Flowsheet synthesis and complex distillation column design for separating ternary heterogeneous...

Journal Identification = CHERD Article Identification = 720 Date: June 6, 2011 Time: 11:26 am

chemical engineering research and design 8 9 ( 2 0 1 1 ) 1362–1376

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

j ourna l ho me pa ge: www.elsev ier .com/ locate /cherd

Flowsheet synthesis and complex distillation columndesign for separating ternary heterogeneousazeotropic mixtures

Paritta Prayoonyonga, Megan Jobsonb,∗

a Chemical Engineering Department, Faculty of Engineering, Mahidol University, 25/25 Phuttamonthon 4 Rd.,Nakornprathom 73170, Thailandb Centre for Process Integration, School of Chemical Engineering and Analytical Science, The Mill, Sackville Street,The University of Manchester, Manchester M13 9PL, UK

a b s t r a c t

The boundary value method for design of distillation columns separating ternary heterogeneous azeotropic mixtures

is extended to include complex configurations, i.e. columns with integrated decanters and with multiple heteroge-

neous stages; double-feed columns; columns with intermediate decanters. The methods can be used for establishing

product feasibility in a column and evaluating a column design in terms of cost. Multiple feasible designs can be

generated for a given set of product specifications, according to the reflux ratio, number of heterogeneous stages

and liquid phase ratio; these designs may be evaluated with respect to operating and equipment costs. Case studies

illustrate the design method.

An algorithmic approach is presented for synthesising novel sequences for separating, by distillation and decant-

ing, ternary heterogeneous azeotropic mixtures. Existing synthesis procedures that consider only simple single-feed

columns with decanters and integrated decanters are extended to include more industrially relevant options such

as columns with several heterogeneous stages, double-feed columns, columns with intermediate decanters, and

those accepting a heterogeneous liquid feed. With these advanced column configurations included in the syn-

thesis method, a wider range of sequences may be considered systematically, allowing sequences that are more

economically attractive than conventional designs to be identified. A case study illustrates the approach.

© 2011 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: Heterogeneous azeotropic distillation; Complex columns; Boundary value method; Sequence synthesis;

Recycle heuristics

from previous work (Pham et al., 1989; Wasylkiewicz et al.,

1. Introduction

Separation of heterogeneous azeotropic mixtures using dis-tillation columns is an important industrial process forwhich systematic design approaches are still lacking. Columnand flowsheet design using conventional process simulationsoftware is challenging, because of difficulties convergingsimulations involving these non-ideal mixtures, and becauseconventional simulation software does not lend itself tosystematic and comprehensive exploration of design alterna-tives. The boundary value method for assessing feasibility of
proposed separations and for designing columns separating
∗ Corresponding author.E-mail addresses: [email protected] (P. Prayoonyong), megaReceived 2 November 2010; Received in revised form 4 February 2011;

0263-8762/$ – see front matter © 2011 The Institution of Chemical Engidoi:10.1016/j.cherd.2011.02.015

ternary azeotropic mixtures can overcome these difficulties.While this method is well established, there remain significantgaps in its capabilities in the area of heterogeneous azeotropicdistillation.

This paper presents a methodology to generate, simulateand evaluate complex column configurations and a frameworkfor flowsheet synthesis to separate ternary heterogeneousazeotropic mixtures. The column design method exploitsthe boundary value method for establishing separation fea-sibility and for assessing economic performance, following

[email protected] (M. Jobson). Accepted 8 February 2011

2000; Urdaneta et al., 2002). Single- and double-feed columns

neers. Published by Elsevier B.V. All rights reserved.

http://www.sciencedirect.com/science/journal/02638762

mailto:[email protected]

mailto:[email protected]

dx.doi.org/10.1016/j.cherd.2011.02.015


chemical engineering research and design 8 9 ( 2 0 1 1 ) 1362–1376 1363

wdlfcmts(n

rafTao

abpsracoVhtssddatcass

1a

Ead1ria(pu2o2s

(pppdf

ith integrated decanters and columns with intermediateecanters are addressed. The design method provides pre-

iminary column design information, e.g. number of stages,eed locations and heat duties, which allows estimation ofapital and operating costs. Thus many alternative designsay be readily generated and evaluated, which is valuable in

he synthesis of separation flowsheets. The approach for flow-heet synthesis is heuristically driven and partly evolutionaryNishida et al., 1981), rather than applying optimisation tech-iques to determine the flowsheet structure.

The synthesis problem, for a given feed and products withequired purities and recoveries, involves the selection of sep-rations, the unit operations (and their operating conditions)or achieving these and the flow connections between them.he choices made should allow the resulting flowsheet tochieve the separation aims with a good performance in termsf energy demand or processing costs.

Separation sequence synthesis for heterogeneouszeotropic mixtures is challenging – it may not be possi-le to achieve desired purities or recoveries because of theresence of azeotropes; the number of alternative flowsheettructures to choose from may be large; recycles may beequired to improve product recoveries or purities. Severallternative flowsheet structures may exist – both distillationolumns and decanters may be used; the columns may carryut direct, indirect or intermediate (e.g. ‘saddle-type’) splits.arious column configurations may be used – columns mayave one or two feeds, and decanters may be integrated withhe column, at the top of the column or on an intermediatetage. Verifying that a proposed separation is feasible is also aignificant challenge; identifying the most economic columnesigns poses yet another challenge, exacerbated by theifficulty of converging column simulations. When recyclesre present, their source and destinations need to be selected;o solve the material balance over each column, the recycleompositions and flow rates need to be determined in an iter-tive process. These challenges motivate the development ofystematic approach for synthesising, testing and evaluatingeparation sequences.

.1. State of the art in design of ternary heterogeneouszeotropic distillation columns

stablished methods for assessing separation feasibilitynd/or column design to ternary heterogeneous azeotropicistillation – namely the boundary value method (Pham et al.,989; Wasylkiewicz et al., 2000) and the continuous distillationegion-based method (Urdaneta et al., 2002) – have certain lim-tations to their applicability. Existing boundary value designpproaches restrict the two liquid phases to the top decanterPham et al., 1989), allow the appearance of the two liquidhases in the column but do not explain how to model the col-mn with multiple heterogeneous stages (Wasylkiewicz et al.,000), do not include the reflux phase split ratio as a degreef freedom for design (Pham et al., 1989; Wasylkiewicz et al.,000), and need the number of heterogeneous stages to bepecified up front (Urdaneta et al., 2002).

The continuous distillation region-based methodUrdaneta et al., 2002) does not restrict the two liquidhases to appear only in the decanter. However, it is com-utationally intensive as it requires the calculation of pinchoints and composition profiles to compute the continuous
istillation regions; also it relies on visual inspection to checkor intersection of continuous distillation regions. It can pro-
vide details of a feasible column design, e.g. number of stages,feed location. Recent developments of pinch-based methodsaim to overcome some of its limitations (e.g. Kraemer et al.,submitted for publication). This work extends the boundaryvalue method to address shortcomings of previous work;furthermore, in order to be relevant and applicable to flow-sheet synthesis, a design method that is not computationallyintensive is sought.

The boundary value method may be used to assess feasibil-ity of a proposed separation and to generate column designs,given fully specified product compositions. Liquid compo-sition profiles for the rectifying and stripping sections aregenerated in mole fraction space by stage-by-stage materialbalance and phase equilibrium calculations (and often, asin this work, with enthalpy balances as well) starting withthe product compositions. An intersection of the composi-tion profiles indicates that a continuous profile connecting theproduct compositions may be obtained, and thus that the pro-posed separation is feasible. Information about the design – i.e.number of stages in each section and reflux ratio – of the cor-responding column may be retrieved and used to evaluate thecolumn design (Pham et al., 1989). The capital and/or operat-ing costs of the column can thus be estimated, allowing designalternatives to be compared. The most attractive designs needto be evaluated further, based on other important issues, suchas operability and controllability, and the overall separationflowsheet also needs to be generated and evaluated.

The boundary value method has several well-known lim-itations, including the sensitivity of the composition profilesto the product composition and the need for approximationswhen extending the method to multicomponent systems, e.g.Amminudin et al. (2001).

By repeated application of the boundary value method,varying the reflux and/or reboil ratio and/or feed condition of aproposed column, different designs may be readily generatedand evaluated with respect to their energy demand or eco-nomic performance (Pham et al., 1989; Van Dongen et al., 1985;Levy et al., 1985; Castillo et al., 1998). This capability makesthe boundary value method a powerful tool for sequence syn-thesis and preliminary column design. The boundary valuemethod is therefore adopted in this work, and extended toaddress some significant gaps in the design of ternary het-erogeneous azeotropic distillation columns, in particular thecomplex column configurations depicted in Fig. 1.

1.2. Existing approaches to flowsheet synthesis

Published approaches for the synthesis of separationsequences for heterogeneous azeotropic mixtures are under-pinned by heuristics, using inspection of residue curve maps(Pham et al., 1990; Wasylkiewicz et al., 2003) or algorithmicapproaches (Tao et al., 2003; Vanage, 2005). A sub-problemof sequence synthesis is identifying which streams shouldbe mixed or recycled to recover high-purity products – exist-ing approaches use heuristics (Thong et al., 2001; Tao et al.,2003; Vanage, 2005). The existing algorithms and heuristics arerestricted in the range of column and flowsheet configurationsthey address. Mathematical optimisation approaches have, todate, not been applied for synthesis flowsheets addressing thisclass of separation problem.

The existing sequence synthesis algorithms (Tao et al.,2003; Vanage, 2005) first generate separation sequences that
are most probably feasible. Opportunities to use simplecolumns (with a single feed and two products) carrying out


1364 chemical engineering research and design 8 9 ( 2 0 1 1 ) 1362–1376

B

DIII

F

a b

Fʹ

L1stage 2 L1

I = (1 - φ1) L1

L1II = φ1 L1

Feed

III

Distillate

Bottom produ ct

Condenser

Rebo iler

Sidestrea m produ ct

Draw

Return

Fig. 1 – (a) Column with integrated decanter (i.e. stage 1) and one or two feeds; (b) column with intermediate decanter
(Prayoonyong, 2009).
direct and indirect splits, and stand-alone decanters (separat-ing a column product) or integrated decanters (separating thetwo liquid formed on condensing the overhead vapour andreturning part of the condensate as reflux) are addressed. Aclassification system allows those streams that are not desiredproducts to be identified. Options for mixing these streams orrecycling them are then generated. Those sequences identi-fied as containing infeasible separations may be made feasibleby introducing mixing of non-product streams or recycling ofpure components so that the resulting stream can be sepa-rated in a decanter or a column.

Existing algorithms and heuristics (Tao et al., 2003; Vanage,2005) do not, however, take into account double-feed columns,columns with intermediate decanters or columns with het-erogeneous feeds (where the two phases lie in differentdistillation regions). The configurations have been shown tobring benefits for process economics, energy demand or fea-sibility (Prayoonyong, 2009). The existing synthesis approachis extended in this work to include such complex columnconfigurations. Including these novel options in the synthesisalgorithm will allow a wider range of industrially importantflowsheet structures to be identified, designed and evaluated,and thus will facilitate the development of flowsheets withbetter performance in terms of energy demand, cost, yield,etc.

2. Extended design methods for complexcolumn configurations

In these extensions to the boundary value method, the feedcomposition and product composition of one product needto be fully specified, and one purity specification needs tobe given for the remaining product. These specificationsfix the material balance. The operating pressure is speci-fied and assumed constant in the column. Two variablesfrom the reflux ratio, reboil ratio and feed condition fix thecolumn energy balance. As explained in Section 2.1, the com-position profiles are generated for each section, applyingstage-by-stage material balances and calculating the phaseequilibrium. Energy balances are also incorporated, as theassumption of constant molar overflow is found to be inade-quate for highly non-ideal systems. The intersection betweenthe profiles is identified using a line intersection searchalgorithm (Hölz, 2006). As with the classical McCabe–Thieleboundary value method for column design, a fractional num-ber of stages may result. In order to verify the design approach
and its results, column designs generated by the boundaryvalue method are simulated in Aspen HYSYS 2004.1; the per-
formance of the columns, in terms of composition profiles,product purities and reboiler and condenser duties, etc., iscompared to that predicted by the boundary value method.The existence of input multiplicities – where different designinputs and specifications give identical outputs (i.e. productcompositions and flow rates) – in these processes is wellknown, e.g. Gani and Jørgensen (1994). The multiplicities pre-dicted by the boundary value method may be verified byconventional simulation (Prayoonyong, 2009).

Designs generated by the boundary value method maybe evaluated in terms of their economic performance usingappropriate cost models (Prayoonyong, 2009). The capital costmodels used take into account column height, diameter, shellthickness; heat transfer area of heat exchangers; vessel sizeof decanters, etc. (Peters et al., 2003) and operating cost mod-els take into account the quality and quantity of heating andcooling required (Smith, 2005). Several designs may be gener-ated by varying key degrees of freedom – such as reflux ratio,feed condition and phase split ratio – and evaluated using thecost models.

In this work the models have been implemented in Matlab7.0 code; physical, thermal and phase equilibrium propertiesof fluid streams are calculated in HYSYS and retrieved by Mat-lab (Prayoonyong, 2009). Prior to applying the boundary valuemethod, the vapour–liquid–liquid-equilibrium behaviour iscomputed (at saturation conditions), and represented asa liquid–liquid region and a ‘vapour line’, i.e. the set ofcorresponding equilibrium vapour compositions. These pre-calculations reduce the need for repeated evaluation of phaseequilibrium relationships during generation of compositionprofiles.

2.1. Single-feed column with an integrated overheaddecanter

In an integrated decanter the condensed overheads of thecolumn – a two-phase liquid – are separated by decanting,allowing the reflux to have a different composition to the dis-tillate. The design method of Pham et al. (1990) restricts theexistence of two-phase liquids to the decanter. However, inindustrial columns two liquid phases frequently occur withinthe column as well (Pucci et al., 1986; Kovach and Seider, 1987;Repke and Wozny, 2002) and such columns may be economi-cally advantageous (Urdaneta et al., 2002). The boundary valuemethod is extended to accommodate a two-phase reflux beingreturned to the column from the overhead decanter, as illus-trated in Fig. 1(a).

To generate the rectifying composition profile, the refluxphase split ratio, �1, (the ratio of the molar flow of the heavy



VRRR hV ,,,222

y

VRn

Rn

Rn

h

V

,

,

,

1

1

1

+

+

+

y

LRn

Rn

Rn

h

L

,

,

,

x

VRn

Rn

Rn

h

V

,

,

,

1

1

1

+

+

+

y

IILRn

IIRn

Rnn

h

L

,,

, ,

,

x

φ

( )

ILRn

IRn

Rnn

h

L

,,

, ,

,

x

φ−1

VRRR hV ,,,222

y

2

n

… III D, xD, hD

QC

2

n

…

II D, xD, hDI

Qa b

C

neou

lbfircrtdlrtcb

lhoioTbtnt

sedtohtnr(s

i

(

(

sonably represented by the UNIQUAC-Virial model in HYSYS2004.1 using the parameters given in Table 1. Experimental

2

m

B, xB, hB

Reboiler

…

QR

VSSS hV ,,,111

y

LSSS hL ,,,222

x

VSm

Sm

Sm

h

V

,

,

,

y

LSm

Sm

Sm

h

L

,

,

,

1

1

1

+

+

+

x

Fig. 2 – Schematic of rectifying sections with (a) homoge

iquid to that of the total liquid flow, Prayoonyong, 2009) muste specified. Unlike conventional columns, the rectifying pro-le of the column with integrated decanter starts from theeflux composition and depends not only on the top productomposition but also on the reflux phase split ratio. For a giveneflux ratio, there is a reflux phase split ratio (in the range 0–1)hat causes two liquid phases to appear within the column,enoted by �o

1. For other values of the phase split ratio, twoiquid phases appear only in the decanter at the given refluxatio. The number of heterogeneous stages in the column andhe phase split ratio on the last heterogeneous stage of theolumn are however not unique, allowing multiple profiles toe generated.

When generating heterogeneous rectifying profiles, pre-iminary tests establish whether the final heterogeneous stageas been reached and hence determine the maximum numberf heterogeneous stages that may exist. Once the rectify-

ng profile leaves the two-phase region, the computationf the profile is carried out in the conventional manner.he corresponding stripping composition profile must alsoe generated. Heterogeneous stages may be encountered inhe stripping section; in this case the phase split ratio doesot need to be specified as it is implicit in the calcula-ions.

The liquid compositions on each stage are calculated byolving simultaneously material and energy balances, fromach end of the column, along with phase equilibrium. Theecanter is taken to be stage 1 of the rectifying section, andhe reboiler as stage 1 in the stripping section. As the liquidn a stage in the rectifying section may be heterogeneous oromogeneous; the material and energy balance equations forhe two cases are different (Eqs. (1) and (2) refer to homoge-eous stages, as represented in Fig. 2(a) and Eqs. (3) and (4)efer to heterogeneous stages, as presented in Fig. 2(b). Eqs.5) and (6) provide the governing equations for the strippingection represented in Fig. 3).

The component material balances for homogeneous stagesn the rectifying section are:

LRn + D)yR

n+1 − LRnxR

n − DxD = 0 (1)

The corresponding energy balance is:

LRn + D)hR,V

n+1 − LRnhR,L

n − DhD + QC = 0 (2)

s liquid on stage n, (b) heterogeneous liquid on stage n.

Where two liquid phases are present, the component bal-ances are:

(LRn + D)yR

n+1 − LRn ((1 − �n)xR,I

n + �nxR,IIn ) − DxD = 0 (3)

and the energy balances account for the enthalpy of both liq-uid phases:

(LRn + D)hR,V

n+1 − LRn ((1 − �n)hR,L,I

n + �nhR,L,IIn ) − DhD + QC = 0 (4)

where

QC = ((r + 1)hR,V2 − r((1 − �1)hR,L,I

1 + �1hR,L,II1 ) − hD)D

In the stripping section, the material and energy balancesare:

(VSm + B)xS

m+1 − VSmym − BxB = 0 (5)

(VSm + B)hS,L

m+1 − VSmhS,V

m − BhB + QR = 0 (6)

where QR = (shS,V1 − (s + 1)hS,L

2 + hB)B if s = VS1/B is specified, or

QR = DhD + BhB − FhF + QC if the feed condition is specified.

2.1.1. Example – single-feed column with integrateddecanterFig. 4 presents boundary value method results for a columnseparating the mixture acetic acid–n-butyl acetate–water. Thevapour–liquid–liquid equilibrium of this mixture may be rea-

(stage m=1 )

Fig. 3 – Schematic of the stripping section of a column.



n-B utyl acetate

Acetic acid

B

D

F

Stripping profile

Rectifying profile

Water Water n-Butyl acetate

B

D

F

φ10 = 0

φ10 = 0.2

φ10 = 0.4

φ10 = 0.6

φ10 = 0.8

φ10 = 1

Stripping profile

Rectifying profile

Acetic acid

(ii)(i)

Fig. 4 – Composition profiles of a column recovering acetic acid: Case (i) two-phase liquid occurs only in the decanter(�1 = 0.3); Case (ii) heterogeneous liquid within column (�◦

1 = 0.589). Profiles for six different phase split ratios on the finalstage are shown (Prayoonyong, 2009).

Table 1 – Water/acetic acid/n-butyl acetate: UNIQUACinteraction parameters (aij) from HYSYS 2004.1 (entriesin row and column represent component i and j,respectively). The binary parameters bij are all zero.

Component Water Acetic acid n-Butylacetate

Water – 427.741 738.169Acetic acid −305.452 – 712.235n-Butyl acetate 398.290 −298.434 –

ble designs and to select between them using an appropriate

data of Gomis et al. (2000) and Ou-yang et al. (1985) validatethe model and parameters (Prayoonyong, 2009). Note that thevapour phase model accounts for dimerisation of acetic acid.Table 2 presents the separation specifications.

For the specified reflux ratio, the separation is not feasi-ble when a two-phase liquid appears only in the decanter and� > �o

1 (in Case (i) the rectifying profile leaves the two-liquidphase region near the n-butyl acetate vertex and does notintersect the stripping profile). However, it is feasible for a col-umn with a heterogeneous liquid leaving the top stage (Case
ii). As can be seen in Fig. 2, in this case, the composition pro-
Table 2 – Specifications for column separating water/acetic acid

Feed, F

Composition (mole fraction)Water 0.4499a

Acetic acid 0.4501a

n-Butyl acetate 0.1000

Feed condition (q) 1a

Flow rate (kmol/h) 100a

Reflux ratio

Phase split ratio of reflux Case (i): �1 /= �o1

Case (ii): �1 = �o1

�o1 = phase split ratio of the reflux that may cause the liquid leaving stage

a Specified values.

file leaving the last heterogeneous stage depends strongly onthe phase split ratio on that stage, in terms of both directionand change in composition from one stage to the next. Feasi-ble designs are also found when � < �o

1 causing the rectifyingprofile leaving the two-liquid phase region near the water ver-tex and crossing the stripping profile (results are not shown).Note that the reboil ratio depends on the phase split ratio ofthe reflux via the column energy balance. The maximum num-ber of heterogeneous stages in the column is 10; four of thesix values of the phase split ratio on the last heterogeneousstage (0.4, 0.6, 0.8 and 1) lead to feasible column designs. Theboundary value method may be applied repeatedly to searchfor feasible column designs corresponding to different num-bers of heterogeneous stages as well as different reflux ratios.In this case, designs with fewer heterogeneous stages at thesame reflux ratio are found to need more theoretical stagesoverall, which would indicate that their capital costs wouldbe higher. Heuristics are not known to be available to pro-vide guidance for selecting economically viable and technicalfeasible values of the design variables (reflux ratio and phasesplit ratio), so this approach aims to generate multiple feasi-

performance indicator.

/n-butyl acetate mixture.

Distillate, D Bottoms, B

0.8177 0.00190.0003a 0.9980a

0.1820 0.0001a

54.91 45.09

1.45

0.30.5890

1, next to the decanter, to be heterogeneous.



FL, xFL, hFL

2

m

B, xB, hB

Reboiler (stage m=1 )

…

QR

VSSS hV ,,,111

y

LSSS hL ,,,222

x

VMm

Mm

Mm

h

V

,

,

,

y

LMm

Mm

Mm

h

L

,

,

,

1

1

1

+

+

+

x

Smx

Sm 1−y

Fig. 5 – Schematic of the stripping section including a lowerfeed.

2d

Diheo2if

ceinad

mtrutotw(tsrc

mpi(l(tt

(

Table 3 – 1-propanol/water/1-butanol: UNIQUACinteraction parameters (aij) estimated by UNIFAC-VLE inHYSYS 2004.1 (entries in rows and columns representcomponents i and j, respectively). The binary parametersbij are all zero.

Component 1-Propanol Water 1-Butanol

1-Propanol – 414.781 9.289Water 26.845 – 134.2181-Butanol −2.927 402.753 –

tive column designs, but would most probably be plagued by

.2. Double-feed column with an integrated overheadecanter

ouble-feed columns are commonly used for separat-ng homogeneous azeotropic mixtures by introducing aigh-boiling component towards the top of the column. Het-rogeneous azeotropic distillation may also be made feasibler more economic by using two feeds (Wasylkiewicz et al.,000). Typically a heavy boiling compound that is partlymmiscible with the binary feed mixture is added as an uppereed, shown as F′ in Fig. 1(a).

Available graphical design methods for double-feedolumns (Wasylkiewicz et al., 2000; Urdaneta, 2005) do notxplain how to accommodate multiple heterogeneous stagesn the rectifying section, are computationally intensive or doot directly generate column design information. The bound-ry value method for a single-feed column with an integratedecanter is extended to include double-feed columns.

Additional inputs and specifications to the boundary valueethod are the composition and flow rate of a second feed,

he phase split ratio of the reflux stream and either the reboilatio or the thermal condition of the second feed. The prod-ct specifications must satisfy product purity constraints andhe overall material balance. Liquid phase splitting may occurnly in the decanter or two liquid phases may appear inhe column, depending on the reflux phase split ratio. Asith double-feed homogeneous distillation columns designs

Knapp and Doherty, 1994; Brüggemann and Marquardt, 2004),here may be a maximum reflux ratio above which the desiredeparation is no longer feasible, and there may be a limitedange of upper- to lower-feed flow ratios leading to feasibleolumn designs.

The new method generates the composition profile of theiddle section after generating the rectifying and stripping

rofiles as described in Section 2.1. The middle section profiles generated upwards starting from a specified stripping stagethe starting stage is a degree of freedom in design); the calcu-ation is similar to the calculation of the stripping profile. Eqs.7) and (8) represent the material and energy balance equa-ions for calculating the middle-section profile starting fromhe stripping section, as represented in Fig. 5:

VMm + B − FL)xM

m+1 − VMm yM

m − BxB − FLxFL = 0 (7)

(VMm + B − FL)hM,L

m+1 − VMm hM,V

m − BhB + FLhFL + QR = 0 (8)

where QR = (shS,V1 − (s + 1)hS,L

2 + hB)B if s = VS1/B is specified, or,

QR = DhD + BhB − FUhFU − FLhFL + QC if the upper feed conditionis specified.

If a middle-section profile leaves the composition space,it cannot intersect the rectifying profile. Feasible columndesigns, i.e. the number of theoretical stages in each section,can be identified from the transition from the stripping sec-tion to the middle section and from the intersection of themiddle-section and rectifying composition profiles.

2.2.1. Example – double-feed column with an integratedoverhead decanterA double-feed column separating a mixture of 1-propanol andwater using 1-butanol as an entrainer demonstrates the newapproach. The phase equilibrium of the 1-propanol/water/1-butanol mixture is calculated using the UNIQUAC modelfor the liquid phase and the ideal gas law for the vapourphase. Among the parameters available within HYSYS, theUNIFAC-VLE model provides the best estimate of the UNI-QUAC parameters; these are presented in Table 3. While thebinary azeotropes are well predicted (compared to the dataof Lee and Shen, 2003; Berg and Yeh, 1986), the two-phaseenvelope is not accurately represented by these parameters,when compared to the experimental data of Newsham andVahdat (1977). Table 4 presents the column specifications forthis example.

Fig. 6 illustrates the performance of columns where two-phase liquids are restricted to the decanter (Case i) and wherethey exist within the column (Case ii). The two cases are com-parable, as both designs require a similar number of stagesand the reboiler and condenser duties are similar. In this case,designs with multiple heterogeneous stages are found notto be as attractive as those in which two-phase liquids arerestricted to the decanter.

2.3. Column with an intermediate decanter

In columns with intermediate decanters, a heterogeneousliquid is withdrawn to a decanter – one phase is recoveredas a side-stream (typically the aqueous phase); the otheris returned to the column one stage below the draw stage(Schneider et al., 1997), as shown in Fig. 1(b). Alternatively, abuilt-in ‘donut’ decanter tray on the draw stage with an inter-face controller can facilitate the removal of the aqueous phase(Pucci et al., 1986; Ciric et al., 2000). As will be discussed inSection 3.3, the use of an intermediate decanter may reducethe number of columns and/or decanters required in a sepa-ration sequence. Design of such a column configuration couldbe carried out by trial and error, through simulation of alterna-

convergence issues. The method applying the concept of con-



Table 4 – Specifications of a double-feed column separating a mixture of 1-propanol, water and 1-butanol.

Upper feed, FU Lower feed, FL Distillate, D1 Bottoms, B1

Composition (mole fraction)1-Propanol 0a 0.5a 0.0001a 0.2159Water 0 0.5 0.9599 0.0150a

1-Butanol 1a 0.0a 0.0400a 0.7691Feed condition 1a 1a

Flow rate (kmol/h) 72a 40a 19.39 92.61

Reflux ratio 5

Phase split ratio of reflux Case (i): �1 /= �o1 0.25

Case (ii): �1 = �o1 0.45

�o1 = phase split ratio of the reflux that may cause the liquid leaving stage 1, next to the decanter, to be heterogeneous.

a Specified values.

III

S, xS, hS

VMm

Mm

Mm

h

V

,

,

,

y

2

m

B, xB, hB

Reboile r (stag e m=1 )

…

VSSS hV ,,,111

y

LSSS hL ,,,222

x

ILMm

IMm

IMm

h

L

,,

,

,

,

,

1

1

1

+

+

+

x

IILMm

IIMm

IIMm

h

L

,,

,

,

,

,

1

1

1

+

+

+

x

Smx

Sm 1−y

Fig. 7 – Schematic of a stripping section including an

tinuous distillation regions (Urdaneta, 2005) allows feasibilityto be assessed but does not give directly column design details.

In this development of the boundary value method,additional specifications for design of a column with an inter-mediate decanter are the composition of the side-stream,specified by choosing a water-rich composition on the bound-ary of the two-phase region, and the flow rate, which mustsatisfy the overall column material balance. As in the bound-ary value design methods discussed in Sections 2.1 and 2.2,the column design is obtained by the intersection of compo-sition profiles. The new method generates the compositionprofile of the middle section after generating the rectifying andstripping profiles as described in Section 2.1. Note that the rec-tifying section may or may not include an integrated decanter.

The middle section profile is calculated upwards from thestripping profile once the decanter position has been speci-fied – this is a degree of freedom in the design. Eqs. (9) and (10)present the material and energy balances for calculating themiddle section profile starting from the stripping section rep-resented in Fig. 7. It is assumed that the ‘return’ stage, to whichthe organic liquid phase from the intermediate decanter isreturned, is one stage below the decanter draw stream. Thecomposition, flow rate and enthalpy of the liquid leaving thereturn stage are known from the stripping section profile.

The material and energy balances over the column sec-tion (including the return stage and decanter) determine the

(i)

F

1-Butan olWater

1-Propa nol

2342

F

D1

B1

Fʹ

FU

11 =≠oφφ

Rectifying profil e

Str ipp ingpro file

Midd le section profi les

102

56

0.25

Fig. 6 – Liquid composition profiles for double-feed distillation cotwo-phase liquid appears only in the decanter (�1 = 0.25); Case (istage 2 �2 = 0) (Prayoonyong, 2009).

intermediate decanter.

flow rate and composition of the liquid entering the decanter.This liquid must be heterogeneous, if any separation is to takeplace in the decanter – this provides an additional feasibility
constraint. Multiple middle-section profiles may be gener-ated for a given stripping profile, as illustrated in Fig. 8. The
(ii)

1-Butan olWater

1-Propa nol

6102

234

2F

D1

B1

ʹ

FU

5

0.45 11== oφφ

φ2 = 0

Rectif ying profile

Str ipp ingpro file

Middle section

lumn with integrated overhead decanter. Case (i)i) heterogeneous liquid within column (decanter �◦

1 = 0.45;



(ii) (i)

1-Butanol Water

Acet one

F

D

B

Middle section profile from stage 4of the stripping section

Rectifying profile

Stripping profile

Middle section

Stage 4 of the stripping section

Acetone

1-Butanol Water

F

D

B

Middle section profil e from stage 3 of the stripping sect ion

Rectifying pro fil e

Stripping profil e

Middle section

Stage 3 of the stripping section

Fig. 8 – Composition profiles for a column with an intermediate decanter. Case (i) return stage 3: liquid on draw stage is noth quid

idstec

(

(

wos

2TtaeNatgT

atd

vcecaToaofl

eterogeneous – infeasible design; Case (ii) return stage 4: li

ntersection information and transition from stripping to mid-le section determine the number of stages in each columnection, providing initial design information and rather quali-ative insights. The details also allow design alternatives to bevaluated quantitatively, e.g. in terms of capital and operatingosts of the column.

VMm + S + B)xM

m+1 − VMm yM

m − SxS − BxB = 0 (9)

VMm + S + B)hM,L

m+1 − VMm hM,V

m − ShS − BhB + QR = 0 (10)

here QR = (shS,V1 − (s + 1)hS,L

2 + hB)B if s = VS1/B is specified,

r QR = DhD + ShS + BhB − FhF + QC if the feed condition ispecified.

.3.1. Example – column with an intermediate decanterhe recovery of 1-butanol, from acetone–butanol–water mix-

ures is relevant to dehydration of a butyl-beer mixture from butyl fermentation process (Pucci et al., 1986). The phasequilibrium of the mixture is obtained from HYSYS using theRTL-ideal gas property package. The interaction parameters

ij and ˛ij for the NTRL model are listed in Table 5. Experimen-al data validating the choice of model and parameters areiven in Berg and Yeh (1986) and Chronis and Amrhein (1980).able 6 presents the design specifications.

Fig. 8 presents composition profiles for the proposed sep-ration for two different side draw locations. It may be seenhat the feasibility of the design is sensitive to location of theraw stream.

These examples demonstrate that the extended boundaryalue method can be used to generate and evaluate novel andomplex column configurations for separating ternary het-rogeneous azeotropic mixtures. In some cases, these novelonfigurations will bring benefits in terms of capital and oper-ting costs for the particular separation to be carried out.he complex configurations can also have a significant impactn the structure and cost of the overall flowsheet. Therefore,

systematic approach is also needed to generate flowsheet
ptions including these complex configurations, so that theseowsheets may be evaluated.
on draw stage lies in two-phase region – feasible design.

3. Extended synthesis algorithm forternary heterogeneous azeotropic mixtures

The existing approach to synthesis discussed in Section 1.2is extended to include new features: heterogeneous columnfeeds where the compositions of the two-liquid phases lie indifferent distillation regions and double-feed columns withintegrated decanters. A new evolutionary algorithm to identifyopportunities to use columns with intermediate decanters isalso proposed. The synthesis algorithm, together with the twoextensions discussed in Sections 3.1 and 3.2, is represented inFig. 9.

3.1. Heterogeneous feed to distillation column

In the existing flowsheet synthesis approach, a heterogeneousliquid where its two equilibrium phases are in different distil-lation regions is always separated in a decanter before beingfurther separated. The option of introducing such a feed to acolumn directly is added to the algorithm, as such a columncan allow new, potentially attractive, flowsheet structures toemerge (indicated by “[two-phase feed]” in Fig. 9), comple-menting flowsheet options that use a decanter.

3.2. Double-feed column with integrated decanter

A double-feed heterogeneous azeotropic distillation columnis analogous to an extractive distillation column separating ahomogenous azeotropic mixture; as such, it allows the recov-ery of intermediate-boiling (‘saddle-type’) products. A heavycomponent must be added near the top of the column to actas an entrainer. However, during flowsheet synthesis, it is notknown what streams are available for this purpose. This sep-aration option is added to the synthesis algorithm, initiallyassuming the heaviest component is the entrainer and laterupdating the assumption, once more information is availableabout what streams are present in the whole sequence. Theextended boundary value method described above provides afeasibility check for the proposed column; the ratio of upper
to lower feed flow rates is an important variable as it has astrong influence on feasibility.



Table 5 – Acetone/water/1-butanol mixture: NRTL interaction parameters (aij and ˛ij) from HYSYS 2004.1 (elements in rowand column represent component i and j, respectively). The binary parameters bij are all zero.

Component i, j Acetone Water 1-Butanol

aij

Acetone – 1299.395 594.609Water 750.318 – 570.1361-Butanol −85.730 2794.666 –

˛ij

Acetone – 0.586 0.300Water 0.586 – 0.4701-Butanol 0.300 0.470 –

Table 7 – UNIQUAC interaction parameters (aij) for ethylacetate/water/n-butyl acetate mixture from HYSYS2004.1 (components in row and column representingcomponent i and j, respectively).

Component Ethylacetate

Water n-Butylacetate

Ethyl acetate – 151.642 49.355Water 757.585 – 738.169n-Butyl acetate −36.183 398.290 –

3.3. Column with intermediate decanter

A column with an intermediate decanter is a column in whicha heterogeneous liquid on a stage within a column is with-drawn, separated in a decanter to produce a side streamproduct and a second stream that is returned to the column,typically one stage below the side draw. This complex columnconfiguration is commonly used to recover water (Ciric et al.,2000) and can reduce significantly energy requirements and/orcapital cost (Stichlmair and Fair, 1998).

A sequence with two columns and with two decanters thatperform similar separations (Fig. 10a) can be systematicallymerged into a sequence with two columns and a commondecanter (Fig. 10b) and subsequently into a column with anintermediate decanter (Fig. 10c). An algorithm is added to theoverall sequence synthesis algorithm to identify opportunitiesto use this complex configuration.

3.4. Flowsheet feasibility assessment and evaluation

A number of feasible sequences that separate a mixture into(nearly) pure components may be generated; these sequencesneed to be quantitatively evaluated, e.g. in terms of economicperformance. For design of an individual column, the materialbalance of the sequence determines the flow rates and compo-sitions of all streams in the sequence. For sequences withouta recycle loop, this calculation is straightforward, proceedingsequentially in the direction of material flow. For sequencesthat have one or more recycle loops, where the flow rate andcomposition of the recycle are not known at the outset, itera-tive approaches using ‘tear’ streams are need. The feasibility ofproposed separations may be sensitive to the recycle composi-tion and flow rate. Therefore, the feasibility of each separation(distillation or decanting) must be assessed once the materialbalance of the flowsheet is solved (Tao et al., 2003). Then eachcolumn in the sequence can be designed for its best perfor-mance, e.g. applying the boundary value method for a range
of key design variables, to allow the overall flowsheet to beevaluated.
Table 6 – Specifications of a column with an intermediate decan

Feed, F Distillate

Composition (mole fraction)Acetone 0.0995a 0.9724a

Water 0.7859a 0.0276

1-Butanol 0.1146 7 × 10−8a

Feed condition 1a

Flow rate (kmol/h) 100a 10.10

Reflux ratio

a Specified values.

3.5. Example – n-butyl acetate–ethyl acetate–waterseparation

The synthesis of sequences for separation of an industri-ally relevant ethyl acetate–water–n-butyl acetate mixture(Van Acker et al., 2000) illustrates the approach. Alternativesequences are generated using existing synthesis algorithms(Tao et al., 2003; Vanage, 2005) and the new extensions. Thecomposition and flow rate of the feed stream and the desiredproduct purities are given; the pressure of the separation unitsis fixed. The residue curve map is constructed, and presentedin Fig. 11, showing the azeotropes, distillation boundaries and‘compartments’ (continuous distillation regions), the hetero-geneous boiling envelope and associated vapour line for thismixture.

The UNIQUAC-ideal gas property package in HYSYS2004.1 is used to predict the phase equilibrium of ethylacetate/water/n-butyl acetate mixture. The interactionparameters aij are listed in Table 7. The interaction param-eters bij are all zero. Experimental data of Lee et al. (1996),Gomis et al. (2000) and Berg and Yeh (1986) support the use ofthis model and parameters.

In this case study, the feed to be separated is a saturatedliquid at 1 atm. It consists of 15 mol% ethyl acetate, 70 mol%water, and 15 mol% n-butyl acetate. The feed flow rate is100 kmol/h. All separation units to separate the feed streamwill be operated at 1 atm. Separation sequences will be gener-
ated to recover all components in the mixture with 97 mol%minimum purity.
ter separating a mixture of acetone, water and 1-butanol.

, D Sidestream, S Bottoms, B

0.0016a 1 × 10−6a

0.9855a 0.00240.0129 0.9976a

79.44 10.46

20



Pick a str eam in a sequen ce

Stop

Use a de canter

Two stre ams aregene rated

Use a col umn

Two stre ams are gene rated

Classify as a non-p rodu ct stream

Match a desi red purity?

Heterog ene ous liquid?

Are all str eams conside red?

Liquid composition s in different distilla tion

region s?

Start

Possible t o use a column?

Direct,

Indirect split, and/or

Saddle-type split

o

yes

no

yes

no

no

no

yesyes

yes

no

Yes[two-p hase fee d]

Fig. 9 – Existing algorithm extended to include double-feed column and use of heterogeneous column feed (wheree regio

flwf

Fi

quilibrium liquid compositions are in different distillation

The synthesis algorithm is applied to generate fourowsheets – Option 1 (Fig. 12) uses single-feed columns
ith integrated decanters; Option 2 (Fig. 13) applies a two-
eed column; Option 3 (Fig. 14) introduces a heterogeneous

F

B1

D1

S

L1

B2

W 2

C-1

D-1 C-2

D-2

W1

F

a b

C-1

ig. 10 – Two columns with two decanters carrying out similar sentermediate decanter.

ns).

liquid (with its equilibrium compositions in differentdistillation regions) to a single-feed column; Option 4 (Fig. 11)
evolves Option 3 further to identify a column with an inter-mediate decanter.
B1

D c1

B2

D-1

C-2

W1

F

B

D

WD-1

C-1

parations can be merged to give a column with an



Water (100oC)

n-Butyl acetate (126.1oC)

Ethyl acetate (77.1oC)

azeotrope (91 oC)

Region 1

azeotrope (71 oC)

Feed

Distilla tion bounda ry

Region 2

Residue curve

Vapour line

Azeotrope

Two-liquid phase region

Liquid-liquid tie line

Compartment boundary

Fig. 11 – Residue curve map for n-butyl acetate–ethyl acetate–water mixture at 1 atm.

Water n-Butyl ace tat e

Ethyl acetate

D1

L1

F

W1 B1

L2, L3

B2

W2, W3

C-1

C-2

D-2,

D-3 D-1

F

B1

L2

C-1B2

W3C-2

D-1

D-2

D-3

W1

L1

W2

Water

Water

n-But yl acetate

Water

Ethyl acetate

(Note that prod uct D1 is the net

decanter product, i.e. L2 + W2.)

Fig. 12 – Option 1: Sequence with single-feed columns and integrated decanters.

Water n-Buty l acetate

Eth yl a cetate

L1

F

W1

D-1

B1

D1

L2

B2

D2

W2

C-2

D-2

C-1

FU

F

B1

L2

C-1

D-1

D-2

W1

L1 W2

Water

Water

B2

D2

C-2

n-Butyl acetate rich

Ethyl acetate

n-Butyl acetate

BR

FU

(Note that prod uct D1 is the net

decanter product, i.e. L2 + W2.)

Fig. 13 – Option 2: Sequence with double-feed column and recycle.



Watern-Butyl acetate

Ethyl acetate

F

C-1

L1

W1, W2

B1 L3W3, W4

B3

B2

C-2

C-3D-3, D-4

D-1,

D-2

F

B1

L1

C-1

D-1

W1

Water

D-3

W3

Water B3

W4C-3

D-4

Water

n-But yl acetate

B2

W2C-2

D-2

Water

Ethyl acetate

L3

colu

Oaavctfebr

vcaicrac

a

Fig. 14 – Option 3: Sequence with single-feed

To select the most economically attractive sequences,ptions 1–4 are evaluated with respect to their total annu-lised costs. Each separation unit in a sequence is designednd evaluated for a range of values of relevant designariables so that its total near-minimum annualised costan be estimated. The cost of a decanter is a function ofhe size of the decanter which in turn depends on theeed flow rate of the decanter. The cost of a column isstimated based on the column design details such as num-er of stages, heat duties, product compositions and flowates.

Each column in a sequence is designed using the boundaryalue method appropriate to the type of the column. Feasibleolumn designs are evaluated for a range of reflux ratios and

range of reflux phase split ratios, in cases that the columns coupled with an integrated decanter. For the double-feedolumn in Option 2, column designs are also generated for aange of upper-to-lower feed rate ratios. Based on the totalnnualised cost, the economically near-optimal design of theolumn can thus be identified.

Table 8 presents a summary of design results and totalnnualised costs of each unit, taking into account the installed

Table 8 – Summary of design and costs of flowsheet options.

Option Separation unit NR NS NTotal

Option 1 (Fig. 12) D-1 – – –

C-1, D-2 14 9 22

C-2, D-3 2 11 12

Option 2 (Fig. 13) D-1 – – –

C-1, D-2 2 6 18

C-2 8 8 15

Option 3 (Fig. 14) C-1, D-1 10 18 27

C-2, D-2 2 11 12

D-3 – – –

C-3, D-4 3 5 7

Option 4 (Fig. 15) C-1, D-1, D-2 11 20 32

C-2, D-3 2 11 12

TAC = total annualised cost (3-year period and 5% interest rate); NR, NS: nuand condenser duty, respectively.

mns, and heterogeneous feed to Column 1.

equipment costs and operating costs for each flowsheet alter-native. Of the four alternatives, the most economic is Option1 in which two simple single-feed columns with integrateddecanters are used. The economic performance is very similarto that of the sequence consisting of a double-feed col-umn placed after a decanter (Option 2). Compared with thesequences in which the heterogeneous feed is first separatedin a column (Options 3 and 4), Options 1 and 2 (a decanter isfirst used) are considerably more attractive. This is becauseabout 65% of the feed is simply removed as nearly pure waterby using the decanter. Having some of the feed removed sub-stantially reduces the feed to subsequent columns. The costsof Options 3 and 4 are very similar. Option 4 is generated bymodifying the sequence in Option 3; however, in this case,the column with an intermediate decanter does not providecost savings – although Option 4 requires fewer columns anddecanters than Option 3, the costs are dominated by Column 1which requires both more stages and more heating in Option4 than in Option 3.

Note that the estimations of capital costs include only

QC (kW) QR (kW) TAC (£/year) Total cost (£/year)

– – 5,000 318,700423 463 183,100356 360 130,600

– – 5,000 330,500348 385 154,500421 464 171,000

458 477 183,600 375,200377 382 136,400

– – 5,00055 92 50,200

484 540 236,400 372,800378 382 136,400

mber of rectifying and stripping stages, respectively; QR, QC: reboiler

columns and decanters. Further details are presented inPrayoonyong (2009).



Water n-Butyl acetate

Ethyl acetate

F

C-1

L1

W1, W3

W2

B1

B2

C-2

D-2

D-1,

D-3

F

D-2

C-1

W1

D-1

WaterB2

W3C-2

D-3

Water

Ethyl acetate

L1

W2

Water

B1

n-bu tyl acetate

n 3,
Fig. 15 – Option 4: Sequence evolved from Optio
The extended synthesis approach allows additionalsequences employing a wider range of features to be gen-erated. In the case presented, however, complex columnconfigurations are not found to be economically attractive.Among the four options, the sequence using only simplecolumns (Option 1) has the lowest cost as most of the water inthe feed is removed by using the decanter. In sequence syn-thesis in general, however, including those complex columnconfigurations is valuable because they may make a signif-icant difference to separation feasibility and/or the processeconomics.

4. Conclusions

Complex column configurations for separating ternary het-erogeneous azeotropic mixtures may offer advantages oversimpler configurations with respect to feasibility and cost.However, the complexity of these columns presents a chal-lenge for design and simulation as well as for flowsheetsynthesis. Extended boundary value methods presented inthis paper address single- and double-feed columns with inte-grated decanters and columns with intermediate decanters.These methods allow multiple designs to be generated andscreened for feasibility and cost, and also provide useful ini-tial values for conventional process simulation software suchas HYSYS. The new methods overcome limitations of existingdesign methods, avoid the time-consuming, iterative option ofdesign-by-simulation, and allow multiple designs to be gener-ated and hence screened prior to more detailed simulationand design.

The boundary value method provides a useful approachfor preliminary design and evaluation of ternary azeotropicdistillation columns and facilitates generation and screeningof alternative flowsheets.

A drawback of boundary value methods is that the productcompositions must be fully specified and that no system-atic approach exists for doing so. The feasibility and designresults can be highly sensitive to the mole fraction of minorcomponents in the products, especially for multicomponentmixtures. Another limitation of boundary value methods isthat fractional numbers of stages may be found; such resultslimit the usefulness of the boundary value methods for prelim-
inary design or for initialisation of more conventional processsimulations.
including column with intermediate decanter.

This work extends existing approaches for synthesisingsequences to separate ternary heterogeneous azeotropic mix-tures. Two novel options are introduced: heterogeneous liquidstreams may be separated in either a decanter or a column,and double-feed columns with an integrated decanter can per-form a saddle-type split. A new algorithm has been proposedfor evolving sequences employing columns with intermedi-ate decanters. The new features in the synthesis approachincrease the number of separation flowsheet options, poten-tially allowing more attractive sequences to be found. Theproposed approach is illustrated for the separation of a mix-ture of ethyl acetate–water–n-butyl acetate.

Separation sequences generated from the proposed syn-thesis approach can be evaluated using the iterativeapplication of boundary value methods (Prayoonyong, 2009;Prayoonyong and Jobson, 2010) (e.g. varying reflux ratios,phase reflux ratios, and upper-to-lower feed rate ratios). Thetotal annualised cost of a flowsheet can then be estimated toallow quantitative evaluation of flowsheet alternatives.

The feasibility of a proposed separation in a single-feed col-umn may be evaluated using the common saddle test (Rookset al., 1998). A limitation of the synthesis approach is the lackof such a test for two-feed columns, for which feasibility maybe assessed using the new boundary value method, given fullproduct specifications. The proposed synthesis approach maybe readily extended to consider the effects of operating pres-sure and decanter temperature.

Nomenclature

B molar flow rate of bottom productC columnD molar flow rate of distillate productD decanterF, F′ molar flow rate of feed and second feed, respectivelyh molar enthalpy flow rateL molar flow rate of liquidQ heating or cooling dutyr reflux ratio (molar)s reboil ratio (molar)
S molar flow rate of sidestreamx vector of liquid mole fractions



yV�

SBCDFLLMmRRSSUVI

A

TRt

R

A

B

B

C

C

C

G

G

H.

K

K

K

vector of vapour mole fractions molar flow rate of vapour

phase split ratio (molar): flow of heavy liquid to totalliquid flow

ubscripts and superscripts bottom product

condenserdistillatefeed

liquid phase lower feed

middle section, n stage number

rectifying section reboiler

stripping section side stream

upper feed vapour phase

, II liquid phase 1 or 2, respectively

cknowledgement

he authors gratefully acknowledge the sponsorship of theoyal Thai government for Dr Prayoonyong’s PhD studies athe University of Manchester.

eferences

mminudin, K.A., Smith, R., Thong, D.Y.C., Towler, G.P., 2001.Design and optimization of fully thermally coupled distillationcolumns. Part 1: Preliminary design and optimizationmethodology. Chem. Eng. Res. Des. 79 (A7), 701–715.

erg, L., Yeh, A.I., 1986. The breaking of ternaryacetate–alcohol–water azeotropes by extractive distillation.Chem. Eng. Commun. 48, 93–101.

rüggemann, S., Marquardt, W., 2004. Shortcut methods fornonideal multicomponent distillation: 3. Extractivedistillation columns. AIChE J. 50 (6), 1129–1149.

astillo, F.J.L., Thong, D.Y.-C., Towler, G.P., 1998. Homogeneousazeotropic distillation. 1. Design procedure for single-feedcolumns at nontotal reflux. Ind. Eng. Chem. Res. 37 (3),987–997.

iric, A.R., Mumtaz, H.S., Corbett, G., Reagan, M., Seider, W.D.,Fabiano, L.A., Kolesar, D.M., Widagdo, S., 2000. Azeotropicdistillation with an internal decanter. Comput. Chem. Eng. 24(11), 2435–2446.

hronis, J.H., Amrhein, M., 1980. Vapour–liquid EquilibriumModel for a Partially Miscible Multicomponent System:Isopropanol–Acetone–Ethanol–n-Butnol–Water. Oak RidgeNational Laboratory.

ani, R., Jørgensen, S.B., 1994. Multiplicity in numerical solutionof non-linear models: separation processes. Comput. Chem.Eng. (Suppl.) 18, S55–S61.

omis, V., Ruiz, F., Asensi, J.C., 2000. The application of ultrasoundin the determination of isobaric vapour–liquid–liquidequilibrium data. Fluid Phase Equilib. 200, 245–259.

ölz, S., 2006. Curve intersect.http://www.mathworks.co.uk/matlabcentral/fileexchange/8908

napp, J.P., Doherty, M.F., 1994. Minimum entrainer flows forextractive distillation: a bifurcation theoretic approach. AIChEJ. 40 (2), 243–268.

ovach, J.W., Seider, W.D., 1987. Heterogeneous azeotropicdistillation: experimental and simulation results. AIChE J. 33(8), 1300–1314.

raemer, K., Harwardt, A., Skiborowski, M., Mitra, S., Marquardt,
M., submitted for publication. Shortcut-based Design ofMulticomponent Heteroazeotropic Distillation, ChERD.
Lee, L.S., Chen, W.C., Huang, J.F., 1996. Experiments andcorrelations of phase equilibria of ethanol–ethylacetate–water ternary mixture. J. Chem. Eng. Jpn. 29 (3),427–438.

Lee, L.-S., Shen, H.-C., 2003. Azeotropic behaviour of awater + n-propanol + cyclohexane mixture using cyclohexaneas an entrainer for separating the water + n-propanol mixtureat 760 mm Hg. Ind. Eng. Chem. Res. 42 (23), 5905–5914.

Levy, S.G., Van Dongen, D.B., Doherty, M.F., 1985. Design andsynthesis of homogeneous azeotropic distillations. 2.Minimum reflux calculations for nonideal and azeotropiccolumns. Ind. Eng. Chem. Fundam. 24 (4), 463.

Newsham, D.M.T., Vahdat, N., 1977. Prediction ofvapour–liquid–liquid equilibria from liquid–liquid equilibria.Part I: Experimental results for the systemsmethanol–water–n-butanol, ethanol–water–n-butanol andn-propanol–water–n-butanol. Chem. Eng. J. 13, 27–31.

Nishida, N., Stephanopoulos, G., Westerberg, A.W., 1981. A reviewof process synthesis. AIChE J. 27, 321–351.

Ou-yang, F., Wang, G., Gao, W., 1985. Determination andcorrelation of liquid–liquid equilibrium data for the ternarysystems water–ethyl acetate–acetic acid and water–butylacetate–acetic acid. Huagong Xuebao 36 (1), 110–118.

Peters, M.S., Timmerhaus, K.D., West, R.E., 2003. Plant Design andEconomics for Chemical Engineers, 5th ed. McGraw-Hill, NewYork.

Pham, H.N., Ryan, P.J., Doherty, M.F., 1989. Design and minimumreflux for heterogeneous azeotropic distillation columns.AIChE J. 35 (10), 1585.

Pham, H.N., Doherty, M.F., 1990. Design and synthesis ofheterogeneous azeotropic distillations – 3. Column sequences.Chem. Eng. Sci. 45 (7), 1845–1854.

Prayoonyong, P., 2009. Synthesis and Design of TernaryHeterogeneous Azeotropic Distillation Processes includingAdvanced Complex Column Configurations. PhD Thesis. TheUniversity of Manchester, UK.

Prayoonyong, P., Jobson, M., 2010. Synthesis of novel ternaryheterogeneous azeotropic distillation flowsheets, PO16,Distillation and Absorption 2010, Netherlands, September2010.

Pucci, A., Mikitenko, P., Asselineau, L., 1986. Three-phasedistillation. Simulation and application to the separation offermentation products. Chem. Eng. Sci. 41 (3), 485–494.

Repke, J.U., Wozny, G., 2002. Experimental investigations ofthree-phase distillation in a packed column. Chem. Eng.Technol. 25 (5), 513–519.

Rooks, R.E., Julka, V., Doherty, M.F., Malone, M.F., 1998. Structureof distillation regions for multicomponent azeotropicmixtures. AIChE J. 44 (6), 1382.

Schneider, R., Wozny, G., Fieg, G., 1997. Optimization of anindustrial three phase distillation column train withexperimental verification. Comput. Chem. Eng. (Suppl.) 21,S1131–S1136.

Smith, 2005. Chemical Process Design and Integration. Wiley (Ch.2.5).

Stichlmair, J.G., Fair, J.R., 1998. Distillation: Principles andPractice. Wiley-VCH, New York.

Tao, L., Malone, M.F., Doherty, M.F., 2003. Synthesis of azeotropicdistillation systems with recycles. Ind. Eng. Chem. Res. 42 (8),1783–1794.

Thong, D.Y.-C., Jobson, M., 2001. Multicomponent homogeneousazeotropic distillation. 3. Column sequence synthesis. Chem.Eng. Sci. 56 (14), 4417–4432.

Urdaneta, R.Y., Bausa, J., Bruggemann, S., Marquardt, W., 2002.Analysis and conceptual design of ternary heterogeneousdistillation processes. Ind. Eng. Chem. Res. 41 (16),3849–3866.

Urdaneta, R.Y., 2005. Targeting and Conceptual Design ofHeteroazotropic Distillation Processes. PhD Thesis. ProcessSystems Engineering, RWTH Aachen University, Aachen,Germany.

Van Acker, P.E., Mathieu, O., Milner, R.J., Pacynko, W.F., 2000. EsterCo-production. US patent 6,093,845.

http://www.mathworks.co.uk/matlabcentral/fileexchange/8908



Vanage, P.V., 2005. Synthesis of Ternary HeterogeneousAzeotropic Reaction-Separation Systems. PhD Thesis.University of Manchester, Manchester, UK.

Van Dongen, D.B., Doherty, M.F., 1985. Design and synthesis of
homogeneous azeotropic distillations. 1. Problem formulationfor a single column. Ind. Eng. Chem. Fundam. 24 (4), 454–463.
Wasylkiewicz, S.K., Kobylka, L.C., Castillo, F.J.L., 2000. Optimaldesign of complex azeotropic distillation columns. Chem.Eng. J. 79, 219.

Wasylkiewicz, S.K., Kobylka, L.C., Castillo, F.J.L., 2003. Synthesis
and design of heterogeneous separation systems with recyclestreams. Chem. Eng. J. 92 (1–3), 201–208.

Flowsheet synthesis and complex distillation column design for separating ternary heterogeneous...

Documents

Transcript of Flowsheet synthesis and complex distillation column design for separating ternary heterogeneous...