Professional Documents
Culture Documents
(
(
(
(
(
(
(
+ +
+ +
+
+
+ +
+
+
+
(
(
(
(
(
(
(
(
(
(
+ +
+
+ +
+
=
(
(
(
3
2
1
6 . 2 6 . 0
44 . 3 5 . 10
12
3
2
1
6 . 2 4 . 9 2 . 9
5 . 0 3 5 . 6
5 . 3 6 . 2
3
2
1
) 1 76 . 7 (
32 . 0
) 1 01 . 7 (
61 . 11
) 1 94 . 8 )( 1 29 . 7 (
) 1 62 . 19 ( 0032 . 0
) 1 9 . 6 (
53 . 0
) 1 63 . 14 )( 1 85 . 7 (
) 1 32 . 25 ( 011 . 0
) 1 2 . 6 (
1 . 0
) 1 29 . 7 )( 1 94 . 8 (
) 1 62 . 19 ( 0032 . 0
) 1 9 . 10 (
2 . 46
) 1 15 . 3 (
68 . 34
) 1 09 . 7 (
012 . 0
) 1 6 (
8 . 2
) 1 26 . 3 (
11 . 1
) 1 06 . 9 (
0049 . 0
) 1 64 . 8 (
61 . 0
) 1 7 . 6 (
66 . 0
d
d
d
S
e
S
e
S S
e S
S
e
S S
e S
S
e
u
u
u
S S
e S
S
e
S
e
S
e
S
e
S
e
S
e
S
e
S
e
Y
Y
Y
S S
S S
S S
S S S
S S S
S S S
MIMO model
FOPDT model
e
k
S
rs
rs rs
s s u
s Y
+
=
1 ) (
) (
(r=1,2R;s=1,2..S)
T= )) 5 . 0 , 1 (. (
rs rs
Max Min r=1,2..R,s=1,2..S)
P=N=
) 1 5 ( + +
T T
Max
rs rs
( r=1,2..R,s=1,2..S)
M=Max
) 1 ( + +
T T
rs rs
(r=1,2..R,s=1,2..S)
( )
=
(
(
+
|
|
.
|
\
|
|
|
.
|
\
|
+
|
|
.
|
\
|
+
|
|
.
|
\
|
+
=
R
r
rs rs rs rs
r
M
T T
P
T T
M
R
Q
1
2
2 1 2
2
3
1 1 1 500
s=123,S
Prediction horizon,P
Past Future
Set point(target)
Control horizon,M
k-1 k k+1 k+2 k+M-1 k+P
Sampling instant
y
u
u
Past output
Predicted future
Past control action
Future control action
112
Above equation given by (Ogunnaike and ray) [8] to
separate ethanol and water, where Y
1
is overhead ethanol mole
fraction, Y
2
side stream ethanol mole fraction and Y
3
is 19 tray
temperature (
0
C) (corresponding to bottoms composition). The
inputs are u
1
reflux flow rate (m
3
/s), u
2
side stream product
flow rate (m
3
/s), u
3
reboiler stream (kPa). The disturbance is d
1
feed flow rate (m
3
/s) and d
2
feed temperature (
0
C).[4]
When used DMC tuning strategy from table 1 then M=30,
P=90, T= 0.71
Fig 4: Manipulated Variables Plot without Default
Fig 4 shows the manipulated variable response taking M
control horizon, P predictive horizon, T sample time
according to given table. We find that there is more ringing
present in controller output, that affect on output response.
Fig 5: Response of 3x3 Processes without Default
Fig 5 shows the output (controlled variable) response for
33 process. We find there is more peaks over shoot occur in
output and hence output settle at long time.
So now trying to reduce ringing with the change of
parameter to T=1.5, M=5, P=10, N=60 by default.
Fig 6: Manipulated Variables Plot with Default
Fig 7: Response of 3x3 Processes with Default
Fig 6 shows the manipulated variable response in which
ringing is removed. Fig 7 shows output response for33
processes in which no peak overshoot occurs and system is
taken less time for settle compare to above.
Disturbance Rejection For 33 Process by using default
values:
Fig 8: Disturbance Rejection Plot For 33 Process by using default values
113
Fig 8 shows the disturbance rejection by using default
values. Using default value disturbance rejection taken place
earlier in 33 process.
B Dokukas and Luyben 44 Process:
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
+ +
+ + + + + +
+
+
+ + + +
=
(
(
(
(
(
4
3
2
1
191 . 1 48 . 0 52 . 0
42 . 0 59 . 0 68 . 0
2
72 . 8 24 . 2 71 . 0 60
33 . 27 59 . 1 75 . 7 39 . 3
4
3
2
1
54 . 15
) 1 19 . 12 (
67 . 11
) 1 90 . 6 )( 1 90 . 6 (
176 . 0
) 1 11 . 11 (
488 . 4
513 . 0
) 1 43 . 1 )( 1 43 . 1 (
38 . 2
) 1 14 . 7 )( 1 14 . 7 (
0204 . 0
) 1 38 . 2 )( 1 38 . 2 (
33 . 0
) 1 250 (
422 . 0
) 1 29 . 14 (
984 . 5
) 1 7 . 66 (
986 . 1
) 1 400 (
24 . 5
) 1 3 . 33 (
3711 . 2
) 1 36 . 11 (
811 . 9
) 1 2 . 22 )( 1 6 . 31 (
3746 . 0
) 1 74 . 21 )( 1 74 . 21 (
3 . 11
u
u
u
u
e
S
e
S S
e
S
e
e
S S
e
S S
e
S S
e
S
e
S
e
S
e
S
e
S
e
S
e
S S
e
S S
e
Y
Y
Y
Y
S
S S S
S
S S S
S S S S
S S S S
The above 44 process is presented by Doukas and
Luyben [9]. They studied the dynamic of a distillation column
producing a liquid side stream product. The controlled and
manipulated variables are Y
1
toluene impurity in the bottom),
Y
2
(toluene impurity in the distillate), Y
3
(benzene impurity in
the side stream), and Y
4
(xylem impurity in the side stream);
u
1
(side stream flow rat), u
2
(reflux ratio), u
3
(reboil duty), and
u
4
(side draw location).
Above process has eight FOPDTs, six SOPDTs, and two
dead times. The six SOPDT models are changed to FOPDTs
model in order to apply DMC tuning, DMC tuning find out
from table (1) is M=1087, T=0.5 and P=4370. The huge DMC
parameters are caused by the transfer function g24. MPC
controller gives an error statement using DMC tuning from
table (1).
Below MPC controller output at come at default setting
where T=0.1, M=5 P=10, N=60
0 10 20 30 40 50 60 70 80 90 100
-0.5
0
0.5
1
1.5
2
2.5
Outputs
Time
0 10 20 30 40 50 60 70 80 90 100
-25
-20
-15
-10
-5
0
5
Manipulated Variables
Time
Y1& u1
Y2 &u2
Y3 &u3
Y4 &u4
Fig 9: Response and Manipulated variables plot with default for 44 Process.
Figure 9 shows the manipulated and control variable
response for 44 process.
IV. CONCLUSION
A model predictive control is designed and implemented
for distillation column. It is found that at certain tuning
parameter the performance of control system is better than any
classical control system. in this paper we studied Ogunnaike
and ray(33) process and Doulas and Luyben(44) process.
It is found that by using MPC a overhead ethanol mole
fraction, side stream ethanol mole fraction and 19 tray
temperature (corresponding to bottoms composition), are
easily controllable in Ogunnaike and ray (33) process
compare to conventional control. Toluene impurity in the
bottom, toluene impurity in the distillate, benzene impurity in
the side stream and xylene impurity in the side stream in the
Doukas and Luyben (44) process are also controllable by
using MPC. MPC can be implemented for multivariable
processes like crystallization process.
REFERENCES
[1] Pradeep B. Deshpande, Charles A. Plank, Distillation Dynamics and
Control, Instrument Society of America, 1985.
[2] Dale E. Seborg, Thomas F. Edgar, Duncan A.Mellichamp, Process
Dynamics and control., 2
nd
Edition, Wiley & Sons (Asia) Pte.Ltd.
[3] Allgower, F., T. A. Badgwell, S. J. Qin, J. B. Rawlings, and S. J.
Wright, Nonlinear Predictive Control and Moving Horizon
Estimation-An Introductory Overview, Advances in control, P.M.
Frank (Ed), Springer, pp. 391., New York, 1999.
[4] Qin, S. J., and T. A. Badgwell, A Survey of Industrial Model
Predictive Control Technology, control Technology, Control Eng.
Practice, vol. 11, pp. 733-746, 2003.
[5] Sridhar, R.; Cooper, D. J., A Tuning Strategy for Unconstrained
Multivariable Model Predictive control, Ind. Eng. Chem. Res. 1998,
37, pp. 4003-4016.
[6] Rahul Shridhar, Douglas J. Cooper, A novel tuning strategy for
multivariable predictive control, ISA Transactions, Vol. 36, No 4 pp,
273-280, 1998.
[7] V. T. Minh, W. Mansor W. Muhamed, Model Predictive Control of
a Condensate Distillation Column, IJSC, vol. 1, pp. 4-12, 2010.
[8] Ogunnaike, et al. (1983). Advanced Multivariable Control of a Pilot-
Plant Distillation Column. AlChE Journal (29/4): 632-640
[9] Doukas, N. And Luyben, W. L. Control of Side stream Columns
Separating ternary mixtures. Instrumentation Technology, (1978), 25,
43-48.
[10] Rakesh Kumar Mishra, Rohit Khalkho,Brajesh Kumar, Tarun Kumar
Dan, Effect of Tuning Parameters of a Model Predictive Binary
Distillation Column IEEE International Conference on Emerging
Trends in Computing, Communication and Nanotechnology
2013.(reference)
114