(CHAPTER,
THREE
CONTROL OF LUMPED PARAMETER SYSTEMS
3.1 INTRODUCTION
In this chapter we shall consider the control of lumped parameter processes,
processes described by ordinary differential equations. These include both
single-variable and multivariable dynamic systems. We shall begin with a
discussion of some of the key concepts and practical difficulties encountered in
lumped parameter control system design. Some of the control system design
procedures available for linear multivariable systems shall be presented and
illustrated with examples. Following this we shall introduce optimal control
theory, which represents one approach to control system design, and which
provides a theoretical foundation helpful in understanding material to be
‘covered later in the book. Finally, we shall discuss special techniques which may
be used for nonlinear dynamic systems.
To provide a very simple example of the type of problem we shall be
attacking and to clarify a number of concepts, let us consider the simple,
well-stirred, steam-heated mixing tank shown in Fig. 3.1. It is used to mix and
preheat a recipe of reactants before passing them to a chemical reactor. Now let
us suppose that we wish to design a scheme for controlling the effluent
temperature T(J) to a desired value T, by manipulating the steam rate Q(¢) with
valve v.
‘An open-loop control scheme for such a process would involve programming
the steam valve position over time without benefit of feedback information, such
as a tank effluent temperature measurement. Such a scheme might be useful if @
very good mathematical model of the process were available, €.2.,
a FG-7), OW
a ¥— * Voc,
70) = To G12)
and one had the start-up problem of driving 7(1) from the initial condition Ty to
GD)
940 ADVANCED PROCESS CONTROL
stem
Openoor
A} —FA, set
ew
A se +
Figure 3:1. seam-heated mixing tank.
Ge ni yl Tn opp ec pay ol 5 ere
tee Fant cbc Pope on gel be eae ain a
2 eee
Se rom de oc wan el
ee eo nen eee
foe cae eee ee ay one peice
See ee a ee aay tie te Solgar
og a ene epcieetleaeat triers
(PID) controller given by
a= aealen- met fin na +X Danny
wes tg ty he ont ance Hove vl lio
Tene cn une chars oer be al
Le ee ae eh eee eter ol
schemes for processes described by ordinary differential equations.
3.2 LINEAR MULTIVARIABLE CONTROL SYSTEMS
Perhaps the most commonly encountered control system design problem in the
process industries is the design of multivariable control systems. If the process
under study is linear, a very general model in the time domain (so-called state‘CONTROL OF LUMPED PARAMETER SYSTEMS 41
cariable notation) is
Ba acs Buta X(t) = Xo (32)
y=Cx (322)
where
is a k vector of disturbances,
is an n vector of states,
is an m vector of controls (manipulated variables), and
va
v2
yo
w
is an [ vector of outputs (those states or combination of states which can be
measured). The matrices
a a a bu ba bm]
an a by 2 . 2m
a-|® fy male ”
Gn 4, Ont Pam
cu te ein Mm Me Ye
aed Sol eel ™
“ns ve