T
Reflux (R) Distillate (D)
Heating
Sharper V duty
separation
L B
Q B
Temperature
Bottoms (B)
Fig. 8.17. Distillation column with five CV's, MV's and temperature profile.

In the above, the dominant variables and its pairing are selected are selected for unit
control. There remain many loops to be closed, such as product quality and inven-
tory loops, which still leaves many options open. These may lead to both static and
dynamic interactions, and avoid closure of the loops.
When we return to the distillation example in Figure 8.17, during the dominant
variable selection we found that the separation is dominant and influenced by the
heat input. A normal distillation has five CVs, which include two qualities Q T and
Q B , two levels L T and L B , and a pressure P. Five MVs are involved: reboiler heat duty
(V); cooling duty (C); reflux (R); bottom (B); and distillate (D) valves ± hence it is a
five-by-five system.
One loop is selected as dominant variable V, while another dominant variable is
the mass balance effected by bottom flow B, distillate flow D or indirectly by reflux
R. The mass balance control shifts the concentration profile, reflected in a tempera-
ture profile, up and down the column, while an increased separation is reflected in a
steeper temperature profile.
There are still many pairing options open, and even when pressure is controlled
with cooling duty the remaining options are: Q T by V or D,B, R; Q B by V or D,B, R;
L T and L B by D,B,R in this list we did not include reflux ratio as a MV. The question
is which is the preferred option for a specific distillation column.

8.4.7.4 Self-optimizing control
A design methodology is discussed based on work from Skogestad et al. (1999) for
the selection of the controlled variables to obtain a self-optimizing control structure.
Definition:

A self-optimizing control structure is designed to achieve acceptable loss with constant
set point values for the controlled variables.