Modeling, Simulation and Control of a

multicomponent
Batch Distillation Processes
Thesis submitted in partial fulfillment of the requirements for the award of
the degree of
Master of Technology

in

Chemical Engineering

by

P.V.Radha Krishna Adari

(06CH6027)

under the guidance

of

Prof. Amiya K. Jana

DEPARTMENT OF CHEMICAL ENGINEERING

INDIAN INSTITUTE OF TECHNOLOGY
KHARAGPUR, INDIA
2008
 Department of Chemical Engineering

Indian Institute of Technology

Kharagpur-721302

CERTIFICATE
This is to certify that the thesis entitled “Modeling, Simulation and
Control of a multicomponent Batch Distillation Processes” submitted by
P.V.Radha Krishna Adari (Roll no. 06CH6027), in the partial fulfillment for the
requirement for the award of degree of Master of Technology in Chemical
Engineering of Indian Institute of Technology, Kharagpur during the academic
session 2007-2008 is a bonafide record of the project work carried out by him under my
supervision and guidance. The approval does not necessarily endorse or accept every
statement made, opinion expressed or conclusion drawn as recorded in the thesis. It only
signifies the acceptance of the thesis for the purpose for which it is submitted.

____________________
Dr. Amiya K. Jana
Date: _______________ Department of Chemical Engineering
Place: IIT Kharagpur Indian Institute of Technology
Kharagpur-721302
 ACKNOWLEDGMENTS

I would like to express my sincere gratitude to Prof. Amiya K. Jana for his
guidance and assistance in this project. The technical discussions with Prof. Amiya K.
Jana were always been very insightful, and I will always be indebted to him for all the
knowledge he shared with me. His prompt responses and availability despite his
constantly busy schedule were truly appreciated. The reality is that Prof. Amiya K. Jana
was much more than an advisor for me. He always helped me in all the technical and
non-technical issues during the production of this work. His encouragement and efforts
led this project to successful completion in a timely fashion.
I would like to extend my sincere thanks to my classmates. Finally, I express my
deep sincere thanks to my Parents who motivated and encouraged me for higher studies,
without which wouldn’t have been possible.

P.V.Radha Krishna Adari

 International Journal Publications

Adari, P. V. Radha Krishna and Jana, Amiya K. (2008) "Comparative

Control Study of a High-Purity Ternary Batch Distillation," Chemical Product
and Process Modeling: Vol. 3: Iss. 1, Article 26.
Avialable at http://www.bepress.com/cppm/vol3/iss1/26

P.V. Radha Krishna Adari and Amiya K.Jana (2009) “Nonlinear Adaptive
Control of a Batch Reactive Distillation Column,” Chemical Engineering
Journal: Vol 150: Iss. 2-3, pages 516-526.
Available at doi:10.1016/j.cej.2009.03.015
 CONTENTS
Page. No

List of Tables………………………………………………………. iii

List of Figures……………………………………………………... iv

Abstract ………………………………………………………… vi

Chapter 1: Introduction……………………………………………. 1

Chapter 2: Literature review………………………………………. 5

Chapter 3: Modeling, Simulation and Control of a multicomponent

Batch Distillation Column

3.1. Process Description……………………………………. 8

3.2. Assumptions…………………………………………… 10

3.3. Modeling Equations ……………………….…………... 11

3.4. Simulation Algorithm………………………………….. 18

3.5. Synthesis of Control Scheme…………………………... 19

3.6. Simulation Results and Discussion…………....……….. 20

Chapter 4: Modeling, Simulation and Control of a multicomponent

Reactive Batch Distillation Column

4.1. Process Description……………………………………. 34

4.2. Assumptions…………………………………………… 35

4.3. Modeling Equations ………………………….….……. 35

4.4. Simulation Algorithm………………………………….. 40

i
 4.5. Synthesis of nonlinear Control scheme……………….. 41

4.6. Simulation Results and Discussion……………………. 48

Chapter5: Conclusions and Scope of future work………………….. 63

Notation…………………………………………………………….. 64

References…………………………………………………………... 65

Appendix……………………………………………………………. 67

ii
List of Tables

Table page

1. Batch distillation column specifications …………………………. 17

2. Tuning parameters and ISE values for servo performance with respect
to component (XD,1)……....…………..…………………...……… 31
3. Tuning parameters and ISE values for servo performance with respect
to component (XD,2)…………………………………………….….. 31
4. Tuning parameters and ISE values for regulatory performance with
respect to component (XD,1) (for heat input)..…………………....... 32
5. Tuning parameters and ISE values for regulatory performance with
respect to component (XD,2) (for heat input )..…………...…..…….. 32
6. Tuning parameters and ISE values for regulatory performance with
respect to component (XD,1) (for XB )..….……………..……….........33
7. Tuning parameters and ISE values for regulatory performance with
respect to component (XD,2) (for XB )..………….…………..……... 33
B

8. Reactive batch distillation column specifications...………..………. 39

9. Tuning parameters for ASE 1 and ASE 2 ……………..….….……..62
10. GSPI and GMC controller tuning parameters….....……….………..62
11. ISE values for comparison of ASE 1 and ASE 2 …………..…...….62
12. ISE values for comparison of GSPI and GMC ..………..….……….62

iii
List of Figures

Figure page
1. Schematic diagram of multicomponent batch distillation process...…… 18
2. Bubble point calculations…………………………………………….... 23
3. Open-loop process dynamics at stat-up phase ……….……………….. 31
4. Open-loop process dynamics at production phase………………….….. 32
5a.Comparative closed loop performance of PI, NLPI and GSPI control
algorithms with a set point change in XD,1.............................................. 34
5b.Comparative closed loop performance of PI, NLPI and GSPI control
algorithms with a set point change in XD,2............................................. 35
5c.Third component in the bottom after second slop cut ............................ 36
6a.Comparative Regulatory performance of PI, NLPI and GSPI control
algorithms with step change in reboiler heat duty for XD,1………….….37
6b. Comparative Regulatory performance of PI, NLPI and GSPI control

algorithms with step change in reboiler heat duty for XD,2………….….38

7a. Comparative Regulatory performance of PI, NLPI and GSPI control
algorithms with step change in XB,1…………………………………….39
7b. Comparative Regulatory performance of PI, NLPI and GSPI control
algorithms with step change in XB,2………………………………….....40
8. Block diagram of an adaptive state estimator control strategy………...51
9. Open loop performance of reactive batch distillation without reaction.. 51
10. a. Open-loop response of the batch reactive distillation column under
total reflux ratio………………………………..………...................... 52
10. b. Open-loop response of the batch reactive distillation column under
reflux ratio 0.8……………………………………….…….…………. 53

iv
11. Comparison of the estimated outputs (ASE 1, ASE 2) and process
outputs with out any change for production stage…………………….. 54
12. Comparison of the estimated outputs (ASE 1, ASE 2) and process
outputs with two consecutive step changes in heat input to reboiler…..55
13. Comparison of the estimated outputs (ASE 1, ASE 2) and process
outputs with two consecutive step changes in Tray efficiency……...…56
14. Comparison of the estimated outputs (ASE 1, ASE 2) and process
outputs with initialization error in VnT ynT,3 ..……………………..….…57
15. Comparison of the estimated outputs (ASE 2) and process outputs
with initialization error in both VnT ynT,3 and r3…………………..……..58
16. Comparative closed loop performance of GSPI and GMC control
algorithms ………………………………………………………………59
17. Comparative Regulatory performance of GSPI and GMC control
algorithms with step change in reboiler heat duty for XD,3......................60
18. Comparative closed loop performance of GSPI and GMC control
algorithms with a set point change in XD,3 ..............................................61

v
 ABSTRACT

Modeling, Simulation and Control of a multicomponent

Batch Distillation Processes
BY
P.V.Radha Krishna Adari

The present work is devoted to study the comparative closed-loop performance of

multicomponent non-reactive batch distillation followed by a multi-component reactive
batch distillation column. First, the dynamic model of the non-reactive batch column has
been developed and then simulated. In the next, three different control schemes, namely
proportional-integral (PI), Nonlinear Proportional-Integral (NLPI), and Gain Scheduled
Proportional-Integral (GSPI), have been designed. Then the control structures have been
employed on the simulated non-reactive batch distillation column for comparative control
study. Several simulation experiments have been conducted to examine the servo as well
as regulatory performance.
In the subsequent part, the model structure of a multi component reactive batch
distillation column has been constructed. Then dynamic simulation has been performed to
observe the column performance at start-up and production phase. In the next, two non-
linear adaptive state estimators have been designed for the concerned process.
Interestingly, the state predictor includes only the component mass balance equation
around condenser-reflux drum system. Subsequently, an adaptive control scheme has
been synthesized for the example reactive column. Comparative performance study has
been conducted between the non-linear adaptive controller and the GSPI using the
simulative rectifier.

vi
 Chapter-1
Introduction
______________________________________________________

Distillation is the oldest separation process and the most widely used unit
operation in the industry. The process of distillation has been around since 4000 BC,
where it was first discovered by the Babylonians. It wasn’t until an Arabian alchemist
named Jabir Ibn Hayyan (721-815 AD) invented the alembic still that distillation became
used for beverage purposes. The process of distillation quickly spread after the Persian
physician and scientist Rhazes (865-925 AD) published a book on the process and theory
of distilling alcohol. In the early 19th century the basics of modern techniques including
pre-heating and reflux were developed, particularly by the French, then in 1830 a British
Patent was issued to Aeneas Coffey for a whiskey distillation column, which worked
continuously and may be regarded as the archetype of modern petrochemical units. In
1877, Ernest Solvay was granted a U.S. Patent for a tray column for ammonia distillation
and the same and subsequent years saw developments of this theme for oil and spirits.

Distillation separates two or more liquid components in a mixture using the

principle of relative volatility or boiling points. The greater the difference in relative
volatility the greater the nonlinearity and the easier it is to separate the mixture using
distillation. The process involves production of vapour by boiling the liquid mixture in a
still and removal of the vapour from the still by condensation. Due to differences in
relative volatility or boiling points, the vapour is rich in light components and the liquid
is rich in heavy components.

Often a part of the condensate is returned (reflux) back to the still and is mixed
with the outgoing vapour. This allows further transfer of lighter components to the
vapour phase from the liquid phase and transfer of heavier components to the liquid
phase from the vapour phase. Consequently, the vapour stream becomes richer in light
components and the liquid stream becomes richer in heavy components.

1
 Different types of devices called plates, trays or packing are used to bring the
vapour and liquid phases into intimate contact to enhance the mass transfer. Depending
on the relative volatility and the separation task (i.e. purity of the separated components)
more trays (or more packing materials) are stacked one above the other in a cylindrical
shell to form a column. The distillation process can be carried out in continuous, batch or
in semi-batch (or semi-continuous) mode.
In continuous distillation column feed is fed to the column at one or more points
along the column. Liquid runs down the column due to gravity while the vapour runs up
the column. The vapour is produced by partial vaporisation of the liquid reaching the
bottom of the column. The remaining liquid is withdrawn from the column as bottom
product rich in heavy components. The vapour reaching the top of the column is partially
or fully condensed. Part of the condensed liquid is refluxed to the column while the
remainder is withdrawn as the distillate product. The column section above the feed tray
rectifies the vapour stream with light components and therefore is termed as rectifying
section. The column section below the feed tray strips heavy components from the vapour
stream to the liquid stream and is termed as stripping section.
Batch distillation is the oldest operation used for separation of liquid mixtures.
For centuries and also today, batch distillation is widely used for the production of fine
chemicals and specialised products such as alcoholic beverages, essential oils, perfume,
pharmaceutical and petroleum products. It is the most frequent separation method in
batch processes (Lucet et al., 1992). The essential features of a batch distillation column
are:
• A bottom receiver/reboiler which is charged with the feed to be processed and
which provides the heat transfer surface.
• A rectifying column (either a tray or packed column) superimposed on the
reboiler, coupled with either a total condenser or a partial condenser system.
• A series of product accumulator tanks connected to the product streams to
collect the main and or the intermediate distillate fractions.

2
 Operation of such a column involves carrying out the fractionation until a desired
amount has been distilled off. The overhead composition varies during the operation and
usually a number of cuts are made. Some of the cuts are desired products (main-cuts)
while others are intermediate fractions (off-cuts) that can be recycled to subsequent
batches to obtain further separation. A residual bottom fraction may or may not be
recovered as product (Mujtaba, 1989).
The batch processing is a very attractive separation operation mainly for the
following reasons (Barolo and Berto, 1998a)
(i) A single batch column can separate a multicomponent feed mixture into
several products within a single operation,
(ii) The flexibility of a batch process is such that the frequent change of market
demands and strict product purity requirements can be accommodated,
(iii) The production amounts in a batch process are usually small with minimum
raw material inventories; this often results in an economic incentive.
The basic difference between a batch column and continuous column is that in
continuous column feed is continuously entering the column, while in batch distillation
the reboiler is normally fed at the beginning of the operation. Also, while the top products
are removed continuously in both batch and continuous operation, there is no bottom
product in the batch distillation. Since in a continuous operation the total product flow
rate equals that of incoming feed, the process reach steady state. In batch distillation, on
the other hand, the reboiler gets depleted over time, so the process is unsteady.
Batch distillation requires the least amount of capital for separating relatively pure
components. Continuous distillation generally requires a separate column for each
component. The same operation could be obtained with one batch distillation column
with product cuts.
Reactive distillation is one major step in the history of distillation in achieving
these goals. Reactive distillation processes couple chemical reactions and physical
separations into a single unit operation. These processes as a whole are not a new concept
as the first patent dates back to the 1920s. The initial publications dealt with
homogeneous self catalyzed reactions such as esterifications and hydrolysis.
Heterogeneous catalysis in reactive distillation is a more recent development and was

3
first described by Spes. While the concept existed much earlier, the first real world
implementation of reactive distillation took place in 1980s.
Batch distillation with or without chemical reaction is used in industry for the
production of small amounts of products with high added value and for processes where
flexibility is required. Distillation with chemical reaction is well suited for processes
where one of the products has a lower boiling point than other products and reactants.
The higher volatility of this product induces a decrease of its concentration in the liquid
phase, thus leading to higher reactant conversions than with reaction alone. The higher
volatility of this product results in a decrease in its concentration in the liquid phase,
therefore increasing the liquid temperature and hence reaction rate, in the case of an
irreversible reaction. With reversible reactions, elimination of products by distillation
favors the forward reaction. In both cases higher conversion of the reactants is expected
than by reaction alone. Therefore, in both cases, a higher amount of distillate
(proportional to the increase in conversion of the reactant) with desired purity is expected
than that obtained by distillation alone (as in traditional approach).

4
 Chapter-2
Literature Review
The flexibility provided by a batch rectifier, however, gives rise to challenging
control problems that are basically owing to the nonlinear, nonstationary, and finite time
duration nature of the underlying dynamics. The batch distillation is inherently an
unsteady state process and therefore, there is no normal condition at which the
conventional input/output linear model can be formulated and the employed controller
can be tuned. So, it is really a tough job to achieve satisfactory closed-loop process
response (Quintero-Marmol et al., 1991; Kim, 1999). Finefrock et al. (1994) mentioned
that if a PI composition controller is implemented, the controller gain should be varied
during the operation in order to enable control to be maintained. They suggested using a
gain-scheduled PI (GSPI) controller to accomplish the control objective. Frattini-Fileti
and Rocha-Pereira (1997) applied the GSPI control scheme on a binary batch distillation
column. In their strategy, the controller gain is increased during the batch operation in
order to meet the control objective. Predictive and adaptive control approaches have been
used to tackle the nonstationary nature of the batch rectifier. Li and Wozny (1997)
showed that the optimal profiles cannot be tracked with conventional linear controllers.
To realize the optimum of multiple-fraction batch distillation, subsequently they
(Li and Wozny, 2001) presented a predefined optimal policy that can be tracked with an
adaptive control law. To follow the rapidly changing process dynamics, they have applied
a recursive least-square estimation technique with a variable forgetting factor for the
online plant identification. In recent years, various approaches for designing the nonlinear
control methodologies have been reported in the literature. Unless they are using “ad
hoc” designs, most of the nonlinear controllers, such as globally linearizing controller
(GLC) (Hunt et al., 1983; Kravaris and Chung, 1987; Isidori, 1989; Jana et al., 2005),
generic model controller (GMC) (Lee and Sullivan, 1988; Signal and Lee, 1992), internal
model controller (IMC) (Henson and Seborg, 1991; Barolo and Berto, 1998a) and many
others, require the feedback of state variables to implement the control laws.
Although direct integration of the process model, i.e., open-loop observer
(Kravaris and Chung, 1987), may be employed to estimate the required state feedback

5
and performs satisfactorily for a limited number of applications, a closed-loop
observation scheme with feedback correction is still desirable, especially, when the
system consists of modeling error(s) or pure integrator(s). The development of closed-
loop observers for nonlinear systems still provides an open area for research. Many
observation approaches have been developed so far and few of them are applied with
good results on the batch distillation processes (Quintero-Marmol et al., 1991; Quintero-
Marmol and Luyben, 1992; Barolo et al., 2000; Oisiovici and Cruz, 2000; Venkateswarlu
and Avantika, 2001; Zamprogna et al., 2004). It is recognized that the performance of a
nonlinear modelbased control system largely depends on the convergence ability as well
as robustness of the state estimator. Barolo and Berto (1998a) constructed a control
strategy within the framework of nonlinear internal model control (Henson and Seborg,
1991). To estimate the distillate composition from the selected tray temperature
measurements, the authors have used the extended Luenberger observer (ELO)
(Quintero-Marmol et al., 1991). The derived control law is then successfully applied on a
binary and a ternary batch rectifier. The authors reported the significant improvement
provided by a nonlinear controller over a linear control law.
The nonlinear model-based controller provided very fast process response
towards the desired operating point, and the reflux rate profile was relatively smooth.
Barolo and Berto recommended the use of a stochastic estimator (like an extended
Kalman filter (EKF)) when a large degree of noise is expected. Dechechi et al. (1998)
developed a nonlinear model predictive controller for overhead composition regulation of
a batch column. In their approach, an ELO coupled with an optimization problem has
been solved online and this approach can lead to a very complex controller (Barolo and
Berto, 1998a).
Alvarez-Ramirez et al. (2000) have presented a strategy that comprises of a
controller (classical PID with antireset windup) and an observer to estimate the modeling
error. A drawback of this control approach (Monroy-Loperena and Alvarez-Ramirez,
2003) is that they strongly rely on an accurate model of the batch distillation process.
Subsequently, Oisiovici and Cruz (2001) proposed an inferential control structure that
consisted of the GLC and the EKF (Baratti et al., 1995). They have shown that the
inferential control scheme yields good results for a high-purity multicomponent batch

6
distillation column. They pointed out that the performance of the EKF is greatly
influenced by the vapor–liquid equilibrium (VLE) data. The authors also found that the
tracking performance of the filter usually improved when larger sets of secondary
measurements were used, and/or when the sampling frequency was increased. But both
the options increase the design complexity and the computational load. Han and Park
(2001) have used the quasi dynamic estimator (QDE) (Quintero-Marmol et al., 1991) to
estimate the distillate composition in the closed-loop batch rectifier. But Quintero-
Marmol et al. (1991) found that the ELO provided better performance than the QDE for a
ternary batch column. Then, Monroy-Loperena and Alvarez-Ramirez (2003) developed a
technique for the identification and control of a batch distillation process. The feedback
controller is derived in the framework of robust nonlinear control (Alvarez-Ramirez,
1999) with modeling error compensation technique for the control of distillate
composition via manipulations of the reflux ratio.
Cuille and Reklaitis (1986) considered the simulation of reactive batch
distillation, with reaction occurring on the plates, in the condenser, and in the reboiler.
They considered the esterification of 1-propanol with acetic acid, but the example was
not suitable for use in batch distillation. Since 1-propanol (one of the reactants) is the
more volatile component in the system, the removal of species by distillation causes the
removal of reactant from the column thus decreasing conversion..
Wilson discussed the optimal design of batch distillation processes using a
simplified column model involving chemical reaction and using repeated simulation. For
a commercially used complex parallel reaction scheme and using a simple economic
model, he showed the benefit of integrating reaction and distillation. He generated a
number of plots of process efficiency for a range of alternative process and design
variable choices and suggested an optimal design and operation of reactive batch
distillation. (Wilson, J. A., 1987)
Albet et al. (1991) presented a method for obtaining operational policies for both
reactive and non reactive batch distillation systems using repeated simulation techniques.
Mujtaba and Macchietto, (1991, 1993) discussed reactive batch distillation is
presented as a proper dynamic optimization problem incorporating a detailed dynamic
model.

7
 Chapter-3

Modeling, Simulation and Control of a multicomponent Batch

Distillation Column

3.1. Process Description

In batch distillation, a liquid mixture is charged into a vessel and heat is added to
produce vapor that is fed into a rectifying column. The liquid mixture can be a fresh feed
and also with any recycled slop cuts. During the initial startup period, the column
operates under total reflux condition in which vapor from the top of the column is
condensed and returned to the column. The operation of batch distillation described here
corresponds to a ternary system. During the column operation under total reflux
condition, the concentration of the lightest component buildup on the upper trays in the
column and the concentrations of the intermediate component and heaviest component
decreases in the top of the column but increases in the still pot. When the concentration
of the lightest component in the distillate reaches its specified purity level, then the
distillate product withdrawal is begun. During the withdrawal of the first product, there is
a composition front located in the lower part of the column that separates the lightest and
intermediate components. This front moves up the column as light product is removed.
When this front nears the top of the column, the distillate stream is diverted to another
tank as the 1rst slop cut. When the concentration of the intermediate component in the
distillate reaches its speci1ed purity level, the distillate is diverted to another tank in
which second product is collected. When the purity of the material in this tank drops to
the speci1ed purity level, the distillate stream is diverted into another tank, and the
second slop cut is collected until the average composition of the material remaining in the
still pot and on the trays in the column meets the purity speci1cation of the heavy
product.
In order to represent realistic operation of actual batch distillation column, a
rigorous nonlinear model that considers simultaneous effect of heat and mass transfer
operations and fluid flow on the plates is needed. Such batch distillation model is derived
from first principles involving dynamic material and component, and algebraic energy

8
equations supported by vapor–liquid equilibrium and physical properties. The
multicomponent batch distillation dynamics simulator has major computation functions
like vapor flow, liquid flow and tray holdup calculations, enthalpy calculations, average
molecular weight and density calculations, and vapor–liquid equilibrium calculations.
As assumed, the production phase the reflux drum holdup is kept constant
employing Proportional controller.
The operation of batch distillation described here corresponds to a ternary system
of cyclohexane–toluene–chlorobenzene. Among these constituent feed components,
cyclohexane is the lightest component, toluene is the intermediate component, and
chlorobenzene is the heaviest component. The model structure of the ternary distillation.

Fig. 1. schematic representation of the multicomponent Batch Distillation Process

9
3.2. Assumptions
Process is developed based on the following assumptions

• Staged batch distillation column with trays numbered from the bottom and top
(total 14 trays including still pot and reflux drum).
• Perfect mixing and equilibrium on all trays.
• Constant stage pressures (atmospheric) and tray efficiencies
(Vapor-phase Murphree efficiency = 75%).
• Negligible tray vapor holdups.
• Total condensation with no sub cooling in the condenser.
• Nonlinear Francis weir formula (Luyben, 1990) for tray Hydraulics calculations.
• Variable liquid holdup in each tray.
• Constant liquid holdup in the reflux drum (perfectly controlled by a conventional
proportional (P) controller with Proportional gain = −0.0005).
• Raoult’s law for the vapor–liquid equilibrium.

Batch distillation is inherently an unsteady state process. Consider the distillation

column as shown in Fig.1. Normally in batch processes the feed is charged in the still at
the bottom of the column. For this study feed is a cyclohexane-toluene-chlorobenzene
mixture, which has composition 0.4-0.4-0.2 respectively. This mixture is to be separated
by this batch process. The column shown in Fig.1. has 12 plates excluding reboiler and
reflux drum. It has divided into five sections for doing the material and energy balances
easily. MB is the amount of holdup in the still. QR amount of heat is supplied to still. Due
to that heat feed in the still changes its phase according relative volatility and VB amount
of vapors are formed which tries to go up through the first plate and so on. Now on first
plate there are four streams two inlet and outlet each. The suffix in the stream name
shows the stream coming from that plate. For example L1 is the amount of liquid stream
coming from first plate. V is used for vapor rate, M for hold up Hl for liquid enthalpy, Hv
for vapor enthalpy, x is liquid composition and y is vapor composition.

10
3.3. Modeling Equations
Material balance, component balance and enthalpy balance equations can be
written accordingly,
The change in the heat energy for a very small amount of time can be considered
negligible i.e. the change is very less. So d(M Hl)/dt is very small, d(M Hl)/dt = 0; on
rearrangement, we get
3.3.1. Reboiler Section

Total mass balance:

dM B
= L1 − V B (3.1)
dt

Component mass balance:

d ( M B xB, j )
= L1 x1. j − V B y B , j (3.2)
dt
Where j = 1,2….c (c = No. of components)

dxB , j L1 V
= ( x1. j − xB , j ) − B ( yB , j − xB , j ) (3.3)
dt MB MB

Energy balance:

d ( M B H Bl )
= L1 H 1l − V B H Bv + Q R (3.4)
dt

( L1 H l
+ Q )
VB = 1
v
R (3.5)
H B

3.3.2. First Plate

dM 1
= V B + L2 − V1 − L1 (3.6)
dt

d ( M1 x1, j )
= VB yB, j + L2 x2 , j − V1 y1, j − L1 x1, j (3.7)
dt
Where j = 1,2….c (c = No. of components)

11
dx1, j VB L2 V1
= ( yB, j − x1, j ) + ( x2 j − x1 j ) − ( y1, j − x1, j ) (3.8)
dt M1 M1 M1

d ( M 1 H 1l )
= V B H Bv + L 2 H 2l − V1 H 1v − L1 H 1l (3.9)
dt

V B H Bv + L2 H 2l − L1 H 1l
V1 = (3.10)
H 1v

3.3.3. nth Plate

dM n
= Vn−1 + Ln+1 − Vn − Ln (3.11)
dt

d(Mn xn, j )
= Vn−1 yn−1, j + Ln+1xn+1, j −Vn yn, j − Ln xn, j (3.12)
dt
Where j = 1,2….c (c = No. of components)

dxn, j Vn−1 L V
= ( yn−1, j − xn, j ) + n+1 ( xn+1, j − xn, j ) − n ( yn, j − xn, j ) (3.13)
dt Mn Mn Mn

d(Mn Hnl )
= Vn−1Hnv−1 + Ln+1Hnl +1 −Vn Hnv − Ln Hnl (3.14)
dt

V n −1 H nv−1 + L n +1 H nl +1 − L n H nl
Vn = (3.15)
H nv

3.3.4. Top Plate (nTth)

dM nT
= R + VnT −1 − LnT − VnT (3.16)
dt

d ( M n T xn T , j )
= Rx D , j + VnT −1 ynT −1, j − LnT xnT , j − VnT ynT , j (3.17)
dt
Where j = 1, 2….c (c = No. of component

dxnT ,i VnT − 1 R V
= ( ynT −1,i − xnT ,i ) + ( xD ,i − xnT ,i ) − nT ( ynT ,i − xnT ,i ) (3.18)
dt M nT M nT M nT

12
d (M nT H nl T )
= RH Dl + VnT −1 H nv T −1 − LnT H nl T − VnT H nv T (3.19)
dt

For start-up phase, R = VnT

VnT −1 H nv T −1 − LnT H nl T
VnT = (3.20)
( H nv T − H Dl )

For production phase, for any R

RH l
+ V n T −1 H nv T −1 − L n T H nl T
VnT = D
(3.21)
H nv T

3.3.5. Reflux Drum

dM D
= V nT − R − D (3.22)
dt

d ( M D xD , j )
= VnT ynT , j − RxD, j − DxD, j (3.23)
dt
Where j = 1,2….c (c = No. of components)

13
3.3.6. Bubble point calculations
For solving the vapor rates i.e. the energy balance equation, one requires the
enthalpy data. And to calculate the enthalpy; temperature should be known. Therefore, it
is necessary to have the temperature-composition correlation. The vapor-phase
composition in equilibrium with the liquid-phase is given by ,

yi* = k i ( xi , Ti , Pi ) xi (3.24)

y i* = k i xi (3.25)
Where k is equilibrium ratio, for this study k is calculated as follows,

Pi s
ki = (3.26)
Pt
Where Pi s the vapor pressure was calculated by using Antoine equation, and Pt
was the total pressure.
The Antoine equation is given as,

⎡ B ⎤
Pi s = exp⎢ A − (3.27)
⎣ T + C ⎥⎦

Where the A, B, C are the constants and T is the temperature. Here the pressure is
in mmHg and temperature is in K

So now the equilibrium ratio becomes,

⎡ B ⎤
exp ⎢ A −
⎣ T + C ⎥⎦
ki = (3.28)
Pt

14
 Assume temperature T

Calculate equilibrium ratio Ki from

equation (28)

Calculate yi from equation (25) using

given xi

No
Check Tnew=Told - f 1( T )
f(T)=∑(yi) -1=0.00001 f (T )

Yes

Obtain yi and T

Fig.2. Calculation of Temperature (T) and vapor composition (yi)

3.3.7. Enthalpy Calculations

The liquid and vapor enthalpies were calculated from the following equations.
Where the temperature is in (K) and enthalpy is in (kJ/kmol)
H l = a + bT + cT 2 + d T 3 + eT 4 (3.29)

H v = a + bT + cT 2 + d T 3 + eT 4 (3.30)

15
3.3.8. Liquid flow rate Calculations
For the calculation of the liquid flow rates the Francis weir formula was used,
which is given by the following equation. Liquid flow rate is function of density,
molecular weight and tray parameters which includes the tray holdup, height, length and
column diameter. The change in the liquid rate is accounted by changes in the density and
indirectly the liquid composition.
1 .5
⎛ 183 .2 M m W H ⎞
ρ W L (999 )⎜⎜ − ⎟
⎝ ρ D c2 12 ⎟⎠
L= (3.31)
m

Where L liquid rate (lb-mol/hr)

ρ Liquid density (lb/ft3)
WL Weir Length (in)
WH Weir Height (in)
M Hold up (lb-mol)
m Average molecular weight (lbm/lb-mol)
Dc Column Diameter (in)

Density and molecular weight of liquid on the plate is calculated by the sum of
product of liquid composition & density, and liquid composition & molecular weight of
all the components present in the system respectively, is given by the following
mathematical relation.
ρ = ∑ xi ρ i (3.32)

m = ∑ xi mi (3.33)

16
Table. 1. Batch Distillation Column specification

System Cyclohexane / Toluene / Chlorobenzene

Feed (kmol) 30

Feed composition 0.40 / 0.40 / 0.20

Number of trays (excluding
12
reboiler)
Tray hold-up (kmol) 0.03

Heat input to still (kJ/min) 25,000

Distillate composition 0.99

Reflux drum hold-up (kmol) 1.0

Production rate (kmol/min) 0.10

Time step (min) 0.005

Column diameter (inch) 18

Weir Length (inch) 12

Weir Height (inch) 0.30

Murphee Tray Efficiency 0.75

17
3.4. Simulation algorithm

Modeling equations are solve by a stepwise procedure is given below.

Step1: Declare all the variables and initialize the start up phase liquid composition (xi)

and liquid hold up (Mn) on each plate.

Step2: Calculate vapor phase composition (yi) and temperature (T) on each plate

as shown in fig.2 bubble point calculation.

Step3: Calculate actual vapor phase composition (yi) by multiplying the Murphee

tray efficiency.

Step4: Find out the vapor and liquid enthalpy using equations (29) and (30) for all

the component on each plate and then total liquid and vapor phase enthalpies

by multiplying the respective compositions on each plate.

Step5: Then estimates the liquid flow rates using the Francis weir equation (31) In

section with the help of density, molecular weight, holdup and tray specifications.

Step6: Find out the vapor flow rates using the energy balance equations.

Step7: Calculate the liquid hold up on each plate by equations as modeled in section 2

Use the total reflux for start up phase.

Step8: Finally calculate the liquid phase composition. And then check for sum.

The sum of it should be equal to zero. If it is not then normalize it using

current and previous value.

18
3.5. Controller synthesis
In closed-loop simulation study, the control objective is to recover the light
component and intermediate component at a constant purity. The manipulated input is the
reflux flow rate. Distillate light component and intermediate components are controlled
with any one of Proportional integral (PI) controller, Nonlinear Proportional integral
(NLPI) controller and Gain Scheduling Proportional Integral (GSPI) controller.
3.5.1. Proportional Integral controller
A PI Controller has two terms, one proportional to the error and other
proportional to the integral of the error. PI controller equation (process systems Analysis
and Control, Coughanor) is given as
t
kc
u = u s + k ce +
ti ∫ edt
0
(3.34)

kc =gain
ti =integral time (min)
us =constant
This equation can be written as
1 t

τ i ∫0
R = Rs + kc(e + edt) (3.35)

e =xDsp, j-xD
3.5.2. Nonlinear Proportional Integral Controller (NLPI)
The idea of NLPI controller is to modify the controller action in some way to
compensate for the nonlinearity of the process. The equation (Bequette, 2006) is given as
t
kc
t i ∫0
u = u s + k ce + edt (3.36)

k c = k co (1 + a . e ) (3.37)

This means that controller output is effectively proportional to square of the error
kco=controller gain with zero error
e =absolute magnitude of error

19
 a=adjustable constant
1 t
R = R s + kc(e +
τi ∫ 0
ed t ) (3.38)

e =xDsp, j-xD
3.5.3 Gain Scheduling Proportional Controller (GSPI)
A GSPI law (Bequette, 2006) is given as
t
kc
u = u s + k ce +
ti ∫ edt
0
(3.39)

1 − xDSP , j
kc = kco for xD,j>xDSP,j
1 − xD , j
kc = kco for xD,j<xDSP,j
This equation can be written as
1 t
R = R s + kc(e +
τi ∫ 0
ed t ) (3.40)

e =xDsp, j-xD
All the above cases liquid holdup in reflux drum controlled by proportional
controller. Generally, a level controller is used to maintain the liquid holdup in the reflux
drum at a desired value. The variation of liquid holdup so small ( ± 1.0%) that is
reasonable to assume constant mD. This mD is controlled by manipulating the distillate
rate D. Proportional controller equation (process systems Analysis and Control,
Coughanor), given as

u = us + k ce (3.41)
D=Ds+kcm*e (3.42)

e=mDsp-mD
kcm = proportional gain

20
3.6. Simulation Results and Discussion
Several simulation experiments have been carried out on the multi component
batch distillation column in open-loop as well as closed loop mode. The column is
employed for the fractionation of a hydrocarbon system, cyclohexane–toluene–

chlorobenzene. In this system, cyclohexane and toluene are separated as distillate

products and chlorobenzene is separated as a product in the still pot.
In the batch distillation operation, first the column may be brought to the steady
state by considering the total reflux startup procedure. Then the production phase is
started and the controller switched on to maintain the specified product quality.
Sometimes, the product is withdrawn as soon as the distillate composition reaches its
desired value, without waiting for the steady state to be attained. Notice that immediately
after the production phase is started, controller responses may be very aggressive. It
happens because of the following two reasons: (i) immediate withdrawal of the distillate
product, and (ii) the change of distillation composition from the steady state value to the
set point value.
Fig. 3 presents the dynamics of the uncontrolled distillate composition of the
example batch column at the start-up phase. The steady state composition of the lightest
component in the distillate product is 0.999 (very close to 1) and the column can be
considered as a high-purity distillation process. In the following, the open-loop as well as
closed-loop dynamics of this multicomponent batch rectifier are discussed in the
production phase only.
Fig. 4 illustrates the uncontrolled process dynamics at the production phase. The
production phase has been started from the steady state and with the withdrawal of
distillate. In the present simulation-based experiment, the distillate is discharged with a
constant flow rate of 0.1 kmol/min.

21
3.6.1. Constant Composition Control
In the present study, the set point composition has been fixed at a value of 0.985
for lightest and 0.97 for intermediate component. The column is started up as usual and
the lightest component withdrawal is begun as soon as this component met the
composition specification ,then set point step change is maintained from 0.995 to 0.985
at time= 30 min. The constant composition control is continued until the distillate
decreases to small (almost zero) value. The time taken to withdrawal the lightest
component is 91 min with purity of 0.985.At this moment the controller is switched off
and the intermediate component starts to reach the top of the column, time period 77 min
(91 min to 168 min). When this component has reached the desired purity (xDsp,2 =0.97),
constant composition control is started again. In between the withdrawal of lightest and
intermediate components, the distillate is collected as the slop cut. At this moment we
must note that in the simulation experiment, the distillate flow rate is manipulated by a
proportional level controller to maintain constant (nearly) liquid holdup in the reflux
accumulator. The time taken to withdrawal the second lightest component is 66 min with
purity of 0.97.At this moment the controller is switched off, during this second slop cut
third component starts to reach the bottom of the column up to purity of 0.938 (234 min
to 294 min).
Withdrawals of lightest and intermediate components are shown in Fig.7.a and 7.b
with servo performance (+5% step change in set point) at time 30 min and 190 min.
Regulatory performances are shown for (+5%) step change in heat input, feed change for
light component and intermediate component at time 30 min and 190 min(Fig.8 &Fig 9).
Comparative study among the PI, NLPI, and GSPI controllers based on the ISE
values.
t

∫e
2
ISE= dt (3.42)
0

e = error

22
Open loop performance

1
0.9
0.8
XD,j (mole fraction)

0.7
0.6 Cyclohexane
0.5 Toluene
0.4 Chlorobenzene
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Time (min) 11.2 min

0.6

0.5
XB,j (mole fraction)

0.4

0.3

0.2 Cyclohexane
0.1 Toluene
Chlorobenzene
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Time (min)

Fig. 3. open-loop process dynamics at start-up phase

23
 1
0.9
0.8
XD,j (mole fraction)

0.7
0.6 Cyclohexane
0.5 Toluene
0.4 Chlorobenzene
0.3
0.2
0.1
0
10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
Time (min) 147 min

0.7
0.6
XB,j (mole fraction)

0.5
0.4
0.3
Cyclohexane
0.2 Toluene
0.1 Chlorobenzene
0
10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
Time (min)

Fig .4. open-loop process dynamics at the production phase

24
Servo Performance

1 Set Point
GSPI
0.995
XDSP,1 / X D,1 (mole fraction)

NLPI
0.99 PI
0.985
0.98
0.975
0.97
0 10 20 30 40 50 60 70 80 90 100
Time (min)

0.8
R (kmol / min)

0.6

0.4 GSPI
NLPI
0.2 PI

0
0 10 20 30 40 50 60 70 80 90 100
Time (min)

Fig.5a. Comparative closed loop performance of PI, NLPI and GSPI control algorithms
with a set point change in XD,1 (0.995 to 0.985 at time=30 min)

25
 Set Point
1
XDSP,2 / X D,2 (mole fraction)
GSPI
0.99
NLPI
0.98 PI
0.97
0.96
0.95
0.94
0.93
140 150 160 170 180 190 200 210 220 230 240
Time (min)

0.8
0.7
0.6
R (kmol/min)

0.5
0.4
0.3 GSPI
0.2 NLPI
0.1 PI
0
140 150 160 170 180 190 200 210 220 230 240
Time (min)

Fig.5b.Comparative closed loop performance of PI, NLPI and GSPI control algorithms
with a set point change in XD, 2 (0.975 to 0.97 at time=190 min)

26
 1
XB,3 (mole fraction) 0.95

0.9

0.85

0.8

0.75

0.7
234 244 254 264 274 284 294
Time (min)

Fig.5c.Third component in the bottom after second slop cut

Regulatory performance

1
0.995
XDSP,1 / X D,1 (mole fraction)

0.99
0.985 Set Point
GSPI
0.98
NLPI
0.975 PI
0.97
0 10 20 30 40 50 60 70 80 90
Time (min)

0.9
R (kmol/min)

0.8

0.7
GSPI
0.6 NLPI
0.5 PI

0.4
0 10 20 30 40 50 60 70 80 90
Time (min)

Fig.6a.Comparative Regulatory performance of PI, NLPI and GSPI control algorithms

with +5% (25000 → 26250 kJ/min ) step change in reboiler heat duty at time =30min.

27
 0.99
XDSP,2 / X D,2 (mole fraction)

0.98

0.97
Set Point
0.96
GSPI
0.95 NLPI
0.94 PI

0.93
140 150 160 170 180 190 200 210 220 230 240
Time (min)

0.9
0.8
0.7
R (kmol/min)

0.6
0.5 GSPI
0.4 NLPI
0.3 PI
0.2
140 150 160 170 180 190 200 210 220 230 240
Time (min)

Fig.6b.Comparative Regulatory performance of PI, NLPI and GSPI control algorithms

with +5% (26250 → 27562 kJ/min ) step change in reboiler heat duty at time =190 min.

28
 0.3

XB,1 (mole fraction) 0.25

0.2
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90 100
Time (min)

1
XDSP,1/ X D,1 (mole fraction)

0.995
0.99
0.985 Set Point
GSPI
0.98
NLPI
0.975 PI
0.97
0 10 20 30 40 50 60 70 80 90 100
Time (min)

1
0.9
0.8
R (kmol/min)

0.7
0.6 GSPI
0.5 NLPI
0.4 PI
0.3
0 10 20 30 40 50 60 70 80 90 100
Time (min)

Fig.7a. Comparative Regulatory performance of PI, NLPI and GSPI control algorithms
with +5 % (0.17128, 0.56771, 0.26110 → 0.22128, 0.53318, 0.24553) step change in XB,1
at time =30min.

29
 0.7
0.6
XB,2 (mole fraction)

0.5
0.4
0.3
0.2
0.1
0
140 150 160 170 180 190 200 210 220 230 240 250
Time (min)

1
0.99
0.98
R (kmol/min)

0.97
Set Point
0.96
GSPI
0.95
NLPI
0.94 PI
0.93
140 150 160 170 180 190 200 210 220 230 240 250
Time (min)

0.8
0.7
0.6
R (kmol/min)

0.5
0.4
GSPI
0.3 NLPI
0.2 PI
0.1
140 150 160 170 180 190 200 210 220 230 240 250
Time (min)

Fig.7b. Comparative Regulatory performance of PI, NLPI and GSPI control algorithms
with +5 % ( 0.000095.48899,0.51067 → 0.000084,0.53899,0.46080) step change in XB,jat
time =190 min.

30
Tuning Parameters

Table2. (Servo Performance)

For component X D,1

Controller Kci τi a ISE

PI 35.65 6.5 ---- 0.000193259

NLPI 35.65 6.5 12.5 0.000182704

GSPI 35.65 6.5 ---- 0.000126476

Liquid level in the reflux drum controlled by proportional controller. Tuning parameter
for proportional controller kcm= -0.0005

Table3. (Servo Performance)

For component X D,2

Controller Kci τi a ISE

PI 38.68 7.5 ---- 0.00000491465

NLPI 38.68 7.5 12.1 0.00000421346

GSPI 38.68 7.5 ---- 0.00000394175

Liquid level in the reflux drum controlled by proportional controller. Tuning parameter
for proportional controller kcm= -0.00048

31
Table4. (Regulatory Performance)

For component X D,1

Controller Kci τi a ISE

PI 35.65 6.5 ---- 0.000221949

NLPI 35.65 6.5 12.5 0.000214235

GSPI 35.65 6.5 ---- 0.000126805

Liquid level in the reflux drum controlled by proportional controller. Tuning parameter
for proportional controller kcm= -0.0005

Table5. (Regulatory Performance)

For component X D,2

Controller Kci τi a ISE

PI 38.68 7.5 ---- 0.000068712

NLPI 38.68 7.5 12.1 0.000062481

GSPI 38.68 7.5 ---- 0.000048687

Liquid level in the reflux drum controlled by proportional controller. Tuning parameter
for proportional controller kcm= -0.00048

32
Table6. (Regulatory Performance) for XB change
For component X D,1

Controller Kci τi a ISE

PI 35.65 6.5 ---- 0.000019565

NLPI 35.65 6.5 12.5 0.000019283

GSPI 35.65 6.5 ---- 0.0000152386

Liquid level in the reflux drum controlled by proportional controller. Tuning parameter
for proportional controller kcm= -0.0005

Table7. (Regulatory Performance) for XB change

For component X D,2

Controller Kci τi a ISE

PI 38.68 7.5 ---- 0.000058637

NLPI 38.68 7.5 12.1 0.000055245

GSPI 38.68 7.5 ---- 0.000043618

Liquid level in the reflux drum controlled by proportional controller. Tuning parameter
for proportional controller kcm= -0.00052

33
 Chapter- 4

Modeling, Simulation and Nonlinear Adaptive Control of a

multicomponent Reactive Batch Distillation Column
4.1. Process Description
Initially batch column is filled with reactants. In This work 4% of the total feed
charge as the total column holdup. Half of this holdup is taken as condenser holdup and
the rest is equally divided for the plate holdups. Plate compositions (M. Mujtaba and
Macchietto, 1997).In this present work the esterification of ethanol and acetic acid
considered. The reaction products are ethyl acetate (main product) and water. Sulfuric
acid in liquid form is used as catalyst.
The reversible reaction scheme is the following:

Acetic Acid (1) + Ethanol (2) Ethyl Acetate (3) + Water (4)

Rate of reaction r = k1C1C2-k2C3C4 (4.0)

Where rate constants are k1 = 4.76 x10-4 and k2 =1.63x10-4
The boiling temperatures are respectively 391.1, 351.5, 350.3, and 373.2 K. Ethyl
acetate, the main product has the lowest boiling temperature in the mixture and
consequently has the highest volatility. The continuous removal of this product by
distillation will shift the chemical equilibrium further to the right and will improve
conversion of reactants.

34
4.2. Assumptions
1. The vapor-phase holdup is assumed to be negligible compared to the liquid-phase
holdup on each phase.
2. Chemical reactions in the vapor phase are neglected.
3. The initial state of the column is the steady-state total reflux condition with no
reactions.
4. The liquid volumetric holdups on the plates will be assumed to be constant. Thus, the
model is directed at simulating the dynamics of the main production period during which
the hydrodynamic conditions are not widely varying.
5. The pressure drops and the plate efficiencies are constant during the operation
4.3. Modeling Equations
Material balance, component balance and enthalpy balance equations can be written
accordingly,
The change in the heat energy for a very small amount of time can be considered
negligible i.e. the change is very less. So d(M Hl)/dt is very small, d(M Hl)/dt = 0; on
rearrangement, we get
4.3.1. Reboiler Section

Total mass balance:

r c
dMB
= L1 − VB + rf ∑∑ γ i , m.rB , i.ε B (4.1)
dt m =1 i =1
Component mass balance:
d ( M B x B ,i ) r
= L1 x 1,i − V B y B ,i + r f ∑ γ i , m . r B , i .ε B (4.2)
dt m =1
Where i = 1,2….c (c = No. of components)
dxB,i L1 VB r r c
= ( x1.i − xB,i ) − ( yB,i − xB,i ) + rf ∑γ i, m.rB, i.ε B − xB.rf ∑∑γ i, m.rB, i.ε B (4.3)
dt MB MB m=1 m=1 i =1

Energy balance:
d ( M B H Bl ) r
= L1 H 1l − VB H Bv + QR + rf ∑ (γ i , m.rB , i .ε B ) (4.4)
dt m =1
r
L1 H + Q R + r f ∑ H B r B , iε B
l l
1
VB = v
m =1
(4.5)
H B

35
4.3.2. First Plate

r c
dM1
= VB + L2 −V1 − L1 + rf ∑∑γ i, m.r1, i.ε 1 (4.6)
dt m=1 i =1

d ( M 1 x1,i ) r
= VB y B ,i + L2 x2 ,i − V1 y1,i − L1 x1,i + rf ∑ γ i , m.r1, i .ε 1 (4.7)
dt m =1
Where i = 1,2….c (c = No. of components)
r r c
dx1, i VB L2 V1
= ( yB, i − x1, i) + (x2, i − x1, i) − ( y1, i − x1, i) + rf ∑γ i, m.r1, i.ε1 − xB.rf ∑∑γ i, m.r1, i.ε1 (4.8)
dt M1 M1 M1 m=1 m=1 i=1

d ( M 1H1l ) r
= VB H Bv + L2 H 2l − V1H1v − L1H1l + rf ∑ (γ i , m.r1, i .ε 1) (4.9)
dt m =1

r
VB H v
B + L 2 H 2l − L1 H 1l + r f ∑ H 1l r 1 , i .ε 1
V1 = m =1
(4.10)
H 1v

4.3.3. nth Plate

r c
dM n
= Vn −1 + Ln +1 − Vn − Ln + rf ∑∑ γ i , m.rn , i.ε n (4.11)
dt m =1 i =1

d ( M n xn ,i ) r
= Vn −1 yn −1,i + Ln +1 xn +1,i − Vn yn ,i − Ln xn ,i + rf ∑ γ i , mrn , iε n (4.12)
dt m =1
Where i = 1,2….c (c = No. of components)

dxn,i Vn−1 L V r
= ( yn−1,i − xn,i ) + n+1 ( xn+1,i − xn,i ) − n ( yn,i − xn,i ) + rf ∑γ i,m.rn,i.ε n
dt Mn Mn Mn m=1

r c
− xn.rf ∑∑ γ i , m.rn , i .ε n (4.13)
m =1 i =1

d(MnHnl ) r
=Vn−1Hnv−1 + Ln+1Hnl +1 −VnHnv − LnHnl ++rf ∑(γ i, m.rn, i.εn) (4.14)
dt m=1
r
Vn −1 H nv−1 + Ln +1 H nl +1 − Ln H nl + rf ∑ H lnrn , i .ε n
Vn = m =1
(4.15)
H nv

36
4.3.4. Top Plate (nTth)

r c
dM nT
= R + VnT −1 − LnT − VnT + rf ∑∑ γ i , m.rnT , i.ε nT (4.16)
dt m =1 i =1

d(MnT xnT ,i ) r
= RxD,i +VnT −1 ynT −1,i − LnT xnT ,i −VnT ynT ,i + rf ∑γ i, m.rnT , i.εnT (4.17)
dt m=1
Where i = 1,2….c (c = No. of components)
dxnT ,i VnT − 1 R V r
= ( ynT −1,i − xnT ,i ) + ( xD,i − xnT ,i ) − nT ( ynT ,i − xnT ,i ) + rf ∑γ i,m.rnT ,i.ε nT
dt MnT MnT MnT m=1

r c
− xnT .rf ∑∑ γ i , m.rnT , i .ε nT (4.18)
m =1 i =1
l r
d (MnT HnT )
= RHDl +VnT −1HnTv −1 − LnT HnT
l
−VnT HnTv + +rf ∑(γ i, m.rnT , i.ε nT ) (4.19)
dt m=1
For start-up phase, R = VnT

−1 − LnT H nT + rf ∑ H nT rnT , i .ε nT
v l l
VnT −1 H nT
VnT = m =1
(4.20)
(H v
nT −H ) l
D

For production phase, for any R

r
RH Dl + VnT −1 H nT −1 − LnT H nT + rf ∑ H nT rnT , i .ε nT
v l l

VnT = v
m =1 (4.21)
H nT

4.3.5. Reflux Drum

r c
dM D
= VnT − R − D + rf ∑∑ γ i , m.rD , i.ε D (4.22)
dt m =1 i =1

d ( M D xD,i ) r
= VnT ynT ,i − RxD,i − DxD,i + rf ∑γ i , m.rD, i.ε D (4.23)
dt m=1
Where i = 1, 2….c (c = No. of components)

37
4.3.6. Calculation of tray volume

θ
D O E
C

B
d=diameter of column
OA=d/2
AB=weir length
AC=BC
AC
Sin θ=
OA
AC
θ=sin-1
OA
AC
tan θ=
OC
OC=tan θ *AC
Area of ADB (down comer area) = Area of OADB – Area of OAB
2θ π d 2
Area of OADB = *
360 4
1
Area of OAB = (AB)*(OC)
2
Area of ABE = (Area of ADBE) – (Area of ADB)
Volume of tray = (Area of ABE) * (weir height)

38
Table. 8. Reactive Batch Distillation Column specification

System Acetic Acid + Ethanol Ethyl Acetate + Water

(1) (2) (3) (4)
Feed (kmol) 30

Feed composition 0.45 / 0.45 / 0.0 / 0.1

Number of trays (excluding
8
reboiler & reflux drum)
Tray hold-up (kmol) 0.075

Heat input to still (kJ/min) 3200

Distillate composition 0.936

Reflux drum hold-up (kmol) 0.6

Production rate (kmol/min) 0.06

Time step (min) 0.005

Column diameter (inch) 18

Weir Length (inch) 12

Weir Height (inch) 0.30

Murphee Tray Efficiency 0.75

39
4.4. Simulation algorithm

Modeling equations are solve by a stepwise procedure is given below.

Step1: Declare all the variables and initialize the start up phase liquid composition (xi)

and liquid hold up (Mn) on each plate.

Step2: Calculate vapor phase composition (yi) and temperature (T) on each plate

as shown in fig 2; bubble point calculation.

Step3: Calculate actual vapor phase composition (yi) by multiplying the Murphee

tray efficiency.

Step4: Find out the vapor and liquid enthalpy using equations (3.29) and (3.30) for all

the component on each plate and then total liquid and vapor phase enthalpies

by multiplying the respective compositions on each plate.

Step5: Then estimates the liquid flow rates using the Francis weir equation (3.31) In

section with the help of density, molecular weight, holdup and tray specifications.

Step6: Find out the vapor flow rates using the energy balance equations.

Step7: Calculate the liquid hold up on each plate by equations as modeled in section 2

Use the total reflux for start up phase.

Step8: Find out volume of tray (v), the rate of reaction using equation (4.0)

Step9: Finally calculate the liquid phase composition. And then check for sum.

The sum of it should be equal to zero. If it is not then normalize it using

current and previous value.

40
4.5. Nonlinear adaptive control algorithm
As stated, the adaptive control structure consisted of the nonlinear GMC and
an ASE. The closed-loop system having different controller elements and the process is
shown in Fig. 3. In the following, the detailed synthesis of the adaptive controller is
presented in generalized form.

Nonlinear
ASE

^
x
Nonlinear Output
GMC Process Map
u x y
ysp e

Fig. 8. Block diagram for the adaptive control algorithm

4.5.1. Generic model control

In nonlinear modeling of dynamic processes (Guo et al., 2001), it may be considered that
the system is nonlinear in the states, disturbances, and control variables but linear in the
model parameters such that
dx
= f (x,d)θ + g1(u, x, d), (4.34.a)
dt
y = cx, (4.34.b)

where the state x ∈ Rn, the model parameter θ ∈ ℜn , the measurable disturbance

ℜ ℜ
q m
d ∈ , and the input u ∈ . Moreover, f and g1 are matrices of nonlinear
functions. In this study, it is assumed that all states are measurable and c (coefficient
matrix) is a unity matrix. From the basic principle of GMC (Lee and Sullivan, 1988),
the following control law can be derived (Guo et al., 2001)

41
 t
f (x,d) θ + g1(u, x, d) − K 1e − K 2 ∫ edt = 0 (4.35)
0

where e is the error (=ysp − y) to the controller, ysp is the set point value of the output y,
and K1 and K2 are diagonal n×n tuning parameter matrices. Eq. (4.35) implies that the
GMC algorithm comprises of dynamic process model, proportional action term, and
integral action term. If g1 is linear with respect to u, then one can write
g1 (u, x, d) = b(x, d) u. Accordingly, Eq. (35) yields
⎡ t
⎤
u = (b(x, d))−1 ⎢ K 1e − K 2 ∫ edt − f ( x, d )θ ⎥ =0 (4.36)
⎣ 0 ⎦
The values of the elements of tuning parameter matrices can be found out based on the
following relationships given by Signal and Lee (1992)

2τ 1i
K 1( i , i ) = (4.37.a)
τ 2i
1
K 2( i , i ) = (4.37.b)
τ
2
2i

where τ1i and τ2i determine the shape and speed of the desired closed-loop trajectory (the
reference trajectory), respectively. The reference trajectory gives pseudo-second order
response for a step change in the set point. However, Yamuna and Gangiah (1991)
confirmed that the above relationships could be applied to compute the specified response
accurately. Once the values of τ1i and τ 2i are obtained, then K1 and K2 can be calculated
from Equations (4.37.a) and (4.37b).
4.5.2. Adaptive state estimation
In practice, there are two types of mismatch, structural mismatch and parameter
mismatch. The structural mismatch exists when there is a difference between the actual
plant model and the predictor model. The parameter mismatch occurs when the numerical
values of parameters in the predictor model differ with the true values. The effects of the
structural discrepancy on the closed-loop performance can be reduced if the imprecisely
known parameters are continuously updated.
In the present study, the model parameters in Eq. (4.34.a and 4.34.b) are supposed
to be time varying. Here, a nonlinear observer proposed by Farza et al. (1998, 1999) is

42
designed to estimate the poorly known parameters of the batch rectifier. It is also
assumed that the parameter dynamics in the nonlinear system (Eq. (4.34)) obey the
following general first-order equation:
dθ
= g2(u, x, d) +ε (4.38)
dt
where g2 is a nonlinear function and ε is an unknown function that may depend on x, θ, u,
d, noise, and so on. The assumptions that have been made are: ε is an unknown but
bounded function and the disturbance d with its time derivative are also bounded.
The nonlinear system equations ((4.34) and (4.38)) can be expressed in the following
condensed form
dZ −
= F ( x , d ) Z + G ( u, x , d ) + ε (4.39.a)
dy
y = CZ, (4.39.b)
⎡ x⎤ ⎡0 f ( x, d ) ⎤ ⎡ g1(u, x, d ) ⎤ − ⎡ 0 ⎤
Where Z = ⎢ ⎥ , F ( x, d ) = ⎢
0 ⎥⎦
, G ( u , x , d ) = ⎢ g 2(u, x, d ) ⎥ , ε = ⎢ε ⎥ , C =[In, 0],
⎣0⎦ ⎣0 ⎣ ⎦ ⎣ ⎦
with In the n×n identity matrix. f is an n × n matrix which is differentiable and the
corresponding partial derivative is continuous. According to Farza et al. (1998), the
nonlinear adaptive observer can be used to track the vector Z as follows:
^
dZ ^ ^
= F ( y , d ) Z + G ( u , y , d ) − Γ ( y , d ) S C (C Z − y )
−1 −1 T
(4.40)
dt

^ ⎡ y⎤ ^
(i) Z = ⎢ ^ ⎥ ∈ ℜ ,and θ ∈ ℜ
2n m

⎣⎢θ ⎦⎥
⎡ In 0 ⎤
(ii) Γ( y , d ) = ⎢ , and
⎣0 f ( y , d ) ⎥⎦
(iii) S is the unique symmetric positive-definite matrix which satisfies the algebraic
Lyapunov equation.
The gain of the estimator is obtained as
⎡ 2α In ⎤
Γ
−1 −1 T
( y, d )S C = ⎢ 2 −1 ⎥ (4.41)
⎢⎣α f ( y , d ) ⎥⎦

Where α >0 is a design parameter (Gauthier et al., 1992).

43
 It is obvious from Eq. (41) that only a single tuning parameter α is involved in the
estimator. When ε = 0, the convergence of the observer error is an exponential one. In the
case where ε ≠ 0, the asymptotic error can be made arbitrarily small by choosing a
sufficiently large value of α. However, a very large value α may make the observer
sensitive to noise. Thus, the choice α of is a compromise between fast convergence and
sensitivity to noise.

4.5.3. Controller synthesis

A nonlinear adaptive controller has been designed for a multicomponent reactive
batch distillation column. This model will be referred to as “the process”. The schematic
representation of the batch rectifier is shown in Fig. 1. In the closed-loop simulation
study, the control objective is to recover the more volatile at a constant purity. The
manipulated input is the reflux flow rate. At the beginning of the operation, it is assumed
that the reboiler, all the trays, and the reflux drum are filled with the liquid feed. During
the startup period, the liquid holdup in the reflux drum remains constant due to the total
reflux condition. As assumed, at the production phase the reflux drum holdup is kept
almost constant employing a traditional proportional controller.

4.5.4. Adaptive control strategy

As stated previously, the adaptive GMC–ASE controller consists of the GMC and
an ASE. The important features of its application to the batch column are:
• The nonlinear batch distillation model can be directly inserted into the control structure,
permitting for the inherent process nonlinearity to be taken into account.
• The relationship between feed forward and feedback control is explicitly accounted for
in the GMC controller.
• This control algorithm allows us to regularly update the parameters in the predictor
model. As a consequence, the effects of the structural mismatch on the controller
performance can be minimized.
• The mathematical formulation and tuning of the adaptive GMC–ASE controller, even
for this large multicomponent distillation system, are relatively simple.

44
4.5.5. Generic model controller
The component mass balance equation for the condenser accumulator system is
given as
d ( M DxD , j )
= V nTy nT , j − ( R + D ) x D , j + γ jrD , jε D (4.42)
dt
Actually, a level controller is employed to maintain the liquid holdup in the reflux
drum at a desired value. The variation of holdup is so small (±1.0%) that it is reasonable
to assume constant MD. Accordingly, the above equation becomes
dxD , j VnTynT , j − ( R + D ) xD , j + γ irD, jε D
= (4.43)
dt mD
Using Eq. (4.42) and simplifying, the following form of GMC controller equation
can be obtained for the concerned process:
t
VnTynT + γ jrD , jε − MD( K 1e + K 2 ∫ edt )
R= 0
−D (4.44)
xD, j
where e = xDsp,j − xD,j , and xDsp,j is the set point value of xD,j . It is obvious from the above
controller Eq. (4.44) that the component vapor flow rate leaving top tray (VnT ynT, j), a
poorly known parameter, is required to estimate for the implementation of the GMC
algorithm. Notice that to obtain satisfactory controller tuning parameters for the closed-
loop batch distillation operation, we need to follow the conventional startup procedure
(Barolo and Berto, 1998b).
4.5.6. Adaptive state estimator
In the present study, the product composition (xD,j) is assumed as measured
variables (true state), whereas the component vapor flow rate leaving top tray (VnT ynT ,j)
and rate of reaction rD are the extra states having no dynamics. It is important to mention
here that although xD,j is obtained through direct measurement, that composition is also
estimated in the ASE to compute the residual, ˆx − x (=estimated value − measured
value).
The predictor model, which is required to design the ASE 1, consisted of the
component mass balance equation around the condenser–reflux drum system and the
extra state equation with no dynamics. The mathematical representation of the predictor
model is

45
dxD, i VnTynT , i − RxD, i − DxD, i + γ i.rD, i.ε D
= (4.45.a)
dt MD
d (V nT y nT )
=0 (4.45.b)
dt
The final structure of the ASE estimator can be obtained in matrix form by combining
Eqs. (4.40), (4.41) , (4.45.a) and (4.45.b) as
⎡ d x^ D , i ⎤
⎢ ⎥ ⎡ 1 ⎤ ⎡ x^ D , i ⎤ ⎡ − RxD , i − DxD , i + γ i.rD , i .ε D ⎤ ⎡ 2α1 ⎤
⎢ dt ⎥ = ⎢0 ⎥+⎢ ⎥ −⎢ ⎡ x^ D , i − xD , i ⎤
⎢ ^ ^ ⎥ ⎢ mD ⎥ ⎢ MD ⎥ ⎢⎣ ⎥⎦
⎥ ⎢^ ^ ⎥ ⎢ ⎥ ⎣⎢α 22MD ⎦⎥
⎢ d V nT y nT ⎥ ⎣ 0 0 ⎦ ⎣V nT y nT ⎦ ⎣ 0 ⎦
⎢⎣ dt ⎥⎦

(4.47)
^ ^ ^
where x D , i and V nT y nT are estimates of xD,i and VnT ynT,i , respectively. In the above
estimator structure, α1 and α2 are the tuning parameters. It is true that the dynamics of xD,i
and VnT ynT ,i are not same. As a consequence, it is better to estimate xD,j and VnT ynT ,i ,
using different tuning parameters, respectively. The values of these parameters have been
determined based on the guidelines suggested by Farza et al. (1999).

In this study, another Estimator ASE 2 is designed with known parameter, rate
of raction (rD), along with VnT ynT . Such attempts have been made in order to provide a
better test scenario for the proposed procedure. It is worth mentioning that xD is directly
obtained from the sensor model, composition of distillate is estimated in the ASE 2 to
compute the residual ( ˆx − x). The predictor model, which is required to design an
adaptive state estimator, consists of only two balance equations around the reflux-
condenser system. The mathematical representation of the predictor model is

46
Sub system 1:

⎡ d x^ D , i ⎤
⎢ ⎥ ⎡0 1 ⎤ ⎡ x^ D , i ⎤ ⎡ − RxD , i − DxD , i + γ i.rD , i .ε D ⎤ ⎡ 2α 1 ⎤
⎡x^ D, i − xD, i ⎤
⎢ dt ⎥=⎢ mD ⎥ ⎢ ⎥ +⎢ ⎥ −⎢ 2 ⎥
⎢ ^ ^ ⎥ ⎢ ⎥ ⎢^ ^ ⎥ ⎢ MD
⎥ ⎢⎣α 2MD ⎥⎦ ⎢⎣ ⎥⎦
⎢ d V nT y nT , i ⎥ ⎣0 0 ⎦ ⎣V nT y nT , i ⎦ ⎣ 0 ⎦
⎢⎣ dt ⎥⎦

(4.48)
Sub system 2:

⎡ d x^ D , i ⎤ ⎡^ ^
⎤ ⎡ 2α 1 ⎤
⎡ γ i.ε D ⎤ ⎡ x D , i ⎤ ⎢V nT y nT − RxD, i − DxD, i ⎥ ⎢
^
⎢ ⎥ 0 ⎥ ⎡ x^ D, i − xD, i ⎤
⎢ dt ⎥ = ⎢ mD ⎥ ⎢ ^ ⎥ + ⎢ − ⎢ α MD ⎥
2

⎢ ⎥⎢ ⎥ M D ⎥ 3 ⎢⎣ ⎥⎦
⎢ ^ ⎥ 0 ⎦⎣ r ⎦ ⎢ ⎥⎦ ⎢⎣ γ j .ε D ⎥⎦
⎣0 ⎣ 0
⎣⎢ r ⎦⎥
(4.49)

In this estimator α1, α2, α3 are tuning parameters

47
4.6. Simulation Results and Discussion
In the batch reactive distillation operation, first the column may be brought to the
steady state by considering the total reflux startup procedure with no reaction. Ethanol is
the lightest component in system reactants, ethanol reaches to top of the column and in
distillate ethanol mole fraction reaches to steady state 0.99385 at top are shown in the
Fig. 10a. Once reactants in the column reaches to steady state then reaction will take
place in the column.
We have carried out several numerical simulations with the example presented by
Mujtaba and Macchietto. Fig.10b and Fig. 10c. Shows that open loop performance of the
composition for total reflux ratio and internal reflux ratio rf (R/VnT). It is noted that the
concentration of the reactant acetic acid in the distillate flow goes to almost zero
immediately.
Reaction mechanism becomes more important than the distillate separation
mechanism. For total reflux operation, the maximum achievable ethyl acetate
concentration is 0.935 mole fraction. This value imposes a limit in the achievable product
purity under batch operation. In fact, for batch operation (0 < rf < 1), the ethyl acetate
mole fraction increases in the first part of the batch time, achieves a maximum value, and
then decreases until the end of the operation. The first part of the batch operation where
the ethyl acetate mole fraction increases can be called the reaction phase because the
chemical reaction is the main drive of ethyl acetate in the distillate product. Compared
with nonreactive batch distillation where the mole fraction of the more volatile
component decreases along the operation of the uncontrolled batch distillation, the last
part of the batch operation can be called the separation phase. In fact, the process is
controlled by the separation during the second phase. It is also noted that the smaller the
reflux ratio, the smaller the operating time where the maximum ethyl acetate mole
fraction is achieved. Hence, inefficient batch operation is obtained with smaller values of
the reflux ratio.

48
4.6.1. Open Loop Performance of Estimators
Comparison of the estimated outputs (ASE 1, ASE 2) and process outputs with
out any change for production stage are shown in Fig 11.
Disturbance in Heat input to reboiler
The performance of the designed state observer has been tested considering +10%
and then -10% step change in the heat input to the reboiler (from 3200 kJ/min to 3520
kJ/min at time = 2000 min and then from 3520 kJ/min to 3200 kJ/min at time=3000 min).
In Fig.12 shown good agreement between the process outputs and estimator outputs.

Uncertain tray efficiency

Fig.13. compares the estimated outputs and true process outputs with two
consecutive step changes, +10% and −10%, in the Murphree e tray efficiency (from 0.75
to 0.825 at time 2000 min and then from 0.825 to 0.75 at time 3000 min). In the present
situation, the ASE estimator again confirmed its convergence ability.

Initialization error

The initialization error performance of the ASE scheme has been shown in Fig 14
comparison of the estimated outputs (ASE 1, ASE 2) and process outputs with 10%
initialization error in VnT ynT,3 (from 0.8692 to 0.9561 ). It is obvious from the figures that
good convergence against the initialization error is achieved by the proposed observation
approach.
The initialization error performance of the ASE scheme has been shown in Fig.15
comparison of the estimated outputs (ASE 2) and process outputs with 10% initialization
error in both VnT ynT,3 (from 0.8692 kmol/min to 0.9561 kmol/min) and rD (from
0.00000417 kmol/lit.min to -0.00000458 kmol/lit.min).

49
4.6.2. Constant composition control
In the present study, the set point composition has been fixed at a value of 0.934
for product ethyl acetate (lightest component). The column is started up as usual and
product withdrawal is begun as soon as this component met the composition
specification. The time taken to withdrawal the product is 5400 min with purity of 0.934.
shown in fig.17. Comparative performance between the GSPI and GMC controllers based
on the ISE values.
Closed-loop servo performance of GSPI and GMC control algorithms with two
consecutive step changes in XD,3 from 0.934 to 0.9 at time 2000 min and then from 0.9
to0.934 at time 3000 min shown in fig. 18.
Closed-loop with regulatory performance (+10% step change in set point) at time
2000 min and 3000 min. shown in fig.19.

50
Open-Loop Performance

1.1
0.9
XD, i (mole fraction)

0.7 Acetic Acid

0.5 Ethanol
Ethyl Acetate
0.3
Water
0.1
-0.1
0 1 2 3 4 5 6 7 8 9 10 11
Time (min)

0.5
0.4
XB, i (mole fraction)

Acetic Acid
0.3 Ethanol
Ethyl Acetate
0.2
Water
0.1
0
-0.1
0 1 2 3 4 5 6 7 8 9 10 11
Time (min)

Fig. 9. Steady state response of batch reactive distillation without reaction.

51
 1.1

0.9

X D, i (mole fraction)
Acetic Acid
0.7
Ethanol
0.5 Ethyl Acetate
0.3 Water

0.1

-0.1
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500
Time (min)

Acetic Acid
0.8
Ethanol
0.7
XB,I (mole fraction)

Ethyl Acetate
0.6
Water
0.5
0.4
0.3
0.2
0.1
0
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500
Time (min)

Fig. 10a. open-loop response of the batch reactive distillation column with total reflux
ratio. (First 10 min without reaction)

52
 Acetic Acid
1.2
Ethanol
XD, i (mole fraction) 1
Ethyl Acetate
0.8 Water
0.6
0.4
0.2
0
-0.2
0 50 100 150 200 250 300 350 400 450 500 550 600 650
Time (min)

1.2 Acetic Acid

Ethanol
1
Ethyl Acetate
XB,I (mole fraction)

0.8 Water
0.6
0.4
0.2
0
-0.2
0 50 100 150 200 250 300 350 400 450 500 550 600 650
Time (min)

Fig. 10b. open-loop performance of the batch reactive distillation column with start-up
phase and production phase (RR=0.8 with chemical reaction)

53
 1
XD,3 (mole fraction) Process
0.8
ASE 1
0.6 ASE 2

0.4

0.2

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

0.5
VnT y nT,3 (kmol/min)

0.4 Process
ASE 1
0.3
ASE 2
0.2

0.1

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1.00E-04
Process
ASE 2
rD (kmol/lit.min)

5.00E-05

0.00E+00

-5.00E-05

-1.00E-04
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

Fig. 11. Comparison of the estimated outputs (ASE 1, ASE 2) and process outputs with
out any change for production stage.

54
 1
XD,3 (mole fraction) Process
0.8
ASE 1
0.6 ASE 2

0.4

0.2

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

0.5

Process
VnT y nT,3 (kmol/min)

0.4
ASE 1
0.3 ASE 2

0.2

0.1

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1.00E-04
Process
ASE 2
rD (kmol/lit.min)

5.00E-05

0.00E+00

-5.00E-05

-1.00E-04
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

Fig. 12. comparison of the estimated outputs (ASE 1, ASE 2) and process outputs with
two consecutive step changes in heat input to reboiler (from 3200 kJ/min to 3520 kJ/min
at time 2000 min and then from 3520 kJ/min to 3200 kJ/min at time 3000 min).

55
 1
XD,3 (mole fraction) Process
0.8 ASE 1
0.6 ASE 2

0.4

0.2

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

0.5
VnT y nT,3 (kmol/min)

0.4
Process
0.3 ASE 1
ASE 2
0.2

0.1

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1.00E-04
Process
ASE 2
rD (kmol/lit.min)

5.00E-05

0.00E+00

-5.00E-05

-1.00E-04
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

Fig. 13. comparison of the estimated outputs (ASE 1, ASE 2) and process outputs with
two consecutive step changes in tray efficiency (from 0.75 to 0.825 at time 2000 min and
then from 0.825 to 0.75 at time 3000 min).

56
 1
XD,3 (mole fraction)
0.8
Process
0.6 ASE 1
ASE 2
0.4

0.2

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1
VnT y nT,3 (kmol/min)

0.8
Process
0.6 ASE 1
ASE 2
0.4

0.2

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1.00E-04
Process
ASE 2
rD (kmol/lit.min)

5.00E-05

0.00E+00

-5.00E-05

-1.00E-04
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

Fig. 14. comparison of the estimated outputs (ASE 1, ASE 2) and process outputs with
10% initialization error in VnT ynT,3 (from 0.8692 to 0.9561 ).

57
 1
XD,3 (mole fraction)
0.8 Process
ASE 2
0.6

0.4

0.2

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1
VnT ynT,3 (kmol/min)

0.8

0.6
Process
0.4
ASE 2
0.2

0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1.00E-04
Process
ASE 2
rD (kmol/lit.min)

5.00E-05

0.00E+00

-5.00E-05

-1.00E-04
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

Fig. 15. comparison of the estimated outputs (ASE 2) and process outputs with 10%
initialization error in both VnT ynT,3 (from 0.8692 kmol/min to 0.9561 kmol/min) and rD
(from -0.00000417 kmol/lit.min to -0.00000458 kmol/lit.min).

58
Closed-loop performance

0.935
XD,3 /XDSP,3 (mole fraction)

0.9345
0.934
Set Point
0.9335
GSPI
0.933 GMC
0.9325
0.932
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1.13
Reflux Rate (kmol/min)

1.12
1.11
1.1
1.09
1.08
1.07 GSPI
1.06 GMC
1.05
1.04
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

Fig. 16. Comparative closed loop performance of GSPI and GMC control algorithms for
production stage.

59
Servo performance

0.94
XD,3 /X DSP,3 (mole fraction)

0.93
0.92
0.91 Set Point
GSPI
0.9
GMC
0.89
0.88
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1.6
Reflux Rate (kmol/min)

1.4
1.2
1
0.8
GSPI
0.6
GMC
0.4
0.2
0
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

Fig. 17. Comparative closed loop performance of GSPI and GMC control algorithms with
two consecutive step changes in XD,3 (from 0.934 to 0.9 at time 2000 min and then from
0.9 to0.934 at time 3000 min).

60
Regulatory Performance

0.939
XD,3 /XDSP,3 (mole fraction)

0.938 Set Point

0.937 GSPI
0.936 GMC
0.935
0.934
0.933
0.932
0.931
0.93
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

1.25
Reflux Rate (kmol/min)

1.2 GSPI
GMC
1.15

1.1

1.05

1
1400 1900 2400 2900 3400 3900 4400 4900 5400
Time (min)

Fig. 18. Comparative Regulatory performance of GSPI and GMC control algorithms
with two consecutive step changes in reboiler heat duty (from 3200 kJ/min to 3520
kJ/min at time 2000 min and then from 3520 kJ/min to 3200 kJ/min at time 3000 min).

61
Tuning parameters

Table. 9. Tuning parameters for ASE 1 and ASE 2

Estimator α1 α2 α3

ASE 1 110 20 ----

ASE 2 110 20 0.0055

Table. 10. GSPI and GMC controller tuning parameters

Controller kco τi K1 K2
GSPI 1.16 0.15 ---- ----
GMC ---- ---- -20 0.000012

Table. 11. ISE values for comparison of ASE 1 and ASE 2

ASE 1 ASE 2
XD,3 VnT ynT,3 XD,3 VnT ynT,3 rD
Closed loop 0.00000626 0.0001088 0.00000121 0.000129 0.00000731
8 9
Q change 0.00004418 0.00011275 0.0000389 0.000258 0.00001371
Efficiency 0.00004395 0.00011126 0.0000369 0.000241 0.00001385
change
Initial 0.00004393 0.00011124 0.0000353 0.000235 0.00001412
change in 6
VnT ynT
Initial ____ _____ 0.00003873 0.0002805 0.00001416
change in
VnT ynT & rD

Table. 12. ISE values for comparison of GSPI and GMC

GSPI GMC

Closed loop 0.000013578 0.0000022574

Regulatory Performance 0.00010325 0.00002173

Servo Performance 0.00986812 0.00122158

62
 Chapter -5

Conclusions and Scope of future work

Conclusion

GSPI controller strategy is proposed for the control of constant composition

operations of a multicomponent batch distillation column compare to other two NLPI and
PI controllers. An adaptive control strategy is proposed for the control constant.
Composition operations of multi component reactive batch distillation column. Structural
and parametric mismatches were considered between the actual process and the predictor
model in order to provide a realistic test scenario for the proposed strategy. In this study,
the open-loop performance of the nonlinear adaptive estimator was inspected. Despite
structural discrepancy, disturbance, and parametric uncertainty, the observation scheme
provided sufficiently fast convergence of the estimation error towards zero. The GMC
control strategy showed relatively better performance than the GSPI controller for
constant composition control in reactive batch distillation.

Future Directions

1) Testing of the proposed adaptive (GMC-ASE) controller on a real time batch

reactive distillation process.
2) In corporation of dead-time and feed forward disturbance comparison.
3) Comparison with a widely used nonlinear model predictive controller ( NLMPC)

63
 Notation

MB =liquid holdup in still pot (kmol)

MD =liquid holdup in reflux drum (kmol)
Mn =liquid holdup in the nth tray (kmol)
nT =total number of trays
QR =Heat input to the still pot(kJ/min)
R=Reflux flow rate (kmol/min)
RS =steady state value of R, (kmol/min)
VB =vapor boil-up rate (kmol/min)
Vn=vapor flow rate of vapor leaving nth tray (kmol/min)
VnT =vapor flow rate of vapor leaving top tray (kmol/min)
D=distillate flow rate (kmol/min)
Ln=liquid flow rate of liquid leaving the nth tray (kmol/min)
xB,i=composition of component i in the still
xD,i=composition of component i in the Distillate
xn,i=composition of component i in Liquid stream leaving the nth tray
r,i=rate of reaction of component (kmol/lit.min)
ε=volume of catalyst (lit)
k1, k2 =rate constants
K1, K2= GMC controller tuning parameters
vb =volume of reboiler (lit)
vd=volume of reflux drum (lit)
v=volume of tray (lit)
rf=multiplication factor (rf=1 for reactive section , rf=0 for non reactive section)
γi,m=stoichiometric coefficient of i th component of m th reaction.
α1, α2, α3=estimator tuning parameters.
rD,i=rate of reaction of component i in distillate.
rB,i= rate of reaction of component i in bottom.
rn,i = rate of reaction of component i in nth tray.

64
 References

Alvarez-Ramirez, J. (1999) “Adaptive control of feedback linearizable systems, a

modeling error compensation approach,” International Journal of Robust and
Nonlinear Control , 9, 361–377.
Albet, J.M., Le Lann, J.M., Joulia, X. and Koehret, B., (1991). “Rigorous simulation
of multicomponent multisequence batch reactive distillation,” Computer-Oriented
Process Engineering, 75, 75–80.
Amiya K. Jana, (2007). “Synthesis of nonlinear adaptive controller for a batch
distillation,” ISA Transactions,46 , 49–57.
Barolo, M., Berto, F. (1998). “Composition control in batch distillation: binary and
multicomponent mixtures,” Industrial & Engineering Chemistry Research, 37,
4689–4698.
Bequette, B.W. (2006). “Process Control: Modeling, Design, and Simulation,”
Prentice-Hall, Inc., New Delhi, 359–362.
Bonsfills, L. Puigjaner. (2004). “Batch distillation: simulation and experimental
validation,” Chemical Engineering and Processing, 43, 1239–1252.
Cuille, P. E., Reklaitis, G. V. (1986). “Dynamic simulation of multicomponent batch
rectification with chemical reaction”. Comput. Chem. Eng. 10, 389-398.
Galindez, H., Fredenslund, A.A., (1988). “Simulation of multicomponent batch
distillation processes,” Computers and Chemical Engineering,12, 281–288.
Gundale Mangesh Murlidhar, Amiya Kumar Jana, (2006). “Nonlinear adaptive
control algorithm for a multicomponent batch distillation column,” Chemical
Engineering Science, 62, 1111 – 1124.
Henley, E.J., Seader, J.D., (1981). “Equilibrium-Stage Separation Operations in
Chemical Engineering,” John Wiley & Sons ,Inc.,2nd edn., New York,714-724.
Jana, A.K., Samanta, A.N., Ganguly, S., (2005). “Globally linearized control system
design of a constrained multivariable distillation column,” Journal of Process
Control, 15, 169–181.
Luyben, W.L., (1990). “Process Modeling, Simulation and Control for Chemical
Engineers,” McGraw-Hill, 2nd edn., Singapore, 72-74,133–138 and157-162.

65
Mujtaba, I. M, Macchietto, S. (1996). “Simultaneous optimization of design and
operation of multicomponent batch distillation column-single and multiple
separation duties,” J. Process Control , 6, 27.
Mujtaba, I. M., Macchietto, S. (1997). “Efficient optimization of batch distillation
with chemical reaction using polynomial curve fitting techniques,” Ind. Eng.
Chem. Res., 36, 2287.
Rosendo Loperena and Alvarez Ramirez.J, (2000). “Output feedback control of
Reactive batch distillation columns,” Process Design and Control, 39,378-386.
Sorensen, E., Skogestad, S. (1994) “Control strategies for reactive batch distillation,”
Journal of Process Control, 4, 205-217.
Venkateswarlu, C., Avantika, S.,(2001). “Optimal state estimation of multicomponent
batch distillation,” Chemical Engineering Science 56, 5771–5786.
Urmila M.Diwekar (1996). “Batch Distillation Simulation, Optimal Design and
control,” Taylor & Francis, Washington, pp 35-52.
Yamuna, K.R., Gangiah, K., (1991). “Adaptive generic model control: dual
composition control of distillation,” A.I.Ch.E. Journal 37, 1634–1642.

66
Appendix. A. Data for Batch Distillation system
Table 1: Constants of enthalpy equations (29&30)
Constants Cyclohexane Toluene Chlorobenzene
Liquid Enthalpy
a 0 0 0
b -220.6 140.1 -1307
c 1.559 -7.615×10-2 7.669
d -3.141×10-3 2.317×10-4 -1.799×10-2
e 2.672×10-6 -2.769×10-13 1.587×10-5
Vapor Enthalpy
a 4.126×104 4.852×104 5.118×104
b -203.0 132.7 -1332
c 1.284 -0.2389 7.623
d -2.55×10-3 5.907×10-4 -1.791×10-2
e 2.084×10-6 -3.914×10-7 1.575×10-5

Table .2: Density, molecular weight and boiling point data

Property Cyclohexane Toluene Chlorobenzene
B.P. (oC) 80 110.8 132.1
ρ (gm/cc) 0.779 0.866 1.107
ρ (lb/ft3) 48.63 54.06 69.11
M.W 84.16 92.13 112.56

Table .3: Antoine’s constants

Constant Cyclohexane Toluene Chlorobenzene
A 15.7527 16.0137 16.0676
B 2766.63 3096.52 3295.12
C -50.50 -53.67 -55.60

67
Appendix. B. Data for Reactive Batch Distillation system
Table .1: Antoine’s constant
Constant Ethanol Acetic Acid Ethyl Acetate Water
A 16.80 18.912 16.1516 18.3086
B 3405.57 3803.98 2790.5 3816.44
C -56.34 -41.68 -57.15 -46.19

Table .2: Density, molecular weight and boiling point data

Property Ethanol Acetic Acid Ethyl Acetate Water

B.P. (K) 351.5 391.1 350.3 373.2
ρ (gm/cc) 0.789 1.049 0.897 1
ρ (lb/ft3) 65.486 49.255 62.4278 55.997
M.W 46.069 60.05 88.107 18

Table .3: specific heat of vapor constant

Constant Ethanol Acetic Acid Ethyl Acetate Water
a 14.6934 14.048 24.9801 7.9857
-1 -1 -1
b 0.22987x10 0.21531x10 0.33297x10 0.46331x10-3
c -0.102199x10-4 -0.21534x10-4 -0.7316x10-6 -0.14028x10-5
d 0.2589x10-8 -0.4607x10-8 -0.1247x10-8 -0.65783x10-9
e -0.8044x10-12 0.1893x10-11 0.48242x10-11 0.9895x10-13

Cp=a+bT+cT2+dT3+eT4
HV=mcpΔT
HL=HV-λ
⎡ B ⎤
Latent heat λ= RT 2 ⎢ 2⎥
⎣⎢ ( C + T ) ⎦⎥

68