2 Isc2011 2

NEURAL NETWORK MODELING OF A
REACTIVE PACKED DISTILLATION

COLUMN
Being A Paper Presentation at the
7th International Statistics Congress
By
Abdulwahab GIWA, Süleyman KARACAN

Department of Chemical Engineering,
Faculty of Engineering, Ankara
University, Ankara, Turkey
April, 2011
INTRODUCTION
Top Products
Inputs
Outputs Feed
Bottom Products
Figure 1. Some Chemical Engineering Processing Units

Top Products
Feed
Bottom Products
Figure 2. Reactive Packed Distillation Column

Top Products
Feed 1
Feed 2
Bottom Products
Figure 3. Reactive Packed Distillation Column

Why Reactive Distillation?
 Reduction of energy consumption;
 Reduction of equipment units;
 Increment in the yield of a reversible reaction;
 Breaking of azeotropes;
 Reduction of the costs of production.

Applications of Reactive Distillation
 Acetylation
 Alkylation
 Amination
 Dehydration
 Esterification
 Etherification
 Hydrolysis
 Transesterification
Esterification Reaction
k

'
f
RCOOH  R OH  RCOOR  H 2O
'
(1)
k
b
 Equilibrium reaction;
 Highly reversible;
 Slow in the absence of a catalyst.

Ethyl Acetate Production
k
 f
CH 3COOH  C2 H 5OH  CH 3COOC2 H 5  H 2O (2)
k
b *
Re action Heat  1.41E  04kJ / kgmol
Table 1. Properties of the Components
Property Acetic Acid Ethanol Ethyl Acetate Water
Molecular Formula CH3COOH C2H5OH CH3COOC2H5 H2O
Molecular Weight (kg/kgmol) 60.05 46.07 88.11 18.02
Boiling Point (oC) 117.95 78.25 77.15 100.00
Density (kg/m3) 1051.50 795.98 904.86 997.99

Source: Aspen HYSYS 3.2
Calculated with Aspen HYSYS 3.2
Uses of Ethyl Acetate
 Surface coating and thinners;
 Pharmaceuticals;
 Flavors and essences;
 Flexible packaging;
 Adhesives;
Uses of Ethyl Acetate Contd.
 Cleaning fluids;
 Inks;
 Nail-polish removers;
 Coated papers;
 Explosives;
 Artificial leather;
 Photographic films & plates.
PILOT PLANT SET-UP
Figure 4. Experimental Set-Up for Data Acquisition

r a l
e u
y N k
h
W etw g? o r
N e lin
o d
M
B e c a u s e
Problems that are difficult for
conventional computers or human
beings can be solved easily by neural
networks.
MODELING
Model Structures
NNNIO: T t   f Rt  1, , Rt  d R  (3)
NNNAR: T t   f T t  1, , T t  dT  (4)
NNNARX: T t   f Rt  1, , Rt  d R , T t  1, , T t  dT  (5)
where f is the neural network function, T is the top section

temperature, R is the reflux ratio, dR is the delay in the reflux
ratio, dT is the delay in the top section temperature and t

stands for time.
Design Process
 Data Collection;
 Network Creation;
 Network Configuration;
 Initialization of the weights and biases;
 Network Training;
 Network Validation;
 Using the Network.
Hidden Layer Output Layer (a)
T(t)
R(t) 1:3 w w
+ +
b b 1
5 Neurons 1 Neuron
Hidden Layer Output Layer (b)
T(t) T(t)
1:3 w w
+ +
b b 1
1
5 Neurons 1 Neuron
R(t) Hidden Layer (c)

Output Layer
1:3 w T(t)
1 w
T(t) 1:3 w + +
b 1
b
1 1 Neuron
5 Neurons
Figure 5. Schematic Views of Reactive Packed Distillation Column Neural Network Models: (a) NNNIO;
(b) NNNAR; (c) NNNARX
RESULTS AND DISCUSSIONS
Experimental Results
Figure 6. Experimental input-output data (Top – Estimation, Middle –

Validation, Down – Step input)
Figure 7. Time-Series Response of the Training
Figure 8. Time-Series Response of the Testing
Figure 9. Regression Plot of the Testing
Table 2. Performance Criteria of the Models
Models Fit Value Mean Squared Error Correlation Coefficient
NNIO -0.07389 0.28232608 0.000000
NNNAR 96.0627 4.3703E-04 0.999458
NNNARX 96.4358 3.5812E-04 0.999468

CONCLUSIONS
The good agreements between the measured temperatures and
those predicted using the NNNAR and NNNARX models
have revealed that they could be used to represent the reactive
packed distillation column successfully for further studies. In
addition, NNNARX model was discovered to be better among
the two due to its smaller mean squared error.

ACKNOWLEDGEMENTS
I wish to acknowledge the support received
from The Scientific and Technological
Research Council of Turkey (TÜBİTAK) for
my PhD Programme. In addition, this research
was supported by Ankara University Scientific
Research Projects.
APPRECIATION
I would like to express my sincere
appreciation to my supervisor, Prof. Dr.
Süleyman KARACAN, for his guidance and
suggestions towards the successful
completion of this work.

References
Balasubramhanya L. and Doyle III F.J. (2000),
Nonlinear model-based control of a batch reactive
distillation column, Journal of Process Control, 10,
209-218.
Beale M.H., Hagan M.T. and Demuth H.B. (2010),

Neural Network Toolbox 7, The MathWorks Inc.,
Apple Hill Drive, Natick MA.
Himmelblau D.M. (2000), Applications of Artificial

Neural Networks in Chemical Engineering, Korean
Journal of Chemical Engineering, 17, 373-392.
References Contd.
Jana A.K., Radha P.V. and Adari K. (2009),
Nonlinear State Estimation and Control of a Batch
Reactive Distillation, Chemical Engineering Journal,
150, 516-526.
Khaledi R. and Young B.R. (2005), Modeling and
Model Predictive Control of Composition and
Conversion in an ETBE Reactive Distillation
Column, Industrial & Engineering Chemistry
Research, 44, 3134-3145.
MacMurray J.C. and Himmelblau D.M. (1995),
Modeling and Control of a Packed Distillation
Column Using Artificial Neural Networks,
Computers and Chemical Engineering Vol. 19, No.
10, 1077–1088.
References Contd.
Pérez-Correa S., González P. and Alvarez J. (2008),
On-line Optimizing Control for a Class of Batch
Reactive Distillation Columns, Proceedings of the
17th World Congress The International Federation of
Automatic Control Seoul, Korea.
Taylor R., Krishna R. (2000), Modelling reactive

distillation, Chemical Engineering Science, 55,
5183-5229.
THANK YOU
VERY MUCH
FOR
LISTENING

2 Isc2011 2

Uploaded by

Copyright:

Available Formats

2 Isc2011 2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2 Isc2011 2

Uploaded by

Copyright:

Available Formats

NEURAL NETWORK MODELING OF A

REACTIVE PACKED DISTILLATION

Abdulwahab GIWA, Süleyman KARACAN

Figure 1. Some Chemical Engineering Processing Units

Figure 2. Reactive Packed Distillation Column

Figure 3. Reactive Packed Distillation Column

 Reduction of energy consumption;

 Reduction of equipment units;

 Increment in the yield of a reversible reaction;

 Reduction of the costs of production.

 Slow in the absence of a catalyst.

Table 1. Properties of the Components

Property Acetic Acid Ethanol Ethyl Acetate Water

Molecular Formula CH3COOH C2H5OH CH3COOC2H5 H2O

Molecular Weight (kg/kgmol) 60.05 46.07 88.11 18.02

Boiling Point (oC) 117.95 78.25 77.15 100.00

Density (kg/m3) 1051.50 795.98 904.86 997.99

 Flavors and essences;

Figure 4. Experimental Set-Up for Data Acquisition

conventional computers or human

beings can be solved easily by neural

NNNAR: T t   f T t  1, , T t  dT  (4)

NNNARX: T t   f Rt  1, , Rt  d R , T t  1, , T t  dT  (5)

where f is the neural network function, T is the top section

ratio, dT is the delay in the top section temperature and t

Hidden Layer Output Layer (b)

R(t) Hidden Layer (c)

Figure 6. Experimental input-output data (Top – Estimation, Middle –

Models Fit Value Mean Squared Error Correlation Coefficient

NNIO -0.07389 0.28232608 0.000000

NNNAR 96.0627 4.3703E-04 0.999458

NNNARX 96.4358 3.5812E-04 0.999468

those predicted using the NNNAR and NNNARX models

have revealed that they could be used to represent the reactive

packed distillation column successfully for further studies. In

addition, NNNARX model was discovered to be better among

the two due to its smaller mean squared error.

I would like to express my sincere

appreciation to my supervisor, Prof. Dr.

Süleyman KARACAN, for his guidance and

suggestions towards the successful

completion of this work.

Beale M.H., Hagan M.T. and Demuth H.B. (2010),

Himmelblau D.M. (2000), Applications of Artificial

Taylor R., Krishna R. (2000), Modelling reactive

You might also like