Beruflich Dokumente
Kultur Dokumente
Nakamoto+
and Norihiro
Watanabe*
Japan
The decoupling and linearization control theory based on non-linear feedback transformation for nonlinear systems was applied to two non-linear and interactive chemical control processes. The first is a level
and temperature control process and the second a level and pH control process. To compensate for the mismatch between the real process and the process model, external PI controllers, which track the outputs from
the reference model, were added. The experimental results showed that even if the process is both non-linear
and interactive and the process model is not exactly known, a satisfactory control performance can be
obtained by applying this theory.
(Keywords: non-linear process; decoupling; linearization)
140
09Ss-1524/91/03014&06
0 1991 Butterworth-Heinemann
Ltd
control strategy which is called the multiinput/multioutput (MIMO) globally linearizing control, and this was
applied to a semi-batch co-polymerization
reactor
through numerical simulations. Paszkiewicti proposed
either a fuzzy logic controller or a saturation controller
as an external controller. These are both outer loop
controllers.
This paper reports the experimental application of the
decoupling and linearization control theory to two chemical control processes, a level and temperature control
process and a level and pH control process. The applicability of this theory to chemical processes will be examined.
i = f(X)
+fgj(X)U,
(la)
j=l
Y =
h(x)
(lb)
Let
(i=
vi = hay 7
(2)
such that the ith output depends linearly on the ith new
input v. Let pi denote the largest integer such that for all
k < pi, 1 5 j 5 m, and x E Rm:
l,...,m)
(8)
(i=
L,L:hi(
x) = 0
where
L/ h.(x)
1
= i
f,(x)2
j=l
8Xj
(44
(4b)
A(x)
+IL,~L;'~I
=
(54
b(x)
l,...,m)
(9)
(3)
(5b)
(i=
with
constant
coefficients
6,
(i 1, . . . , m;
k=O,. . . ,pi+ 1).
It is known that the decoupling and linearization
problem is solvable if and only if the matrix A( x ) is nonsingularj. More general expressions of A( x ) and b( x ),
proposed by Kravaris and Soroush3, can be used. However, Equation (5) is sufficient for practical applications.
a( x ) and l3( x ) are given by:
- 0
l,...,m)
where yy is a reference value for the external PI controller and is the output from the reference model which
generates the desired output. For example, the following
reference model can be used, which is similar to Equation
(9):
Of33
Yiset
u(x)
= -A-(x)b(x)
(64
P(x)
= A-(x)
(6b)
and input-output
Pi+
c Sik2= vi
(i=
I,...,m)
(10)
(i=
l,...,m)
(11)
k=O
(7)
k=O
ri
(i=l,...
Note that from Equation (7) the relationships between
the new inputs and the outputs are split into m independent linear subsystems.
pi + 1
and
,m;k=O,l,
oik = 6&.
. . ,pi+ 1 )
(12)
141
c----
i,
the following
= -
(gx:f (A)&
i, = (p)u,
(1%
+(&&)u*
(15b)
(l6a)
Y2 =
U6b)
x2
t
vi = 6,q,yise + K,,[y,*t-
(i=
l,...,m)
yi + f
I s
i
y,)dt]
(y,-
(13)
. . ,ri)
(14)
142
sc-4 =
x2 -
TON kx,lS
6,0x, ) -
C,PS(
x2 -
S
To)
~,oQSx,
c,psx,
1
(174
(17b)
C,P[(X,
ToI{ kx?
+
sx,
(1W
S(v,
(v2 -
b-41
62flx,)l
+
U8b)
2o0
600
1200
1600
I
600
I
1200
1600
20
600
1200
1800
3025-
2o0
Time
1
2o0
Time
tsecl
Figure 2 Output responses for the step change in the level set points.
With no external PI controllers: (-) measured values; (- - - -) set
points
yiW
( ydd
y,W
)e-6af
600
i= 1,2 )
1200
1600
Esecl
Figure 3 Output responses for the step changes in the level and
temperature set points. With no external PI controllers: (-) measured
values; (- - - -) set points
45
(19)
201
600
1200
1800
4oa--3530 25 -
I
2o0
600
Time
1200
1600
[secl
Figure 4 Output responses for the step change in the level set points.
With external PI controllers: (-) measured values; (- - - -) set points
143
I
600
201
0
I
1200
I
1600
= -
dxz _ -dt
(;)x;/f + (;)(U,
~a%lorlw
+ u*)
(204
1
)
[K(~2)
C*&,
Y2 =
600
Time
1200
[SeCl
Figure 5 Output
responses for the step changes in the level and
temperature
set points. With external PI controllers:
(-) measured
values; (- - - -) set points
a
Figure 6 Schematic
Difference pressure
valves; (e) pH meter
Xl
(214
x2
@lb)
Here xl and x2 are the liquid level and the value of pH,
respectively. uI and u2 are the feed flow-rate of the strong
acid and the strong base, respectively. As the procedure
to obtain the decoupled and linearized system based on
Equations (20) and (21) is the same as the previous
procedure, only the experimental results are given.
The values S and k are the same as the previous process. The feed concentrations were chosen as C,, = Cr,,,
= 0.03 mol cm-3. The feed flow-rates are constrained as
0 I ul, u2 5 22 cm3 s-r. The reference trajectories are
given by Equation (19) and the following parameters
were used: &, = i& = 0.02.
The control experiment with the external PI controllers was carried out as follows. At first, the set points
of the level and the pH were changed from 15 to 30 cm
and from 2.8 to 4, respectively, at the same time. At 900
s, the level set point only was changed from 30 to 35 cm
to examine the effect of the decoupling. This result is
shown in Figure 7. This figure shows that the influence of
the interaction is very small. In this experiment, the
following parameters were used: KP, = 3.0, Tll = 100,
KP = 0.3, and T, = 100.
fn the next expiriment, the set points of the level and
the pH were changed from 2.8 to 7.0 and from 15 to 30
cm, respectively, at the same time. To obtain a satisfactory result, however, the parameters of the external PI
controllers used in this experiment had to be changed to:
KP, = 0.1, Tl, = 300, KP2 = 0.03 and T12 = 300. This
experimental result is shown in Figure 8. The control
result is not good, but is satisfactory.
144
Conclusions
The decoupling and linearization control theory was
experimentally applied to two non-linear MIMO chemical control processes. The control experiments showed
that even if the processes are non-linear and interactive
and modelling errors are present, a satisfactory control
performance can be obtained.
I
600
lo0
I
1200
I00
1600
600
1200
I
1600
765-
2O
600
1200
Time
2O
Time
Figure 8
points. (-)
References
1 Isidori, A., Nonlinear Control Systems: An Introduction, 2nd edn,
Springer-Verlag, Berlin, 1989, p. 234
2 Sastry, S. and Bodson, M., Adaptive control: stability, convergence
and robustness, Prentice-Hall, New Jersey, 1989, p. 294
3 Kravaris, C. and Soroush, M. AIChE J. 1990,36,249
4 Paszkiewicz, D. P., IFAC Adaptive Control of Chemical Processes,
Cqpenhagen, Denmark, 1988, p. 75
5 Kravaris, C. and Chung, C.-B., AIChE J. 1987,33,592
6 Daoutidis, P. and Kravaris, C., AIChE J. 1989,35, 1602
7 Buchholt;F. and Kiimmel, M., Automafica 1979,15,665
g,
h
c
CP
KP
k
Lf
m
n
I
1600
[secl
Output responses for the step changes in the level and pH set
Measured values; (- - - -) set points
order of the reference model defined in Equation (I 1)
cross-sectional area of tank
temperature
resf3 time of external PI controller
input variable
new input variable
state variable
output variable
Greek letters
defined in Equation (6a)
defined in Equation (6b)
parameter
parameter
density
characteristic number
Nomenclature
matrix, defined in Equation (Sa)
vector, defined in Equation (5b)
function
function
function
concentration
specific heat
proportional gain of external PI controller
constant coefficient
Lie derivative defined in Equation (4)
number of input and output
number of state variable
I
1200
600
[secl
Figure 7 Output responses for the step changes in the level and pH set
points. (-) Measured values; (- - - -) set points
A
b
I
1800
Superscripts
ref
set
reference trajectory
set point
145