1. No category

International Conference on Chemical, Civil and Environmental Engineering (ICCCEE'2015) March 23-24, 2015 London (UK)
Design of Control Structure for Binary
Distillation Column
Ramsagar Vooradi, and Sarath Babu Anne

guidelines for selecting suitable controller variables and
corresponding manipulated variables for distillation columns.
Lot of work has been carried out on the control of distillation
columns, but the study on the response of the process subjected
to various disturbances with different control structures has not
been attempted. Luyben [2] presented steady state design of the
distillation column and carried out dynamic simulation
focusing on tuning of controllers at default ranges and the
controller parameters were calculated using Tyreus–Luyben
Tuning rule.
In this paper, the performance of the binary distillation
column has been studied with different control structures.
Controller parameters have been calculated by using Tyreus–
Luyben and Ziegler-Nicholas tuning rules. The column
response has been observed subjecting it to various
disturbances.
Abstract----Design of control structure for distillation columns is
an important aspect in order to minimize the damage caused due to
unexpected disturbances. In this study, the dynamic simulation of a
distillation column has been carried out using Aspen Dynamics. For
the purpose of the present study, a binary mixture of propaneisobutane is selected. The distillation column has been first designed
at steady state using DSTWU model, a shortcut design model in
Aspen Plus. The design is fine-tuned and the column is sized with
RADFRAC model, a rigorous rating model using the design data
obtained from the DSTWU model. The steady state simulation has
been exported to Aspen Dynamics under pressure driven simulation
mode for carrying out dynamic simulation. The steady-state design
parameters of the column are used as initial conditions for the
dynamic simulation. The Dynamic simulation of the distillation
column has been carried with three different processes, which are
classified based on the control structure used as: Process-1 controls
the composition of top product using the temperature controller;
Process-2 controls the composition of top product using a
composition controller and Process-3 controls the composition using
a cascade of temperature and composition controllers. The tuning of
different controllers has been performed using the Relay AutoTuning test, and the control parameters have been calculated using
Tyreus-Luyben and Ziegler-Nicholas tuning rules. The dynamic
response of the three processes mentioned above has been studied
subjecting them to various disturbances. The major conclusions
drawn from this study are: The temperature controller alone (ProcessI) is not suitable to control both top and bottom product
compositions, when the step disturbance in feed composition is
given. Though the composition controller (Process-II) is able to
maintain the desired compositions of the top and bottom products, the
response of the system is very poor. The response has been observed
to be fast and without any oscillations in case of Process-III when
compared to both Process-I and Process-II.
II. PROBLEM DEFINITION
In the present study dynamic simulation of a distillation
column has been carried out using Aspen Dynamics to develop
different control strategies.
The distillation column is used to separate the binary mixture
of propane and isobutene feed with the following specifications
[2]:
Feed: flow rate is 1 mol/s, temperature is 322 K, pressure is 21
atm, composition:- 40 mol% propane and 60 mol% isobutane.
Product Specifications: Isobutane should not exceed 2 mol% in
the distillate and propane should not exceed 1 mol% in the
bottoms.
III. METHODOLOGY
Keywords----Aspen, Control structure, Distillation Column,
Tyreus-Luyben, Ziegler-Nicholas.
I. INTRODUCTION
S
EVERAL aspects of operation and design of the
depropanizer process have been discussed at length but
none addresses the issue of control structure design in a
systematic manner. Choosing a proper control structure for
distillation columns along with optimum controller parameters
is important to maintain desired product quality, production
rate, and minimize the energy usage. Kister [1] suggested
A. Steady State Process Simulation
Steady state simulation of the column is first carried out by
DSTWU model, a short cut design model followed by
RADFRAC model, a rigorous rating model. In order to carry
out “simultaneous design” (steady state and dynamic), pumps
and control valves must be included to permit a “pressuredriven” dynamic simulation. The necessary pumps and control
valves are added to the flowsheet as shown in Fig. 1.
OUTPUMP1
http://dx.doi.org/10.15242/IIE.E0315071
133
TOPPROD
TOP
V1
IPVAL
Ramsagar Vooradi is with National Institute of Technology, Department of
Chemical Engineering,
Warangal–506004,
AP,
INDIA.
(phone:
+918332969406; e-mail: [email protected]).
Sarath Babu Anne, is with National Institute of Technology, Department of
Chemical Engineering,
Warangal–506004,
AP,
INDIA.
(phone:
+919492273973; e-mail: [email protected]).
V2
PUMP1
COL
IPCOL
PUMP2
V3
OUTPUMP2
BOTPROD
BOT
Fig. 1 Steady State Simulation Flowsheet
International Conference on Chemical, Civil and Environmental Engineering (ICCCEE'2015) March 23-24, 2015 London (UK)
As the data from shortcut method is used for the rigorous
RADFRAC column, the results of the simulation did not give
the required purities. In order to achieve the required purities,
using the RADFRAC column, design specifications are used to
fine tune some of the input parameters. Two design
specifications, one each for the top and bottom product purities
and the manipulated variables chosen are: reflux ratio and
distillate rate respectively.
B. Dynamic Simulation
In order to convert the steady-state simulation into a
dynamic mode, the additional information that must be
provided is the physical dimensions of all the equipment.
The diameter of the column has been calculated by carrying
out tray sizing. The sizes of the reflux drum and the column
base have been calculated using simple heuristics. The
dimensions of the reflux drum used are 4.08 m in diameter and
8.16 m in length, while that of the column base are 5.08 m in
diameter and 10.16 m in length.
When the steady state is converted into dynamic mode only
one default controller, the pressure controller is added which is
configured to measure condenser pressure by manipulating
condenser heat removal.
The following four additional controllers have been added to
achieve effective operation of the column:[3]
1. Reflux drum-level controller
2. Base-level controller
3. Feedflow controller
4. Tray temperature or composition controller or cascade
controller
The selection of the manipulated variables to control each of
the above four measured variables is the issue while selecting
the control structure.
The details of all the four controllers used are tabulated as
shown in the following table.
TABLE I
CONTROLLER SPECIFICATION TABLE
Type of controller
Controlled
Manipulated
variable
variable
Reflux drum-level
(lc12)
Base-level (lc11)
Reflux
drum-level
Base-level
Feed flow
Feed flow
rate
9th stage
temperature
Distillate
composition
Distillate
composition
Temperature
Composition
Primary
Cascade
Secondary
th
9 stage
temperature
Fig. 2 Dynamic Simulation Flowsheet for Process-I
Fig. 3 Dynamic Simulation Flowsheet for Process-II
Fig. 4 Dynamic Simulation Flowsheet for Process-III
Dead
time(min)
% opening of
valve v2
% opening of
valve v3
% opening of
valve v1
Reboiler duty
0
Reboiler duty
3
Setpoint of
Secondary
controller
Reboiler duty
3
TABLE II
LIST OF CONTROLLERS IN THE PROCESS
Process
Controllers used
Process-I
Pressure, Reflux drum-level, Baselevel, Feed flow, Temperature.
Process-II
Pressure, Reflux drum-level, Baselevel, Feed flow, Composition.
Process-III
Pressure, Reflux drum-level, Baselevel, Feed flow, Cascade.
0
0
1
IV. RESULTS AND DISCUSSION
1
C. Reflux to feed ratio
The R/F ratio scheme should do a pretty good job of
maintaining product purities in the propane/iso-butane system
in case of feed composition disturbances and feed flow rate
changes. The R/F structure is implemented by using a
multiplier block. The input of this block is the mass flow rate
of the feed, while its output is the mass flow rate of the reflux.
http://dx.doi.org/10.15242/IIE.E0315071
D. Selection of Control Tray for Process-I
The appropriate location for the temperature or composition
sensor is usually chosen based on any one method: slope
criterion, sensitivity criterion etc. in this study 9th stage has
been selected as the controller tray based on the slope criterion.
The column has been studied with three different control
structures which are named as process-I, II and III as shown in
Figures 2, 3 and 4. The list of controllers used in these three
processes are given in Table-2.
A. Steady State Results
The Steady State simulation results with RADFRAC column
obtained without the use of any design specifications are
shown in Table-3. It can be observed that the top and bottom
product purities obtained are not upto the desired requirement.
The desired purities with respect to both top and bottom
products can be achieved by using two design specifications,
one each for the top and bottom product purities and the
manipulated variables chosen are: reflux ratio and distillate rate
respectively.
134
International Conference on Chemical, Civil and Environmental Engineering (ICCCEE'2015) March 23-24, 2015 London (UK)
The results of the two design specifications are shown in
B. Dynamic Simulation Results
Tables 4 and 5 respectively. It can be observed that the value of
The acceptable ranges of top and bottom product
reflux ratio and distillate rate are 3.51 and 0.402 respectively, compositions for all processes are shown in Table 7.
to get the desired top and bottom product purities.
TABLE VII
ACCEPTABLE RANGE OF PRODUCTS
Product composition
Design
Acceptable range
(Mole Fraction)
specification
(within +5% of its final
value)
DISTILLATE (ISO0.02
0.019 - 0.021
BUTANE)
BOTTOMS
0.01
0.0095 - 0.0105
(PROPANE)
TABLE III
DETAILED DESIGN STREAM RESULTS
1) Tuning of Controllers
The ultimate gain and ultimate period have been found by
conducting Relay Auto-Tuning test with the default ranges and
the Controller parameters have been calculated using TyreusLuyben (T-L) and Ziegler-Nicholas (Z-N) tuning rules. The
results of all the controllers are shown in Table 8.
TABLE IV
DESIGN SPECIFICATION-1 RESULTS
TABLE VIII
CONTROLLER PARAMETER RESULTS
TYREUS-LUYBEN
Kc
1.572
τI
11.22
ZIEGLERNICHOLS
KC
τI
2.2640
4.25
0.11
86.064
0.158
32.6
1.572
11.22
2.264
4.25
0.073
49.236
0.106
18.6
PROCESS
PROCESS-I
PROCESS-II
TABLE V
DESIGN SPECIFICATION-2 RESULTS
PROCESSIII
TEMPERATURE
CONTROLLER
COMPOSITIONCONTROLLER
TEMPERATURE
CONTROLLER
COMPOSITIONCONTROLLER
The Ziegler–Nichols settings work well for a small range of
dead time to time constant ratio D/τ, and the performance starts
to degrade for D/τ <0.2 and D/τ >2. A more conservative
tuning rule is proposed by Tyreus and Luyben which works
well for the time-constant dominant processes[4].
2) Reflux to Feed Ratio
The R/F structure is implemented by using a multiplier block.
With the help of multiplier block the reflux flow rate is
maintained constant irrespective of the disturbance but the
The diameter of the column has been obtained as 5.94 m reflux flow rate will change when feed composition or feed
from the tray sizing results as shown in the Table 6.
flow rate changes.
TABLE VI
TRAY SIZING RESULTS
http://dx.doi.org/10.15242/IIE.E0315071
C. Performance Evaluation
1) Feed Flow Controller Set Point Change for Process-I
Process-I is subjected to a 0.05 kmol/s of step change to the
set point of flow controller and its response is observed when
reflux drum level controller has the following set of controller
parameters:
Case-1: Controller gain = 2 and integral time=9999.
Case-2: Controller gain = 1and integral time=20.
135
International Conference on Chemical, Civil and Environmental Engineering (ICCCEE'2015) March 23-24, 2015 London (UK)
Fig. 8 Response of Process-II
Fig. 5 Response of Process-I for Case-1
Fig. 9 Response of Process-III with T-L tuning
Fig. 6 Response of Process-I for Case-2
The process has been observed to show a good response for
Case-1 conditions, i.e., high gain and low integral action
(proportional action). From Fig. 5 it has been observed that the
process exhibits less oscillatory behavior and response time of
process for Case-1 is 8.5 hours. However, the use of
proportional controller has been observed to result in some offset.
With respect to Case-2, it has been observed that the process
exhibits high oscillatory behavior and the response time of
process is 17.5 hours as shown in Fig. 6. This is essentially as a
result of the high integral time. The level controller (LC12) in
this case did not result in any off-set. But the response time is
much more than that of Case-1.
From this observation it can be concluded that level
controllers can be configured with proportional action if the
off-set is tolerable. In this work the two level controllers have
been configured with proportional action in all the processes.
2) Composition Disturbance
The performance of all the three Processes has been studied
by subjecting them to a step disturbance in feed composition.
The mole fractions of propane and iso-butane have been
changed from 0.40/0.60 to 0.50/0.50. The responses of
different processes are shown in Figures 7, 8, 9 and 10.
In case of Process-II, the product compositions have been
maintained at the desired level unlike the case of Process-I.
However, the response of Process-II has been found to be very
poor, when compare with that of Process-I. This is because of
very low controller gain and the very large integral time. Hence
it can be concluded that Process-II alone may not be suitable
for obtaining a fast response and tight controller action.
Response time of process-II is 9 hours.
Fig. 10 Response of Process-III with Z-N tuning
Process-III, as defined earlier, makes use of a cascade of
temperature and composition controllers as primary and
secondary controllers respectively, has been observed to give
fast response and tight controller action.
It can be clearly observed that the responses of Process–III
with T-L and Z-N tuning parameters are fast when compared to
the responses of both Process-I and Process-II. The Response
time of Process-III with T-L tuning parameters is less and
provides tight controller action even though the value of
controller gain is small when compared with Process-II. This is
as a result of cascading the controllers.
Response time of Process-III with Z-N tuning parameters is
less when compared with Process-III with T-L tuning
parameters because of high controller gain and integral action.
V. CONCLUSIONS

Fig. 7 Response of Process-I
http://dx.doi.org/10.15242/IIE.E0315071
136
Response of level controllers is good with proportional
action. The process becomes oscillatory with the
introduction of integral action.
International Conference on Chemical, Civil and Environmental Engineering (ICCCEE'2015) March 23-24, 2015 London (UK)





In case of disturbance in feed composition, the
temperature controller alone (Process-I) is not suitable to
control top and bottom product compositions.
A single Composition controller (Process-II) is able to
maintain the desired compositions of the top and bottom
products even though there is a disturbance in feed
composition, but the response of the system is very poor.
The response of Process-I is fast to that of Process-II for
all types of disturbances but in case feed composition
disturbance the response of Process-I resulted in an off-set.
Cascade controller (Process-III) is suitable for controlling
the feed composition disturbances. The response of
Process-III is fast without oscillatory behavior when
compared to both Process-I and Process-II.
The response of Process-III with Z-N tuning parameters is
better than that of Process-III with T-L tuning parameters.
REFERENCES
[1]
[2]
[3]
[4]
Henry Z. Kister. Distillation Operation, McGraw – Hill Education, 1990.
William L. Luyben, Distillation Design and Control, 2nd edition, Wiley,
2013.
Seborg, et al., Process Dynamics and Control, Wiley & Sons, Inc. 2006.
Cheng-Ching Yu, Autotuning of PID Controllers A Relay Feedback
Approach, Springer Science & Business Media, 2006.
http://dx.doi.org/10.15242/IIE.E0315071
137