Chapter 3

Mathematical Modelling of Batch Distillation Columns:

A Comparative Analysis of Non-Linear and Fuzzy
Models

Adriana del Carmen Téllez-Anguiano,

Mario Heras-Cervantes, Juan Anzurez-Marín,
Gerardo Marx Chávez-Campos and
José Antonio Gutiérrez Gnecchi

Additional information is available at the end of the chapter

http://dx.doi.org/10.5772/66760s

Abstract
Distillation is the process most commonly used in industry to separate chemical mix-
tures; its applications range from cosmetic and pharmaceutical to petrochemical indus-
tries. The equipment required to perform the distillation process is known as distillation
column. Since initial investment and maintenance costs for distillation columns are very
high it is necessary to have an appropriate mathematical model that allows improving
the comprehension of the column dynamics, especially its thermal behaviour, in order to
enhance the control and safety of the process. This chapter presents a general panorama
of the mathematical modelling of distillation columns, having as a specific case of study
the comparison of a space-state non-linear model and a Takagi-Sugeno fuzzy model for
a batch distillation column using a binary mixture (Ethanol-Water).

Keywords: mathematical modelling, distillation column, Takagi-Sugeno, non-linear

models

1. Introduction

Distillation is the process most commonly used in industry to separate chemical mixtures,
being the petrochemical industry one of the most important due to that oil distillation allows
obtaining useful product, such as fuels. Distillation is also widely used in the pharmaceutical
and cosmetics industry in order to obtain specific drugs and in the liquor industry to obtain
wines and liquors, among other applications [1].

© 2017 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.
58 Distillation - Innovative Applications and Modeling

Distillation columns are the essential equipment required to perform the distillation process,
these columns allow producing food, fuel, medicine, among other products. However, distil-
lation columns represent an important investment in the process they are used, that is why it is
necessary to have both, corrective and predictive maintenance, in order to prevent failures in
the process as well as in the equipment.

Through the computational and technological continuous development, the industrial pro-
cesses, such as distillation, have become very complex systems due to the high number of
components they have and the several functions they develop, so their vulnerability has also
increased. Having appropriate techniques to model distillation columns, such that these
models allow implementing efficient and reliable control techniques, is very important to
obtain the desired product quality, the adequate process functioning and to improve the
security of the system and the user.

In the literature, different mathematical models have been used to improve distillation col-
umns dynamics and comprehensions have been reported. Simple linear and non-linear models
are representations that consider only few variables and low-order equations, simplifying the
design and implementation of controllers using computational tools. Kienle [2] presents a low-
order model for an ideal multicomponent distillation process considering the non-linear wave
propagation theory.

Balasubramhanya and Doyle Iii [3] present a low-order model for a reactive multicomponent
distillation column as well as the designing of a (MPC) predictive control to obtain the best quality
of the distilled product. In Ref. [4], a model based on neural networks having the aim of optimiz-
ing the energy efficiency in a binary distillation column is presented. Lopez-Saucedo et al. [5]
present the simulation and optimization of a model for a conventional and nonconventional batch
distillation column.

Astorga et al. [6] and Cervantes et al. [7] present high-gain observers to estimate the light
component composition in a continuous distillation column using a set of models for each
plate of the column. In Ref. [8], a fault tolerant scheme for a distillation column, where
observers are used to detect failures in the temperature sensors considering a non-linear model
of the distillation column, is presented. The parametric identification is other methodology
used to estimate certain variables in distillation columns as presented in Refs. [9, 10].

The Takagi-Sugeno fuzzy model is a useful tool to model and control complex systems
based on the concept of decomposing a non-linear model in a multi-model structure formed
by linear models not necessarily independents and fuzzy logic [11, 12], where the non-
linear system representation is obtained through a weighted sum of all the sub-systems.
The Takagi-Sugeno fuzzy model provides a solution to solve the designing and implemen-
tation issues in control strategies for non-linear systems, for instance, Wang et al. [13]
propose a methodology to design control techniques for systems having a Takagi-Sugeno
form.

The stability analysis of the Takagi-Sugeno fuzzy model can be solved considering the
Lyapunov approach and by using the inner point tool as well as optimization techniques based
on linear matrix inequalities (LMIs) [14].
 Mathematical Modelling of Batch Distillation Columns: A Comparative Analysis of Non-Linear and Fuzzy Models 59
http://dx.doi.org/10.5772/66760s

In this chapter, the design and simulation of a non-linear state-space and Takagi-Sugeno models
for a batch distillation column are presented. These models are simulated and compared in order
to analyse if they aim the objective of representing adequately the process dynamics in order to
facilitate the implementation of control strategies to improve the distilled product quality as well
as the process security.

2. Distillation column operation modes

Due to the variety of substances found in the nature and their different phases (mainly liquid
and vapour), there exist different distillation operation modes in order to separate diverse
mixtures, obtaining different quality of products.
The main distillation operation types are as follows:
• Vacuum distillation: A low-pressure system is used in order to obtain a low-temperature
boiling of the substances in the mixture. Usually, a vacuum pump is used to generate the
low-pressure state, as shown in Figure 1a.
• Destructive distillation: The substance is heated at high temperatures to be decomposed in
other products that can be separated by fractionating, its operation is similar to the one
used in wood and coal, as shown Figure 1b.
• Extractive distillation: Different separation agents are added to azeotropic mixtures, alter-
ing the relative volatility of the mixture components in order to allow their separation (see
Figure 1c).
• Fractionating distillation: Liquid mixtures are separated by heating, considering a high heat
exchange and the liquid and vapour molar rates. This distillation is used to separate
composite mixtures/substances having different but close boiling temperatures. It usually
considers a continuous operation, having a constant feeding flow through a feeding tray.
The section above the feeding tray is named rectifying section, under the feeding tray is
called stripping section, as shown in Figure 2a.
• Batch distillation: Widely used in industry when having small liquid quantities or when
obtaining different products from a single mixture load is required. This operation does
not have steady state due that the mixture composition varies in time; besides, it only
allows enriching or rectifying the distilled (lighter) product (see Figure 2b).
In general, the different distillation operation modes have the same operating principle, mainly
due the physical variables that interact in the process, such as temperature, composition,
pressure and heating energy.
A typical distillation column is formed by a boiler, a condenser and n trays. The boiler is the
element that provides the heating energy necessary to evaporate the mixture into it. The
condenser provides the cooling necessary to condensate vapour, part of this vapour returns to
the column to enrich the mixture, the rest is obtained as a distilled product. The column body
is composed of a set of trays, where a partial separation of the mixture is performed due the
circulation of liquid and vapour flow.
60 Distillation - Innovative Applications and Modeling

Figure 1. Distillation columns operation modes: (a) vacuum, (b) destructive and (c) extractive.

Figure 2. Distillation column operation modes: (a) fractionating and (b) batch.

The vapour flow is generated by the ebullition of the mixture in the boiler, the vapour rises into
the column body and it is enriched by the light element of the mixture in each tray of the
column. The liquid flow, generated by the reflux, descends from the condenser to the boiler by
gravity and it is enriched by the heavy element of the mixture in every tray of the column. This
operation can be described by an adequate mathematical model of the process.
 Mathematical Modelling of Batch Distillation Columns: A Comparative Analysis of Non-Linear and Fuzzy Models 61
http://dx.doi.org/10.5772/66760s

3. Distillation column mathematical modelling

The main objectives of designing a mathematical model of the distillation process are to
simplify the analysis and comprehension of the distillation dynamics, facilitate the design of
control techniques to enhance the distilled product quality and the system performance,
estimate variables difficult to be measured, diagnose failures, among others. In order to deal
with these objectives development of an adequate model is indispensable.

There exist different distillation column models according to its operation, the most commonly
used in industrial applications are the continuous (fractionating) and batch models. Because of
the similarity between the continuous and the batch operating types, in this section, a generic
model that presents adequate results in both cases is presented.

It is well known that having a more complete/complex model implies having more complex
equations difficult to solve, whereas having a simpler representation implies having simpler
equations but the response resolution will have a higher calculation error compared to the real
system response.

In general, there are two main model types according their complexity: simple and complex.
The simple model found in the literature is the differential model, which considers the boiler
and the condenser as trays in the distillation column. The column dynamics is represented by
the component mass balance as shown in Eq. (1).

dWxw Wdxw xw dW
¼ þ ¼ −DyD (1)
dt dt dt

where W is the bottom product, xW is the bottom product composition, D is the distilled
product and yD is the distilled product composition.

The complex model considers each column element individually, i.e. a condenser, a boiler and
trays are modelled individually, such that the response has a better resolution.

The particular study case presented in this chapter considers a complex model of a batch
distillation column using a binary mixture.

4. Non-linear model of a binary batch distillation column

The model for a binary batch distillation column is obtained considering the light component,
this component is obtained as a final (distilled) product [15]. The light component composition
is obtained in each tray of the distillation column, where the liquid and vapour molar flows
interact.

In order to design the distillation column model, the following assumptions are considered
[16]: total condenser, no heating losses in the body column, constant pressure in the body
column, liquid and vapour phases in thermodynamic equilibrium in each plate, variable
relativity volatility according to the component composition.
62 Distillation - Innovative Applications and Modeling

The distillation column dynamics is represented by a set of differential equations that describe
the behaviour of the light component of the mixture, given by Eq. (2).

dxi Lðxi−1 −xi Þ þ Vðyiþ1 − yi Þ

¼ (2)
dt Mi

where xi is the liquid molar composition of the light component in tray i, yi is the vapour molar
composition of the light component in tray i, L is the liquid molar flow, V is the vapour molar
flow and M is the retained mass.
The phase equilibrium is determined by constant K, as shown in Eq. (3) for ideal mixtures.

xi
K¼ (3)
yi

Such that considering the vapour-liquid equilibrium (VLE) and the relative volatility, the
vapour composition as a function of the liquid composition is obtained. This function is
presented in Eq. (4).

y ¼ f ðx, αÞ (4)

This is specifically presented in Eq. (5).

αi xi
yi ¼ (5)
1 þ ðαi −1Þxi

where α is the relative volatility in tray i.

Within each element of the distillation column flow different molar rates/quantities, named
molar flows. These flows are the liquid and the vapour entering and leaving each tray, the
distilled product and the bottom product.
In a binary batch distillation column, the liquid flows in both rectifying and stripping sections
are the same, as well as the vapour flows, because there is not feeding flow.

L ¼ LS ¼ LR
(6)
V ¼ VS ¼ VR

The molar flows considered in the binary batch distillation model are four: vapour (V), liquid
(L), distilled (D) and bottom (B) products, these are expressed in Eqs. (7)–(9) [17].

QB
V¼ vap vap (7)
H1 xn þ H1 ð1−xn Þ

where QB is the heating power (input), xn is the liquid composition of light component in the
vap vap
boiler (tray n), H1 is the vaporization enthalpy of the light component and H 2 is the
vaporization enthalpy of the heavy component.
 Mathematical Modelling of Batch Distillation Columns: A Comparative Analysis of Non-Linear and Fuzzy Models 63
http://dx.doi.org/10.5772/66760s

L ¼ ð1−RÞV (8)

where R is the reflux input.

D ¼ V−L (9)

B, the bottom product, is not calculated, it is considered as the molar flow that remains into the
boiler.

The non-linear model of the binary batch distillation column presented in this chapter is based
on a set of sub-models, each sub-model corresponding to a specific element of the column
(boiler, condenser and trays).

4.1. Condenser sub-model

The condenser is numbered as tray 1. Its dynamics is described by Eq. (10).

dx1 Vy2 −Lx1 −Dx1

¼ (10)
dt M1

By substituting L ¼ ð1−RÞV in D ¼ V−L, in order to represent the condenser as a function of the

reflux, Eq. (11) is obtained.

D ¼ RV (11)

By substituting Eq. (11) in Eq. (10), Eq. (12) is obtained.

dx1 Vy2 −Lx1 −RVx1

¼ (12)
dt M1

Considering that
α2 x2
y2 ¼ (13)
1−ðα2 −1Þx2

the non-linear equation that represents the condenser dynamics is finally represented in
Eq. (14).


dx1 V α2 x2 Lx1 RVx1
¼ − − (14)
dt M1 1−ðα2 −1Þx2 M1 M1

4.2. Tray sub-model

The column body is formed by a set of n-2 trays. Eq. (15) describes its dynamics.

dxi Vyiþ2 −Vyi þ Lxi−1 −Lxi

¼ ; i ¼ 2, 3, …, n−1 (15)
dt Mi

where n is the total number of trays including a boiler and a condenser.

64 Distillation - Innovative Applications and Modeling

Considering that
αi xi
yi ¼ (16)
1−ðαi −1Þxi

the non-linear equation that represents the condenser dynamics is finally represented in
Eq. (17).



dxi V αiþ1 xiþ1 V αi xi Lðxi−1 −xi Þ
¼ − þ (17)
dt Mi 1−ðαiþ1 −1Þxiþ1 Mi 1−ðαi −1Þxi Mi

4.3. Boiler sub-model

The boiler is numbered as tray n. Eq. (18) describes its dynamics.

dxn Vxn −Vyn þ Lxn−1 −Lxn

¼ (18)
dt Mn

Factorizing Eq. (18), Eq. (19) is obtained.

dxn Vðxn −yn Þ þ Lðxn−1 −xn Þ

¼ (19)
dt Mn

Solving V to represent Eq. (19) as a function of the heating power (QB) based on Eq. (7), Eq. (20)
is obtained.
!

dxn QB xn −yn Lðxn−1 −xn Þ
¼ vap vap þ (20)
dt H1 xn þ H1 ð1−xn Þ Mn Mn

Then, considering
αn xn
yn ¼ (21)
1−ðαn −1Þxn

the non-linear equation that represents boiler dynamics is finally represented in Eq. (22).
" !

#
dxn QB xn 1−αn Lðxn−1 −xn Þ
¼ vap vap þ (22)
dt H 1 xn þ H 1 ð1−xn Þ Mn 1−ðαn −1Þxn Mn

4.4. State-space non-linear model for a binary batch distillation column

In this section, the distillation column sub-models shown in Eqs. (14), (17) and (22) are presented
in a state-space representation having the form shown in Eq. (23).
x_ ¼ Ax þ Bu (23)

This representation is used in a specific study case, a 12-tray distillation column including
a boiler and a condenser, using a binary mixture in a batch operation. Compositions
 Mathematical Modelling of Batch Distillation Columns: A Comparative Analysis of Non-Linear and Fuzzy Models 65
http://dx.doi.org/10.5772/66760s

x ¼ ½x1 , x2 , …, x12  are considered as states of the model and u ¼ ½R, QB T as inputs of the model.
Matrices A and B are shown in Eqs. (24) and (25), respectively.
2 3
L Vf ðx2 , α2 Þ
6 − M1 M1
0 0 ⋯ 0 0 7
6 7
6 7
6 L −L−Vf ðx2 , α2 Þ Vf ðx3 , α3 Þ 7
6 0 ⋯ 0 0 7
6 M 7
6 2 M2 M2 7
6 7
A¼6
6 0 −L−Vf ðx3 , α3 Þ Vf ðx4 , α4 Þ 7 (24)
0 7
L
6 ⋯ 0 7
6 M3 M3 M3 7
6 7
6 ⋮ ⋮ ⋮ ⋮ ⋱ ⋮ ⋮ 7
6 7
6 7
4 L L 5
0 0 0 0 ⋯ −
M12 M12
2 3
Vx1
6 M1 0 7
6 7
6 7
6 0 0 7
6 7
6 7
6 7
B¼6 ⋮ ⋮ 7 (25)
6 7
6 0 0 7
6 7
6 7
6 x12 −f ðx12 −α12 Þ 7
4 0   5
vap vap
H1 x12 þ H 1 ð1−x12 Þ M12

5. Takagi-Sugeno fuzzy model for a binary batch distillation column

The Takagi-Sugeno fuzzy representation describes the system dynamics based on linear sub-
models interpolation and fuzzy rules [18].

Rule for model j:

If z1 ðtÞ is M1j j, z2 ðtÞ is M2j ,… and zp ðtÞ is Mpj

Then:
r
xðtÞ ¼ ∑ Aj xðtÞ þ Bj uðtÞ (26)
j¼1

where j = 1, 2,…, r, Mj is the fuzzy set, r is the sub-model number, x is the state vector, u is the
input vector, Aj is the state matrix for sub-model j, Bj is the input matrix for sub-model j and
zj ðtÞ is the scheduling measurable variable (state variables or external disturbances).

Given ½xðt, Þ, uðtÞ, zðtÞ, the complete fuzzy model is obtained by using a singleton-type
fuzzifier, a product-type defuzzifier mechanism and the gravity centre. The Takagi-Sugeno
fuzzy model for the non-linear system is expressed in Eq. (27).
66 Distillation - Innovative Applications and Modeling

 
∑rj¼1 ωj zj ðtÞ ½Aj xðtÞ þ Bj uðtÞ
_ ¼
xðtÞ   (27)
∑rj¼1 ωj zj ðtÞ

 
where the weight ωj zj ðtÞ is 0 or a positive value, such that the sum of all the weights is
positive; thus, the normalized weight, hi, is calculated in every rule from the zj membership
functions in the Mjk set. It is well known by fuzzy logic that hj ¼ hj ½zðtÞ ≥ 0 and
∑rj¼1 hj ½zj ðtÞ ¼ 1, as expressed in Eq. (28).
 
ωj zj ðtÞ
hj ½zj ðtÞ ¼   (28)
∑rj¼1 ωj zj ðtÞ

The system expressed in Eq. (27) is equivalent to the system in Eq. (29).

_ ¼ ∑rj¼1 hj ½Aj xðtÞ þ Bj uðtÞ

xðtÞ (29)

5.1. Application to a binary batch distillation column

In this chapter, the specific study case is a 12-tray distillation column, including a boiler and a
condenser, using an ethanol-water mixture in a batch operation. In the Takagi-Sugeno fuzzy
model the liquid (L) and vapour (V) molar flows are proposed as parameters; the nominal
operating ranges in steady state are:

L ¼ ½0:418783, 2:97801
(30)
V ¼ ½0:418783, 2:97801

According to these parameters, the Takagi-Sugeno fuzzy model that interpolates between four
linear models based on the following rules is obtained:

Rule 1:

if V is V min and if L is Lmin (31)

Then:

x_ 1 ðtÞ ¼ A1 xðtÞ þ B1 uðtÞ (32)

Rule 2:

if V is V min and if L is Lmax (33)

Then:

x_ 2 ðtÞ ¼ A2 xðtÞ þ B1 uðtÞ (34)

 Mathematical Modelling of Batch Distillation Columns: A Comparative Analysis of Non-Linear and Fuzzy Models 67
http://dx.doi.org/10.5772/66760s

Rule 3:
if V is V max and if L is Lmin (35)

Then:
x_ 3 ðtÞ ¼ A3 xðtÞ þ B2 uðtÞ (36)

Rule 4:
if V isV max and if L is Lmax (37)

Then:
x_ 4 ðtÞ ¼ A4 xðtÞ þ B2 uðtÞ (38)

where:
A1 ¼ fV min , Lmin , G{x1 , α1 }, …, G{x12 , α12 }, M1 , …, M12 g
A2 ¼ fV min , Lmax , G{x1 , α1 }, …, G{x12 , α12 }, M1 , …, M12 g
(39)
A3 ¼ fV max , Lmin , G{x1 , α1 }, …, G{x12 , α12 }, M1 , …, M12 g
A4 ¼ fV max , Lmax , G{x1 , α1 }, …, G{x12 , α12 }, M1 , …, M12 g
0 1
V min  x1
B M1 0 C
B C
B C
B 0 0 C
B C
B1 ¼ B
B ⋮ ⋮ C
C
B C
B 0 0 C
B x12 −gðx12 , α12 Þ C
@ 0 A
vap vap
ðH EtoH x12 þ H H2 O ð1−x12 ÞÞ  M12
0 1
V max  x1 (40)
B M1 0 C
B C
B C
B 0 0 C
B C
B C
B2 ¼ B
B ⋮ ⋮ C
C
B C
B 0 0 C
B C
B x12 −gðx12 , α12 Þ C
@ 0 A
vap vap
ðHEtoH x12 þ HH2 O ð1−x12 ÞÞ  M12

The membership functions (µ(z)) for the fuzzy set are determined by:

Eq. (41) for vapour V:

8
< μV V max −V
min ¼
μðVÞ V max −V min (41)
:
μV max ¼ 1−μV min

Eq. (42) for liquid L:

68 Distillation - Innovative Applications and Modeling

8
< μL Lmax −L
min ¼
μðVÞ Lmax −Lmin (42)
:
μLmax ¼ 1−μLmin

The normalized weights are given by Eq. (43):

8
> h1 ðV, LÞ ¼ μV min μLmin
>
< h ðV, LÞ ¼ μV μL
2 min max
(43)
>
> h ðV, LÞ ¼ μV max μLmin
: 3
h4 ðV, LÞ ¼ μV max μLmax

The Takagi-Sugeno fuzzy model proposed for the distillation column is represented in Eq. (44).
r
xðtÞ ¼ ∑ hi ðL, VÞðAi xðtÞ þ Bi uðtÞÞ (44)
i¼1

6. Models experimental validation and comparison

The Takagi-Sugeno fuzzy model is validated in Matlab by using experimental data from a 12-
tray batch distillation column with variable reflux, using an ethanol-water mixture and con-
sidering the characteristics presented in Table 1.

Parameter Value Units

EtOH volume in boiler 2000 mL

H20 volume in boiler 2000 mL
Process total pressure 637.42 mmHg

Table 1. Mixture initial parameters.

The initial molar composition of ethanol in the boiler is 0.2216, considering that the feed
volume corresponds to 96%Vol ethanol.
The characteristics of the process inputs for the study case, the heating power (QB) and the
reflux valve opening (R) are shown in Table 2.
Figure 3 presents the temperatures estimated by the Takagi-Sugeno model for the trays in the
column body. The temperature increment and decrement due the reflux (R) action can be seen
in all the trays.
Figure 4 presents the temperature graphics corresponding to the condenser (a) and to the
boiler (b) in the non-linear and Takagi-Sugeno model. Temperature variations existing during
the heating power (QB) and reflux changes (R) are shown. It can be seen that there exist a
difference between the results obtained by both models due the reflux action, this difference is
provoked by the fixed operating points for liquid and vapour flows in the Takagi-Sugeno
model; however, this difference is small (less than 1.5%).
 Mathematical Modelling of Batch Distillation Columns: A Comparative Analysis of Non-Linear and Fuzzy Models 69
http://dx.doi.org/10.5772/66760s

Input Signal Time

QB Step 0–800 J 0 min

R Total 0 min
QB Step 800–1000 J 3.3 min
QB Step 1000–1250 J 5.98 min
R Pulse (ton = 6 s, toff = 6 s) 12.61 min
QB Step 1250–1100 J 14.78 min
QB Step 1100–950 J 17.15 min
QB Step 950–1100 J 19.36 min
R Total 23 min
QB Step1100–1250 J 24.88 min

Table 2. Input parameters.

Figure 3. Plate temperatures in the distillation column.

In Figure 5, the composition graphics estimated for the distillation column trays by the Takagi-
Sugeno fuzzy model are presented, these composition values vary according to the tray
position.
In Figure 6, the simulation results obtained by the non-linear and Takagi-Sugeno models for
the light component composition in the condenser (a) and the boiler (b) are presented. It can be
seen that the composition behaviour in both trays varies according the heating power (QB) and
reflux (R) changes, as shown in Table 2.
70 Distillation - Innovative Applications and Modeling

Figure 4. a) Condenser and b) Boiler temperatures, non-linear and Takagi-Sugeno models.

Figure 5. Plate temperatures in the distillation column.

Figure 7 shows the liquid and vapour molar flow behaviour during the distillation process. It
can be seen the process dynamics when reflux or heating power changes exist.
The error percentage in the Takagi-Sugeno model compared to the non-linear models, calcu-
lated by the function shown in Eq. (45), is graphically represented in Figure 8. It can be seen
that the error behaviour in the condenser (a) and the boiler (b) has a maximum value of 1.5%
due to the reflux changes.
 Mathematical Modelling of Batch Distillation Columns: A Comparative Analysis of Non-Linear and Fuzzy Models 71
http://dx.doi.org/10.5772/66760s

jxTS −xNL j
e¼ 100% (45)
xNL

Figure 6. a) Condenser and b) Boiler temperatures obtained by non-linear and Takagi-Sugeno models.

Figure 7. Liquid and vapour molar flows.

72 Distillation - Innovative Applications and Modeling

Figure 8. a) Condenser and b) Boiler error percentages.

7. Conclusions

This chapter presents the analysis and design of a state-space non-linear model and the Takagi-
Sugeno fuzzy model for a batch distillation column using a binary mixture. The state-space
non-linear model is based on differential equations considering compositions, temperatures
and molar flows in the column. The linear fuzzy model is based on four rules, considering as
parameters the liquid and vapour molar flows.

Both, the state-space non-linear and the linear fuzzy models are simulated in Matlab consider-
ing real input parameters (heating power and reflux) from a 12-tray batch distillation pilot
plant using an ethanol-water mixture. The light component compositions and the tempera-
tures in each tray of the column are calculated by both models. Besides, the obtained results are
compared considering the same operating parameters, this comparison has the aim to verify
the adequate functioning of the non-linear state-space and the Takagi-Sugeno models in order
to analyse the existing differences.
The Takagi-Sugeno fuzzy model presents small differences in the estimations of the composi-
tion component and the tray temperatures when a reflux disturbance is presented due that the
reflux affects directly the operating points established in this model; however, these differences
are small enough to be neglected and both models converge under any operating condition.
The Takagi-Sugeno fuzzy model for a distillation column represents an alternative tool that
takes advantage of the fuzzy control theory, allowing to facilitate the design and implement
nonconventional control strategies for non-linear systems, however, if a higher resolution
response is required it could be convenient to consider the non-linear model.
 Mathematical Modelling of Batch Distillation Columns: A Comparative Analysis of Non-Linear and Fuzzy Models 73
http://dx.doi.org/10.5772/66760s

Author details

Adriana del Carmen Téllez-Anguiano1*, Mario Heras-Cervantes2, Juan Anzurez-Marín2,

Gerardo Marx Chávez-Campos1 and José Antonio Gutiérrez Gnecchi1

*Address all correspondence to: [email protected]

1 Electronics Department, Instituto Tecnológico de Morelia, Morelia, Michoacán, México

2 Universidad Michoacana de San Nicolas de Hidalgo, FIE, Morelia, Mich, México

References

[1] Mujtaba, I. M. Batch distillation [thesis]. London, England: Imperial College Press; 2004.

[2] Kienle, A. Low-order dynamic models for ideal multicomponent distillation processes
using nonlinear wave propagation theory. Chemical Engineering Science. 2000;55
(10):1817–1828.

[3] Balasubramhanya, L. S., & Doyle Iii, F. J. Nonlinear model-based control of a batch
reactive distillation column. Journal of Process Control. 2000;10(2):209–218.

[4] Osuolale, F. N., & Zhang, J. Energy efficiency optimisation for distillation column using
artificial neural network models. Energy. 2016;106:562–578.

[5] Lopez-Saucedo, E. S., Grossmann, I. E., Segovia-Hernandez, J. G., & Hernández, S.

Rigorous modeling, simulation and optimization of a conventional and nonconventional
batch reactive distillation column: A comparative study of dynamic optimization
approaches. Chemical Engineering Research and Design. 2016;111:83–99.

[6] Astorga, C. M., Quintero-Mármol, E., Vela, L. G., & Torres, L. Observador No Lineal para
la Estimación de Concentraciones en un Proceso de Destilación Metanol/Etanol.
Información Tecnológica. 2006;17(6):35–41.

´
[7] Cervantes, M. H., Anguiano, A. T., Ramırez, M. G., Zaragoza, C. A., & Archundia, E. R.
Validación experimental de un Observador de Alta Ganancia Reconfigurable para una
columna de destilación. Revista Iberoamericana de Automática e Informática Industrial
RIAI. 2015;12(4):397–407.

[8] Ortiz-Torres, G., Escobar, R. F., Adam-Medina, M., Astorga-Zaragoza, C. M., & Guerrero-
Ramírez, G. V. Control Tolerante a Fallas Activo en Sensores y Actuadores: Aplicación a
una Columna de Destilación. Revista Iberoamericana de Automática e Informática Indus-
trial RIAI. 2016;13(1):67–79.
74 Distillation - Innovative Applications and Modeling

[9] Zhu, Y. Distillation column identification for control using Wiener model. In: American
Control Conference, IEEE. San Diego, CA, USA 1999 . Proceedings of the 1999; 1999. pp.
3462–3466.
[10] Safdarnejad, S. M., Gallacher, J. R., & Hedengren, J. D. Dynamic parameter estimation and
optimization for batch distillation. Computers & Chemical Engineering. 2016;86:18–32.
[11] Johanson, T. A., & Murray-Smith, R. Operating regime approach to nonlinear modeling
and control. In: Multiple model approaches to modeling and control. Taylor and Francis,
UK. 1997; pp. 3–72.
[12] Angelov, P. P., & Filev, D. P. An approach to online identification of Takagi-Sugeno fuzzy
models. IEEE Transactions on Systems, Man and Cybernetics, Part B (Cybernetics).
2004;34(1):484–498.
[13] Wang, H. O., Tanaka, K., & Griffin, M. Parallel distributed compensation of nonlinear
systems by Takagi-Sugeno fuzzy model. Fuzzy Systems, 1995. International Joint Confer-
ence of the Fourth IEEE International Conference on Fuzzy Systems and the Second
International Fuzzy Engineering Symposium, Proceedings of 1995 IEEE, Yokohama,
Japan. 1995;2:531–538.
[14] Tanaka, K., & Sugeno, M. Stability analysis and design of fuzzy control systems. Fuzzy
Sets and Systems. 1992;45(2):135–156.
[15] Cingara, A., Jovanovic, M., & Mitrovic, M. Analytical first-order dynamic model of
binary distillation column. Chemical Engineering Science. 1990;45(12):3585–3592.

[16] Skogestad, S. Dynamics and control of distillation columns-a critical survey. Modeling
Identification and Control. Norwegian Society of Automatic Control. 1st ed. 1997.

[17] Aguilera Gonzalez, A., Tellez-Anguiano, A., Astorga-Zaragoza, C. M., Juárez-Romero,

D., & Quintero-Mármol, E. Observador de alta ganancia constante para una clase de
sistema no lineal de forma triangular. Revista Iberoamericana de Automática e
Informática Industrial. 2010;7(2):31–38.

[18] Takagi, T., & Sugeno, M. Fuzzy identification of systems and its applications to modeling
and control. IEEE Transactions on Systems, Man and Cybernetics. 1985;1:116–132.