Proceedings of the 9th WSEAS International Conference on SIGNAL PROCESSING, COMPUTATIONAL GEOMETRY and ARTIFICIAL VISION

Control of Binary Distillation Column Using Fuzzy PI Controllers

SHAHRAM JAVADI * JABBER HOSSEINI

[email protected] [email protected]

Islamic Azad University - Central Tehran Branch

Electrical Engineering Department
* Assistant Professor in Electrical Engineering Department.

IRAN

Abstract: In this paper the automatic control of a binary distillation column is described. This control is
done with fuzzy logic controllers. After a short explanation of the function and dynamic of a binary
distillation column, it‫ۥ‬s operating and control strategies are described. then all the processes of the
distillation column with it‫ۥ‬s fuzzy controllers are simulated in MATLAB software as the results are shown
at the end.

Keywords: Binary Distillation Column, Fuzzy Inference System, Simulation.

1. Introduction perform this task usually involve a model of

Distillation columns are used in the the process, and such a model is not easy to
chemical industry to separate liquid obtain, since distillation columns are
mixtures into their pure components or into complex, highly nonlinear and multivariable
several fractions of predefined composition. processes. Usually operators‫ ۥ‬knowledge
The two major types of classical distillation plays an important role in complex industrial
include continuous distillation and batch processes, and this knowledge may be
distillation. Continuous distillation, as the integrated in the automatization system by
name says, continuously takes a feed and means of fuzzy inference system (FISs).
separates it into two or more products. Batch FISs allow obtaining control actions from
distillation takes one batch at a time for the input variables by using linguistically
feed and splits it into products by selectivity formulated rules; also, FISs don‫ۥ‬t need an
removing the more volatile fractions over analytic model of the process, so their
time. Keeping a constant product quality in a conception is greatly simplified. The
distillation column is very important. A experiments described in this paper have
better control means a smaller variability of been developed and tested with the aid of a
quality around the set point an operation simulator written in MATLAB/Simulink,
closer to specifications and lower energy which is able to reproduce the dynamic
consumption. Traditional techniques to behavior of the column.

ISSN: 1790-5109 145 ISBN: 978-960-474-108-3

Proceedings of the 9th WSEAS International Conference on SIGNAL PROCESSING, COMPUTATIONAL GEOMETRY and ARTIFICIAL VISION

2. The Distillation Column

Distillation column is probably the most temperature and pressure as variables [1].
popular and important process studied in the Overhead vapor leaving top tray is
chemical engineering literature. Distillation totally condensed for negligible liquid
is used in many chemical processes for holdup with condensate flowing to a reflux
separating feed streams and for the drum having constant and perfectly mixed
purification of final and intermediate molar liquid hold up .
product streams. Most column handle
multicomponent feeds, but many can be
approximate by binary or pseudobinary
mixtures. The objective is to split a liquid
two component mixture into its fraction
throughout stripping and rectifying
processes. As it is shown in figure 1 the
mixture is fed onto the feed tray with
feed flow rate of F (lb.mol/min) and
composition of z (refered to more volatile
component). The more volatile component
(water) leaves the top of the column while
the less (methanol) leaves the bottom. The
overhead vapor is totally condensed in a
condenser and flows into the reflux drum
with a composition and holdup liquid of
(moles), and reflux is pumped back to
the top tray of the column at a rate R, while
overhead distillate product is removed at a
rate D. At the base of column, a liquid
product stream is removed at a rate B with a
composition and holdup of (moles).
Figure 1: Binary Distillation Column
Vapor boil up is generated in a reboiler at a
rate V. The column contains trays and
the liquid holdup on each tray n is with The reflux rate is varied by a fuzzy
composition , while the holdup of vapor is proportional-integral (PI) feedback
assumed to be negligible throughout the controller to control distillate composition
system. for a set point of 0.98 for the mole fraction
of the light component. Holdup of reflux
3. Control Strategy in the line leading back to the top tray is
Consider the closed-loop response during neglected. Under dynamic conditions,
the dynamic distillation of an ideal binary may not equal .
mixture in the column shown in figure 1, At the bottom of the column, a liquid sump
under two assumption of constant relative of constant and perfectly mixed molar liquid
volatility ( ) at a value of 2.0 and constant holdup is provided. A portion of liquid
molar vapor flow for a saturated liquid feed flowing from this sump passes to a
to tray . Following the development by thermosiphon reboiler, with the remainder
taken as bottoms product at a molar flow
Luyben it is not necessary to include energy-
rate B. vapor boilup generated in the reboiler
balance equations for each tray or to treat
is varied by a fuzzy PI feedback controller to

ISSN: 1790-5109 146 ISBN: 978-960-474-108-3

Proceedings of the 9th WSEAS International Conference on SIGNAL PROCESSING, COMPUTATIONAL GEOMETRY and ARTIFICIAL VISION

control bottom‫ۥ‬s composition with a set 4. Fuzzy PI controller

point of 0.02 for mole fraction of the The structure of the fuzzy PI controller
light component. Liquid holdup in the which has two inputs and one output is
reboiler and lines leading from the sump are shown in figure 2. The inputs are the
assumed to be negligible. The composition classical error (e) and the rate of the change
of the boilup is assumed to be in of error ( or de) that are defined in the
equilibrium with .The liquid holdup equations 12,13 and the outputs of the
on each of the equilibrium trays is controllers are the reflux and vapor boilup
assumed to be perfectly mixed but will vary generated in the reboiler.
as liquid rates leaving the trays vary. Vapor
holdup is assumed to be negligible
everywhere. Tray molar rates V vary with
time but at any instant in time are
everywhere equal. The dynamic material-
balance and phase equilibrium equations
corresponding to this description are as Figure 2: Fuzzy PI controller
follow:
All trays, n: Triangular membership functions are used
for both input and output variables that the
d = + - (1) input and output values are represented
linguistically. (i.e., N=Negative, Z=Zero and
= + + - - P=Positive) as it is shown in figure 3.
(2)
= +( + )/β (3)

= (4)

Where is nonzero only for tray Ns, y and

x Refer to the light component only such
that the corresponding mole fractions for the
heavy component are (1- y) and (1- x),
and are the initial steady-state values,
and β is a constant that depends on tray
hydraulic. Figure 3: Membership functions for
For the condenser-reflux-drum combination: two inputs (e & de) and output (Reflux or
D=V– (5) Vapor boilup).
The fuzzy controllers rule base composed of
( / )= (6) 9(3×3) rules as shown in tables 1 and 2.
For the reboiler:
B= -V (7)
( / )= (8)

= (9)

ISSN: 1790-5109 147 ISBN: 978-960-474-108-3

Proceedings of the 9th WSEAS International Conference on SIGNAL PROCESSING, COMPUTATIONAL GEOMETRY and ARTIFICIAL VISION

Table 1: FAM Table for 2 inputs (e & de) is the error or deviation from the set point as
and output (Reflux) that refers to . given by:
(12)
(13)
Typical results for the initial steady-state
conditions, fixed conditions, hydraulic
parameters, and disturbance are listed in
table 3.
Table 3: Initial and Fixed Conditions,
Hydraulic Parameter and Disturbance
Other Initial Initial Liquid-Phase
Table 2: FAM Table for 2 inputs (e & de)
Conditions Composition
and output (vapor boilup) that e refers to
=100 Tray
(lb.mol/min) Bottom 0.02
=0.5 1 0.035
2 0.05719
3 0.08885
=128.01(lb.mol/m 4 0.1318
in) 5 0.18622
=178.01 6 0.24951
The input/output mapping of fuzzy (lb.mol/min) 7 0.31618
controllers is also given in figure 4. 8 0.37948
9 0.43391
=50
10 0.47688
(lb.mol/min)
11 0.51526
=50
12 0.56295
(lb.mol/min)

=0.98
=0.02

=10
Figure 4: Crisp input/output map lb.mol
Also we use the Mamdani inference system Fixed Conditions 13 0.61896
as inference engine and centroid method for =20 14 0.68052
defuzzification in the fuzzy controllers. =10 15 0.74345
The two PI controller equations are: =100 lb.mol 16 0.80319
=100 lb.mol 17 0.85603
V= - ( + ) (10)
Hydraulic 18 0.89995
Parameter
= + ( + ) (11)
β=0.1 min 19 0.93458
Where and are initial values, Disturbance at 20 0.96079
and respectively feedback controller gain t=
and feedback reset time for integral action, E = 0.55 Distillate 0.98

ISSN: 1790-5109 148 ISBN: 978-960-474-108-3

Proceedings of the 9th WSEAS International Conference on SIGNAL PROCESSING, COMPUTATIONAL GEOMETRY and ARTIFICIAL VISION

5. Simulation
In this paper two case studies have been
studied. In both simulations, it is used
simulink toolbox of MATLAB software. In
the first case the all processes of a binary
distillation column (methanol/water) is
simulated and in the second, according to
the control strategy, the fuzzy PI controllers
are designed and simulated. For simulation
two external inputs are forced. First a step
with the step time of 70s and final value of
10 is used as feed rate and also a step with Figure 7: Top: Top Composition ( ) ,
step time of zero and final value of 0.05 is Bottom: Bottom Composition ( )
used as disturbance. Note that the initial (Number of epochs = 150)
values of feed rate and disturbance
respectively are 100 and 0.5.

Figure 8: Mole Fraction of the Trays 1 to 20

(Number of epochs = 100)
Figure 5: External Inputs: Top: Feed Rate,
Bottom: Feed Composition (Disturbance))

Figure 9: Mole Fraction of the Trays 1 to 20

(Number of epochs = 150)

Figure 6 : Top: Top Composition ( ) ,

Bottom :Bottom Composition ( )
(Number of epochs = 100)

ISSN: 1790-5109 149 ISBN: 978-960-474-108-3

Proceedings of the 9th WSEAS International Conference on SIGNAL PROCESSING, COMPUTATIONAL GEOMETRY and ARTIFICIAL VISION

Figure 10: Top: Distillate Rate, Middle: Figure 11: Top: Distillate Rate, Middle:
Bottom Rate, Bottom: Total Rate (Distillate Bottom Rate, Bottom: Total Rate (Distillate
Rate + Bottom Rate) Rate + Bottom Rate)
(Number of epochs = 100) (Number of epochs = 150)

6. Conclusion
We have presented the control system References:
design of a binary distillation column under [1] W.L Luyben, “Process Modeling,
disturbance. Control of the top and bottom Simulation and Control for Chemical
compositions of the column is a difficult Engineering”, McGraw-Hill, 1990
task due to presence of the control loop
interactions and nonlinearities. The structure [2] Warren.L McCabe, Julian C.Smith, Peter
allows taking into account dynamic Harriot, “Unit Operation of Chemical
variations of the process and adapting the Engineering”, 1993
controller parameters to this various
conditions. Fuzzy controllers achieved a [3] Robert Heinlein, “Fuzzy Control
accurate performance in controlling the top System”, 1973
and bottom compositions and also in
controlling the feed rate, top and bottom [4] Driankov D, Helerdoon H and Reinfrank
rates. As it is shown in simulation results M, “An Introduction to Fuzzy Control”,
(figures 6 and 7) the defined errors are more Springer-verlag Berlin, 1993
close to zero and the system operates
approximately around it‫ۥ‬s set points (0.02 [5] Passino K.M, Yurkovics, “Fuzzy
and 0.98). Also according to the figures 8 control” Adisson-Wesley, New York, SA,
and 9 we can see that the mole fraction on 1998
the trays 1 and 20 respectively close to 0.02
and 0.98 and for the other trays (2 to 19) the [6] Wang L.X, “A Course in fuzzy System
mole fraction varies from 0.02 to 0.98. and Control”, Prentice Hall, New York,
Finally as it is shown in figures 10 and 11 USA, 1997
the rates of the outputs (Bottom and Top
products) is equal to the feed rate (input) as [7] Yager R.R, Filev D.P “Essentials of
we expected. Fuzzy Modelling and Control”, John Wiley,
New York, 1994
7. Acknowledgment
The authors thank the National Iranian Oil [8] Abonyi J, “Fuzzy Model Identification”,
Company (NIOC) and Iranian Offshore Oil Birkhauser Boston, MA, USA, 2003
Company (IOOC) for their collaborations.

ISSN: 1790-5109 150 ISBN: 978-960-474-108-3

Proceedings of the 9th WSEAS International Conference on SIGNAL PROCESSING, COMPUTATIONAL GEOMETRY and ARTIFICIAL VISION

[9] L.Jouffe, “Fuzzy Inference System

Learning by Reinforcement Methods, 1996

Figure 12: A view of complete system

Figure 13: Binary Distillation Column

ISSN: 1790-5109 151 ISBN: 978-960-474-108-3

Proceedings of the 9th WSEAS International Conference on SIGNAL PROCESSING, COMPUTATIONAL GEOMETRY and ARTIFICIAL VISION

Figure 14: A sample of designed trays (trays 11-15)

Appendix

Nomenclature:
= Liquid Holdup in the Reflux Drum = Initial Value of Distillate Rate
= Liquid holdup in the Column Base = Initial Value of Bottom Rate
= Liquid Holdup on Tray = Initial Value of Reboiler Rate
= Liquid Composition on the Tray
= Initial Value of Feed Rate
= Vapor Composition on the Tray = Initial Value of Feed Composition
= Internal Reflux Rate on the Tray = Number of Trays
D= Distillate Rate = Feed Tray (10th Tray)
B= Bottom Rate = Relative Volatility
V= Reboiler Rate
= Parameter of the Relationship between
L= Reflux Rate
Liquid Holdup and Liquid Rate on each
F= Feed Rate
Tray.
= Feed Composition
= Top Composition
= Bottom Composition
= Initial Value of Reflux Rate

ISSN: 1790-5109 152 ISBN: 978-960-474-108-3

12/31/13 Control of binary distillation column using fuzzy PI controllers

SIGN IN SIGN UP

Control of binary distillation column using fuzzy PI controllers Tools and Resources

Save to Binder
Authors: Shahram Javadi Islamic Azad University, Electrical Engineering Department,
Iran
Export Formats:
Jabber Hosseini Islamic Azad University, Electrical Engineering Department, 2009 Article
BibTeX EndNote ACM Ref
Iran

Published in: Bibliometrics Share:

· Proceeding · Dow nloads (6 Weeks): n/a Share
Share
Share
Share
Share
ShareMore
|
ISCGAV'09 Proceedings of the 9th WSEAS international conference on Signal · Dow nloads (12 Months): n/a
processing, computational geometry and artificial vision · Dow nloads (cumulative): n/a Tags: binary distillation column
· Citation Count: 0 fuzzy inference system simulation
Pages 145-152
World Scientific and Engineering Academy and Society (WSEAS) Stevens Point, Wisconsin, USA ©2009
table of contents ISBN: 978-960-474-108-3

Feedback | Switch to single page view (no tabs)

Abstract Authors References Cited By Index Terms Publication Reviews Comments Table of Contents

In this paper the automatic control of a binary distillation column is described. This control is done with fuzzy logic controllers. After a
short explanation of the function and dynamic of a binary distillation column, it's operating and control strategies are described. then all
the processes of the distillation column with it's fuzzy controllers are simulated in MATLAB software as the results are shown at the end.

Powered by

The ACM Digital Library is published by the Association for Computing Machinery. Copyright © 2013 ACM, Inc.
Terms of Usage Privacy Policy Code of Ethics Contact Us

Useful downloads: Adobe Acrobat QuickTime Windows Media Player Real Player

dl.acm.org/citation.cfm?id=1627559 1/2