Академический Документы
Профессиональный Документы
Культура Документы
3, 2015
287
288
A. Nova-Rincn et al.
Jorge Mario Gmez is Associate Professor and former head of the Chemical
Engineering Department at Universidad de los Andes in Bogot, Colombia.
He received a BS and MS in Chemical Engineering from Universidad Nacional
de Colombia, an MBA from Universidad de los Andes, and a PhD from
Universit de Pau et des Pays de lAdour, France. Currently, he lectures on
optimisation of chemical processes. His research interests include dynamic
optimisation with integer variables and energy optimisation of process systems.
This paper is a revised and expanded version of a paper entitled Analysis of
the influence of heat flows in optimal design and operation of multicomponent
distillation columns: extractive distillation case study presented at 2012 AIChE
Annual Meeting, Pittsburgh, PA, USA, 31 October, 2012.
Introduction
(a) Adiabatic distillation column and (b) diabatic distillation column (see online version
for colours)
Year Author
Mixture to be
separated
Water Ethanol
Objective
Minimum
exergy losses
Entropy
minimisation
2012
Ghazi et al.
(2012)
The study is made for the extractive distillation system for the production of fuel grade
ethanol using glycerol as entrainer (due to its energy, availability, being economical, and
advantages against other solvents and distillation techniques), proposed by GarcaHerreros et al. (2011).
290
A. Nova-Rincn et al.
The extractive distillation system is made up of two distillation columns: the first one
for the extractive distillation process and the second for entrainer regeneration. The
mixture to be separated and an entrainer are fed into the first column to obtain a distillate
with high ethanol purity. The second column regenerates the entrainer that can be reused
in the extractive distillation column (Lei et al., 2003). The diagram of the process is
shown in Figure 2.
The present work is focused on the extractive distillation column that stands for a
multi-component separation system. This column operates at atmospheric pressure
(101.325 kPa, due to the saturation pressure of the azeotropic mixture fed, the not so
practical chemical degradation of the mixture into the column under this condition, and
the assumption that there is adequate heating media viable to carry the separation (Kister,
1992)) and produces, as distillate, ethanol with purity over 99.5 mol %.
The extractive distillation column has 17 equilibrium stages, a total condenser, and a
partial reboiler. The column is fed with 52 kmol/h of glycerol on stage 3 at 305K and
with 100 kmol/h of azeotropic mixture (ethanol: 85% mol) on stage 12 at 351K
(saturation temperature), both streams at atmospheric pressure. The aforementioned
description corresponds to the optimal design for the extractive distillation column for the
production of fuel grade ethanol, computed by Garca-Herreros et al. (2011).
Figure 2
Extractive distillation of fuel grade ethanol, using glycerol as entrainer (see online
version for colours)
The assumptions taken into account on this work to develop a model for the extractive
distillation column are the following:
The process is modelled as a series of counter-current separation stages. At each stage (j),
entering and leaving liquid (Lj1, Lj) and vapour (Vj+1, Vj) flows get in contact in order to
reach thermodynamic equilibrium as shown in Figure 3. The model uses the MESH
equations, which refers to (Taylor et al., 2003):
M: material balances
E: equilibrium relations (to model the assumption that the streams leaving the stage
are in thermodynamic equilibrium)
H: enthalpy balances.
To model heat exchange on each stage, it is assumed that a series of band heaters as
used to heat liquids flowing through pipes (Chen et al., 2003) can be clamped to the
column (one per stage). This assumption is done in order to compute Qin j , Qout j and the
side heat exchanger temperature per stage. The calculation of these variables is necessary
to compute exergy losses per stage, as will be seen on the next section.
Figure 3
292
A. Nova-Rincn et al.
(1)
Vj +1 L j 1 +
= 0,
RR
(2)
L j 1 L j Vj = 0,
j COND
j REB.
(3)
(4)
Vj +1 yi , j +1 L j 1 +
xi , j = 0, i C , j COND
RR
(5)
L j 1 xi , j 1 L j xi , j Vj yi , j = 0, i C , j REB.
(6)
Equilibrium relationships
K ij =
yi , j
xi , j
ij Pijsat
P
K ij P Pijsat = 0, i C , j J .
(7)
Summations
nc
y
i =1
i, j
nc
xi , j = 0, i C , j J .
(8)
i =1
Enthalpy balances
Taking into account the proposed heat exchange by stage, for diabatic operation, enthalpy
balance can be described as follows:
Q in j Q out j + Vj +1 H Vj +1 Vj H Vj + L j 1 H Lj 1 L j H Lj + F jL H LF
= 0, j S
j
(9)
where Q in j and Q out j , are positive variables which represents the quantity of energy
supplied or taken out from the jth stage.
1 L
Vj +1 H Vj +1 L j 1 +
H j Q j = 0, j COND
RR
(10)
(11)
The main inequality constraint of the model refers to the ethanol purity (molar fraction)
on distillate as follows:
(12)
The validity of the presented NLP model (112) on which the MINLP problem is based
was already probed and properly represents the behaviour of the extractive distillation
column (Nova-Rincn et al, 2012).
For conventional distillation with total condenser, specified operating pressure and
feed conditions, there are two degrees of freedom in the MESH model (Hanson et al.,
1962). In order to specify the distillation model, reflux ratio (RR) and reboiler heat duty
(Q nt ) of the column are considered as operating variables in this research. For diabatic
distillation, it is necessary to add a couple of additional variables Q in j and Q out j which
denote heat exchanged (entering or leaving) at each stage. It represents two additional
degrees of freedom per stage (S), giving a total of 36 degrees of freedom for the NLP
model.
max P1 = D CD Q in j CB + Q out j CC
Q in j
j J
j J
(13)
Qout j
RR
with:
The proposed optimisation problem was carried out for both adiabatic and diabatic
distillation columns. In the case of adiabatic distillation column, the terms j Q in j and
j Q out j represent only reboiler and condenser heat duties respectively. Terms of capital
cost were not taken into account.
A. Nova-Rincn et al.
294
1 T0
Ex
loss j = Q j
T
j
+ Ex
streams,in j Exstreams,out j , j J
(14)
where Q j represents the heat load at stage j (Q in j or Q out j ), Tj is the side heat exchanger
temperature (temperature of condenser and reboiler utilities in the case of COND and
REB), Ex
stream, in j the exergy of the streams entering to stage j ( ExV j +1 , ExL j +1 , ExF j ),
Exstream,out j is the exergy of the streams leaving to stage j ( ExV j , ExL j ), and Exloss j is the
exergy loss of the stage due to irreversibilities of the stage.
The exergy of a stream is given by its enthalpy and entropy as:
= H T S , j J
Ex
j
j
o j
(15)
(16)
jJ
out j
RR
Scheme and notation for optimal design of the diabatic extractive distillation column
(see online version for colours)
296
A. Nova-Rincn et al.
(17)
(18)
jS
= 1, k K .
k
j
Constraints related to the reflux location take into account the reflux ratio (RR) and
distillate molar flow rate ( D ), as follows:
fjR = RR D z Rj
(19)
(20)
jS
= 1.
R
j
One known condition about feeds locations in the column is the fact that for extractive
distillation, the solvent is introduced into the distillation column above the entry point of
the feed mixture to be separated (Lee, 1990). Therefore for the present case study,
glycerol must be fed above the azeotropic mixture. This can be modelled by imposing a
logical condition as shown below:
z azj z gj 0, j S .
(21)
j j
These inequalities guarantee that if z azj = 1 for one stage j S, then j j z gj = 1 and Fg
enters on or above stage j.
In the same way, reflux must be fed to the column above the feeds. It is imposed by:
z gj z Rj 0, j S .
(22)
j j
It is not necessary to define the constraint (22) for the azeotropic mixture feed because
constraint (21) states that it must be located below the feed of glycerol.
Some changes in the mass and energy balances for the separation stages with respect
to the previous NLP model are required in order to define the MINLP model, as follows:
Mass balances for the equilibrium stages equations (1) and (4) become:
Vj +1 Vj + L j 1 L j + F jk + F jR = 0, j S , k K
(23)
Vj +1 yi , j +1 Vj yi , j + L j 1 xi , j 1 L j xi , j 1 L j xi , j
K
R
+ F jk xiF + F jR xiF = 0, j C , j S , k K
(24)
where xiF = xi , j , j COND, and xiF are the mole fractions of the reflux and feeds
streams respectively.
In the same way, the enthalpy balance equation (9) is defined by:
z
j j
R
j
K
R
+ F jk H Fj + F jR H F = 0, j S , k K
(25)
(26)
Vj +1 yi , j +1 D ( RR + 1) xi , j = 0, i C , j COND
(27)
Vj +1 H Vj +1 D ( RR + 1) H Lj Q j = 0, i COND.
(28)
There are no changes or additional terms that affect the mass and enthalpy balances in the
reboiler, then equations that describe reboiler heat transfer keep the same form as in the
NLP model.
(29)
j 1 = s s+j , j S
j
(30)
Vj sj = 0
(31)
L j s+j = 0
(32)
where j is a corrector for the jth stage, and sj and s+j are slack (positive) variables for
the jth tray. Their values are relative to the existence of the liquid phase on each stage,
according to the following complementary conditions:
If j > 1 then sj > 0 and V j = 0
(33)
298
A. Nova-Rincn et al.
If j < 1 then s+j > 0 and L j = 0.
(34)
For this work, only the absence of liquid flow is modelled because the vapour stream is
bypassed by the non-existing stages; therefore constraint (31) is not taken into account
j
and the term s is removed from constraint (30) giving:
1 j = s+j , j S .
(35)
value of products: market value of the ethanol produced in one year of operation
(Cp(x))
operating cost: cost of the utilities required for the operation of the column in one
year (Co(x))
an electric cost was assumed for the design of the intermediate band heaters
(2.83[$/MJ] (Mara et al., 2009))
same costs of condenser and reboiler reported in Section 3.1.1.1 were used
capital cost: cost of the column, including cost of stages, condenser and reboiler
(CI(x, y))
(36)
T
z Rj Q j 1 0
Tj
1 j j
+ Ex
streams, in j Exstreams, out j
, j J .
(37)
(38)
jJ
wEx, is a weight factor (104) which ensures economic viability of the system as well as
minimisation of the total exergy losses of the distillation column. This parameter was
calculated iteratively (and offline), according to the following procedure:
arbitrary parameter value was chosen and the optimisation problem was solved
Solution strategy
The NLP is based on the rigorous MESH model, composed by a set of algebraic
equations of equality constraints. This set contains equations for computing enthalpies
and entropies, and equalities to determine equilibrium constants which are function of
activity coefficients computed using the non random two liquids (NRTL) model (Renon
and Prausnitz, 1968) leading to a non-convex, NLP problem.
The MINLP problem, which is based on the NLP model, includes decision variables
related to number of stages and locations for feeds streams in the distillation column.
300
A. Nova-Rincn et al.
It was modelled using linear relationships between some of the operational variables (i.e.,
reflux ratio, distillate rate, and feeds molar flows) and a set of binary variables-related
streams locations.
The NLP and the MINLP problems were modelled in GAMS 23.8 on a quad core
Intel i5 2.7 GHz CPU with 8 GB of RAM. To solve the NLP problem, we used IPOPT
(interior point optimiser) (Wchter and Biegler, 2005). On the other hand, SBB (simple
branch and bound) was used as MINLP solver and IPOPT as NLP root-solver and subsolver for the MINLP problem.
IPOPT uses a primal-dual interior-point algorithm with a filter line-search method
(Wchter and Biegler, 2005), and is part of the open source COIN-OR (computational
infrastructure for operations research) project (Biegler, 2010). IPOPT reports advantages
for the solution of large-scale NLP problems and has shown to converge faster than other
NLP algorithms (Biegler, 2010; Wchter and Biegler, 2005).
The SBB had been developed by ARKI Consulting and Development A/S. It is
available as a commercial solver within GAMS and implements a branch-and-bound
algorithm using nonlinear relaxations for the bounding step. The NLP relaxations are
solved by one (or several) of the NLP solvers available with GAMS (Bussieck et al.,
2010) (i.e., IPOPT). SBB may perform better than other MINLP solvers (i.e., DICOPT)
on models that have fewer discrete variables but more difficult nonlinearities, and
possibly on models that are fairly non-convex (GAMS Development Corporation, 2001)
as the proposed MINLP model.
A combined strategy for search node was used for the solution of the MINLP
problem, which is made up by depth first search (DFS) and best bound or best first (BB)
strategies, and is included into the SBB options (GAMS Development Corporation,
2013).
Diabatic column
Initial values
Initial
values
RR
0.103
0.885
0.012
0.191
0.237
D [kmol/h]
66.435
85.427
69.569
85.427
85.427
Variable
[E]
[P1] Optimisation Optimisation
2.972
3.902
1.2577
4.05
2.605
Profit [$/h]
1982.81
2543.776
1952.54
2548.925
2548.89
Q nt [GJ/h]
4.379
8.048
3.357
5.309
5.573
Q1 [GJ/h]
2.85
6.269
2.809
3.960
4.113
Table 3
Stages heat loads, for the diabatic distillation column. NLP problem
Diabatic column
Initial values
Heat load
Q out j [GJ/h]
Q in j [GJ/h]
[P1] Optimisation
Stage
Stage
Value
Stage
Value
0.001
0.001
0.006
0.006
0.002
0.002
0.003
0.003
0.005
0.005
0.005
0.005
0.005
0.005
12
Value
[E] Optimisation
0.604
0.004
0.004
10
0.002
10
0.002
11
0.003
11
0.003
12
0.230
12
0.230
13
0.025
13
0.025
14
0.020
15
0.018
16
0.018
17
0.019
18
0.019
4.271E-04
0.172
0.002
14
0.020
15
0.018
16
0.018
17
0.019
18
0.019
302
A. Nova-Rincn et al.
Besides, when total energy consumption (Q Total = (Q nt + Q in j ) (Q1 + Q out j )) is calculated,
a value of 1.779 GJ/h is attained for all of them. It means that the system complies with
the first law of thermodynamics. The distillation systems use the same amount of energy,
but it is being used in different ways, i.e. different locations along the column.
The main objective of the thermodynamic analysis is to evaluate whether adiabatic or
diabatic distillation system has lower total exergy losses, which represent a more efficient
use of energy in the system. Additionally the analysis estimates how much energy is
wasted with the optimal configuration achieved for the economic optimisation in contrast
with the system with the minimum exergy loss. For the mentioned analysis, exergy loss
profiles for adiabatic and diabatic distillation columns for the studied optimisations are
depicted in Figure 5.
Figure 5
Exergy losses profiles for adiabatic and diabatic distillation column, thermodynamic
and economic objectives (see online version for colours)
By inspection of the profiles presented in Figure 5, it can be noticed that the major exergy
losses for both distillation systems and optimisations are located at reboiler (stage 19),
condenser (stage 1), and at the surroundings of feed stages of the extractive distillation
column (stages 3 and 12 to 15).
Until this point and based on results on Table 2, a reduction of 33.24% in the exergy
losses of the system has been attached via diabatic operation of the extractive distillation
column.
Presented results highlight the importance of measuring the energy efficiency of the
separation systems since it is possible to get the same net profit with a lower energy
waste, when an additional exergy-based analysis is made.
Taking this into account, exergy analysis was included for the design of the
distillation column to find out if the results of the NLP problem can be extrapolated to the
MINLP problem.
52G
Design variable
35G
Initial
[P2]
[EE]
Initial
[P2]
[EE]
values Optimisation Optimisation values Optimisation Optimisation
Number of stages
27
18
20
25
29
26
20
12
10
18
19
17
Comparing profit and exergy losses values for the proposed optimisations in Table 5, the
systems with the best profit also presents the greatest exergy losses. Besides, the system
with minimum exergy losses (35G for economic-exergetic objective) presents a reduction
of only 0.45% on its profit with respect to the most profitable (52G for economic
objective), but the reduction in exergy losses is 45.09%. It means that in order to get a
small extra-profit, great quantities of energy are being wasted (1.526GJ/h). Furthermore,
the selection of an appropriated entrainer molar flow has a significant influence in the
exergy losses of the system. Similarity in the profit value for both economic and
economicexergetic optimisations confirms the validity of the proposed exergeticeconomic objective function and the proper selection of the wEx parameter value.
From Tables 4 and 5 results, it can be observed that for a 52 kmol/h feed flow of
glycerol, achieved optimal designs present shorter columns with a very small reflux ratio,
when compared with the results for 35 kmol/h feed flow of glycerol for the proposed
304
A. Nova-Rincn et al.
52G
Initial
values
Variable
RR
D [kmol/h]
Exergy loss [GJ/h]
Profit [$/year]
35G
[P2]
[EE]
Optimisation Optimisation
Initial
values
[P2]
[EE]
Optimisation Optimisation
0.150
0.007
0.005
0.263
0.211
0.223
85.427
85.411
85.426
85.414
85.425
85.423
3.538
3.384
2.581
2.902
2.874
1.858
14,932,445 14,970,476
Q nt [GJ/h]
5.661
5.314
5.321
5.318
5.231
5.271
Q1 [GJ/h]
3.824
3.477
3.471
4.199
4.112
4.150
Figure 6
Figure 8
For heat loads at stages, results for the NLP problem reported that additional heat flows
are required in the surrounding of both feed streams for the economic optimisation of the
diabatic extractive distillation column; nevertheless, when the optimal distribution of heat
loads is computed by a simultaneous strategy for operation and design, the additional heat
flows are only necessary at the neighbouring of the azeotropic mixture feed stage, as
shown in Table 6.
Figure 9 compares exergy loss profiles for economic-exergetic objective function
for the studied systems in the NLP problem and the profile achieved for a minimum
exergy loss system achieved for the NLP problem (base case).
306
A. Nova-Rincn et al.
Table 6
52G
Initial
values
Heat load
[P]
[EE]
Optimisation Optimisation
Initial
values
[P]
[EE]
Optimisation Optimisation
Stage Value Stage Value Stage Value Stage Value Stage Value Stage Value
Q in j [GJ/h]
18
0.006
19
0.061
Q out j [GJ/h]
20
0.125
Figure 9
35G
11
0.061
0.052
17
0.026
18
0.031
16
0.028
12
0.127
10
0.124
18
0.083
17
0.086
17
0.084
Comparison of systems with minimum exergy loss (see online version for colours)
From exergy loss profiles in Figure 9, it can be noticed that regardless of the number of
stages and location of feeds, the major exergy losses are located at the glycerol feed stage
and the reboiler for the analysed cases. Detailed values are shown in Table 7.
The system which reported major reduction in exergy losses is the distillation column
designed for a 35 kmol/h flow of glycerol. The system 35G leads to reductions of 34.1%
and 26.82% in exergy losses with respect to the base case for the glycerol feed stage and
the reboiler respectively.
Table 8 summarises the reduction of the exergy losses of the extractive distillation
column for the production of fuel grade ethanol from the adiabatic distillation system
achieved by Garca-Herreros et al. (2011) (adiabatic case), then by the inclusion of
diabatic operation made in the NLP problem (base case), and ending with the optimal
design, taking into account the analysed molar feed flows of entrainer of the MINLP
problem.
Base case
52G
35G
0.793
0.852
0.522
Reboiler
1.078
1.041
0.789
Table 8
Base case
52G
35G
2.604
2.589
1.858
Conclusions
This work presents a new complete formulation and solution for the optimal design and
operation for a diabatic distillation system taking into account a multi-component mixture
(Glycerol, and ethanol-water azeotropic mixture), using an MINLP model. The proposed
model includes binary variables related to reflux and feeds locations, and an MPCC
formulation to model the vapour-liquid equilibrium in the separation stages.
For the proposed objectives and arrangements of feed flows, it was found that heat
loads along the length of the column are only necessary in the surroundings of the
azeotropic mixture feed stage. Aforementioned results differ significantly from the
computed for optimal operation achieved using a consecutive strategy.
The designs achieved for proposed economic-exergetic objective function shows that
the assumption of heat exchange into the equilibrium stages of the column can lead to a
more sustainable process ensuring also the high profit of the system.
Although the implementation of a diabatic distillation system with minimum exergy
losses as proposed in the MINLP problem has several technical limitations (band heaters
along the column), the presented study gives a first approach to the design of more
sustainable distillation columns.
In terms of process intensification, although the results for the studied system shows
an increase in the size of the column (19 to 26 separation stages) it leads to a major
reduction in exergy losses which means a more sustainable separation process.
308
A. Nova-Rincn et al.
Perspectives
The studied economic-exergetic criterion can also be applied to the design of other
energy-transforming systems associated to low thermodynamic efficiencies, taking into
account alternative strategies related to the energy usage of the system (like diabatic
distillation in the case of distillation columns).
References
Arajo, A.C.B., Braskem, Macei, B. and Vasconcelos, L.G.S. (2007) Exergetic and economic
analysis of an industrial distillation column, Braz. J. Chem. Eng., Vol. 24, pp.461469.
Barttfeld, M., Aguirre, P.A. and Grossmann, I.E. (2003) Alternative representations and
formulations for the economic optimisation of multicomponent distillation columns,
Computers & Chemical Engineering, Vol. 27, pp.363383.
Biegler, L.T. (2010) Nonlinear Programming: Concepts, Algorithms, and Applications to Chemical
Processes, Society for Industrial and Applied Mathematics: Mathematical Optimization
Society, Philadelphia.
Bussieck, M.R., Vigerske, S., Cochran, J.J., Cox, L.A., Keskinocak, P., Kharoufeh, J.P.
and Smith, J.C. (2010) MINLP solver software, Wiley Encyclopedia of Operations Research
and Management Science, John Wiley & Sons, Inc., Washington, DC.
Chen, X.D., Ai, G.M. and Duffy, G.G. (2003) Rules-of-thumb of implementing short electric band
heaters (length to diameter ratio <1.5) for external heating of pipe flows, Erfahrungsregeln
fr die Ausfhrung von kurzen elektrischen Bandheizungen (Lnge-Durchmesser-Verhltnis
<1,5) zur ueren Beheizung von Rohrleitungen 133.
De Koeijer, G., Rsjorde, A. and Kjelstrup, S. (2004) Distribution of heat exchange in optimum
diabatic distillation columns, Energy, Vol. 29, pp.24252440.
Demirel, Y. (2004) Thermodynamic analysis of separation systems, Separation Science and
Technology, Vol. 39, pp.38973942.
Demirel, Y. (2006) Assessment of thermodynamic performances for distillation columns,
International Journal of Exergy, Vol. 3, p.345, doi:10.1504/IJEX.2006.010229.
Dincer, I. and Rosen, M. (2007) Exergy: Energy, Environment, and Sustainable Development,
Elsevier, Amsterdam, Boston.
Doherty, M.F. and Malone, M.F. (2001) Conceptual Design of Distillation Systems, Mc-Graw Hill,
New York.
GAMS Development Corporation (2001) GAMS/SSB Solver/#9 [WWW Document], SBB Solver,
http://www.gams.com/presentations/poster_sbb.pdf
GAMS Development Corporation (2013) SBB [WWW Document], SBB Solver Manual,
http://www.gams.com/dd/docs/solvers/sbb.pdf
Garca-Herreros, P., Gmez, J.M., Gil, I.D. and Rodrguez, G. (2011) Optimisation of the design
and operation of an extractive distillation system for the production of fuel grade ethanol using
glycerol as entrainer, Industrial & Engineering Chemistry Research, Vol. 50, pp.39773985.
Ghazi, M., Sotoodeh, A.F. and Amidpour, M. (2012) Modelling and optimisation of a waterethanol distillation column based on exergoeconomic analysis, International Journal of
Exergy, Vol. 11, p.173, doi:10.1504/IJEX.2012.049742.
Gopal, V. and Biegler, L.T. (1999) Smoothing methods for complementarity problems in process
engineering, AIChE Journal, Vol. 45, pp.15351547, doi:10.1002/aic.690450715
Grossmann, I.E., Aguirre, P.A. and Barttfeld, M. (2005) Optimal synthesis of complex distillation
columns using rigorous models, Computers & Chemical Engineering, Vol. 29,
pp.12031215.
310
A. Nova-Rincn et al.
Soares Pinto, F., Zemp, R., Jobson, M. and Smith, R. (2011) Thermodynamic optimisation of
distillation columns, Chemical Engineering Science, Vol. 66, pp.29202934.
Spasojevi, M.D., Jankovi, M.R. and Djakovi, D.D. (2010) A new approach to entropy
production minimisation in diabatic distillation column with trays, Thermal Science, Vol. 14,
pp.317328.
Stougie, L. and van der Kooi, H.J. (2011) The relation between exergy and sustainability
according to literature, Presented at the 2nd International Exergy, Life Cycle Assessment, and
Sustainability Workshop & Symposium, Nisyros, Greece, pp.590597.
Taylor, R., Krishna, R. and Kooijman, H. (2003) Real World Modeling of Distillation, Chemical
Engineering Progress, New York.
Tsatsaronis, G. (1993) Thermoeconomic analysis and optimisation of energy systems, Progress in
Energy and Combustion Science, Vol. 19, pp.227257.
Turton, R., Bailie, R.C., Whiting, W.B. and Shaeiwitz, J.A. (2002) Analysis, synthesis, and design
of chemical processes, Prentice Hall International Series in the Physical and Chemical
Engineering Sciences, p.915.
Viswanathan, J. and Grossmann, I.E. (1993) Optimal feed locations and number of trays for
distillation columns with multiple feeds, Ind. Eng. Chem. Res., Vol. 32, pp.29422949,
doi:10.1021/ie00023a069.
Wchter, A. and Biegler, L.T. (2005) On the implementation of an interior-point filter line-search
algorithm for large-scale nonlinear programming, Mathematical Programming, Vol. 106,
pp.2557.
Nomenclature
Subscripts
B
Boiler
Distillate
Component index
Superscripts
az
Azeotropic mixture.
Glycerol
Liquid
Reflux
Vapour
Latin symbols
AF
CB
CD
Cp x
CI
Co x
D
Operating cost, $
Annualising factor
Capital cost, $
Distillate, kmol/h
Ex
streams, in j
Ex
streams, out j
F k
Feed of species k
fjk
F jL
F R
Reflux, kmol/h
f
R
j
H Lj
H LF
j
H Vj
Kij
L j
nc
nt
Q j
Q in j
Q out j
RR
sat
ij
j
+
sj
Tj
To
Vj
wEx
xi,j
yi,j
R
j
z kj
312
A. Nova-Rincn et al.
Greek symbols
ij