Design and Control of a Complete Azeotropic Distillation System

Design and Control of a Complete Azeotropic Distillation System Incorporating Stripping Columns for Isopropyl Alcohol Dehydration Wen-Teng Chang,?Chi-Tsung Huang,?and Shueh-Hen Cheng *,?

?Department of Chemical and Materials Engineering,Tunghai University,Taichung 40704,Taiwan

The heterogeneous azeotropic distillation systems are fre-quently encountered in the chemical process industry for separating mixtures of close boilers or breaking azeotropes.One advantage of using this type of separation system is that one can utilize a decanter to cross a distillation boundary for obtaining high purity products.It is oftentimes a continuous process in which entrainer selection is a critical element in maintaining the product quality.One often applies this technique in industrial dehydration processes,such as alcohol dehydration and acetic acid dehydration processes.Widagdo and Seider 1gave a good review on azeotropic distillation.They pointed out that the heterogeneous azeotropic distillation column is difficult to operate and control.Pham and Doherty 2studied three different separation sequences for the ethanol ?water ?benzene system,

which includes four-column,three-column,and two-column heterogeneous distillation sequences.The three-column sequence contains a preconcentrator column,an azeotropic column,and an entrainer recovery column.The two-column sequence,in fact,combines the preconcentrator column and the recovery column as https://www.wendangku.net/doc/753037258.html,ter,Ryan and Doherty 3pointed out that the four-column sequence has no advantage over the three-column sequence,and that the three-column sequence has lower operat-ing costs but higher capital costs than the two-column sequence,so that the total annualized cost is about the same for both sequences.Recently,Luyben 4proposed a control strategy of a three-column sequence for ethanol dehydration.Chien et al.5

proposed a design and control system for a two-column sequence

of isopropyl alcohol (IPA)dehydration.Arifin and Chien 6com-pared two-column and three-column sequences of IPA dehy-dration for a diluted IPA feed.They found that the operating cost column sequence,but the capital cost of the former is more than that of the latter.Nevertheless,their results are almost the same as that of Ryan and Doherty.3Furthermore,Wu and Chien 7pres-ented a design and control study for pyridine dehydration using a two-column sequence.

On the other hand,distillation is one of the most energy intensive processes in the chemical and petrochemical industries.Due to rising oil prices and an increasing demand for reducing CO

2emissions,recent process design technologies

tend to go for energy-saving options.In this context,this study

explores the possibility of improving the design and control of an existing heterogeneous azeotropic distillation system for purposes of reducing energy consumption.

2.STEADY-STATE DESIGN OF THE OVERALL PROCESS

The study considers the design of an isopropyl alcohol (IPA)dehydration system with a feed composition of 50mol %IPA and 50mol %water and a feed rate of 100kmol/h at 25°C.Cyclohexane is used as an entrainer for the system,and a decanter is employed for the scheme and is operated at 40°C.Purities are set at 99.9999mol %IPA for the IPA product stream and 99.9mol %H

2O for the wastewater stream.These

specifications were,in fact,adopted by Arifin and Chien.6

Similar to the approach by Arifin and Chien,6Aspen Plus is Received:September 5,2011

Revised:December 22,2011

Accepted:January 4,2012

Published:January 4,2012

employed for the rigorous steady-state simulation in this study.The vapor phase of the system is assumed to be ideal,and the NRTL model is used to describe the nonideality of the liquid phase.In addition,a set of NRTL binary parameters is obtained from Wang et al.9All other physical properties are obtained from the Aspen Plus data bank.A three-column sequence,which was studied by Arifin and Chien,6is shown in Figure 1and is named as Scheme 1in this study.As shown in Figure 1,the fresh dilute feed stream flows into a preconcentrator column (C101).The bottom product of the preconcentrator column is 99.9mol %of water.The distillate of the preconcentrator column is mixed with a recycle stream from an entrainer recovery column (C301)and flows into a heterogeneous azeotropic column (C201).The distillate composition of the preconcentrator column is close to that of the IPA-water azeotrope.A small makeup stream of entrainer is added to the decanter.The organic entrainer-rich phase of the decanter is refluxed back to the heterogeneous azeotropic column,and the aqueous phase is sent to the recovery column.Their respective bottom products are 99.9999mol IPA and 99.9mol %water.An alternative two-column sequence,which

was proposed by Arifin and Chien,6is shown in Figure 2and named as Scheme 2here.One difference from these two Schemes is that the feed stream of Scheme 2,mixed with the aqueous-phase of the decanter,flows directly into the recovery column,and there is no preconcentration column in Scheme 2.According to Ryan and Doherty 3and Chien et al.,5the three-column sequence (Scheme 1)can save more energy than

the

Figure 1.Operating conditions for Scheme

1.

Figure 2.Operating conditions for scheme 2.

two-column sequence (Scheme 2).From the simulation results of Arifin and Chien,6it can be found that the reflux ratio used either in the preconcentration column or the recovery column

in Scheme 1is quite small.Also,the variation of the reflux ratio in either of these two distillation columns may not be so impor-tant,since there is no rigid specification in the top product stream.This can be interpreted as that either of these two columns requires a very pure bottoms product,but a pure top product is not needed.Thus,these two conventional distillation columns can be replaced by two stripping columns.A stripping column can be thought of as a conventional distillation column with no external reflux in the rectifying section.In contrast,one shortcoming associated with the conventional distillation column is that it has a reboiler to vaporize liquid in the stripp-ing section and also has a condenser to condense vapor in the rectifying section.From the perspective of energy conservation,a stripping column is a tower without a condenser,in which energy carried by the overhead vapor is conserved instead of being removed in a condenser,and is considered to be more energy-efficient here than the conventional distillation.In addi-

tion,using the stripping column to replace the conventional distillation column can not only reduce energy cost but also cut

capital cost.Accordingly,a modified separation system,named Scheme 3here,is proposed in this study,for which a simple process flow diagram (PFD)is presented in Figure 3.As shown in Figure 3,the fresh feed (stream 1)flows into a preconcentration stripping column (C101),yielding a high-purity water stream (99.9mol %)at the bottom (stream 10)and an overhead vapor stream (stream 3).This vapor stream then mixes with the overhead stream (stream 9)of another stripping column (C301)to form a feed stream (stream 4)to a heterogeneous azeotropic column (C201).An ultrahigh purity (99.9999mol %)IPA product,i.e,stream 11,can be obtained in the bottoms.The overhead vapor of C201(stream 5),whose composition is close to that of the ternary azeotrope,then passes through a total condenser and is cooled to 40°C.The condensate from the condenser is split into two liquid phases at the decanter where the organic phase (stream 7)is refluxed to the top of the heterogeneous azeotropic column,and the aqueous phase (stream 8)flows into the entrainer-recovery stripping column (C301).A product stream (stream 12)containing 99.9mol %water can be easily obtained at its bottom of this stripping column.In order to facilitate the transport of overhead vapor streams from C101and C301,the top pressures of C101and C301are both set at 1.1atm,while that of C201is set at 1.05atm,instead of fixing all column pressures at 1atm as in the work of Arifin and Chien.6A stream summary for this flowsheet based on rigorous simulation results with Aspen Plus is presented in Table 1.Residue curve map (RCM)and conceptual material balance (MB)lines for Scheme 3including MB lines for Columns C101,C201,and C301can be visualized in Figure 4with major process streams being labeled.

3.ECONOMIC ANALYSIS

In order to evaluate the cost effectiveness of these three schemes,viz.the flowsheets in Figures 1?3,an economic analysis is performed in this study.All of these schemes are designed under the same operating conditions,i.e.the same inputs and outputs,the same cooling water temperature,and the same steam conditions,etc.These operating conditions in fact come from Arifin and Chien.6Column diameter for each distillation unit is estimated using the Aspen Plus,and a tray spacing of 0.6m and flooding of 80%are assumed.The overall tray efficiency for each column is conservatively assumed to be 50%.5,6Thus,the height of each column (H )is estimated from the number of actual plates (N actual )and disengagement heights

as

Figure 3.Operating conditions for Scheme 3.

=?+H N (1)(tray spacing)disengagement heights actual (1)According to Truton et al.,9we add 1.2m for vapor diseng-agement at the top and 1.8m at the bottom for liquid level and reboiler return.In addition,the overall heat transfer coefficients are 850and 1140(W/m 2°C)for condenser and reboiler,respectively.The correction factor 9of 0.9for logarithmic mean temperature difference (LMTD)is used for each heat exchanger.Low pressure steam is assumed to be available at 5barg (160°C)and cool-ing water of 30°C is employed.A stream factor 9of 0.95,which is the fraction of working days in a year,is used for calculating the yearly cost of utilities.The decanter is sized with a liquid holdup of 20min half full and an aspect ratio (length-to-diameter ratio)of 3.It should be noted that the above design specifications together with utility costs are obtained from Truton et al.9Table 2gives an equipment list with sizing results for all major equipment units.Economic analysis in this study follows closely the method of

Turton et al.9To estimate the capital cost,bare module equip-ment cost (C BM )is calculated for each unit.The

bare module cost (C BM )is a function of purchased cost (C p

0),bare module cost factor (F BM ),number of trays (N )in the column,and

quantity factor (f q )based on the number of trays in the column,i.e.

==C C F C C F Nf for vessels and heat exchangers for sieve trays in the column BM p BM BM p BM q 00(2)The purchased cost for base conditions (C p 0)is calculated by =++C K K A K A log

log ()[log ()]p 10012103102(3)where K 1,K 2,and K 3are constants specific to the type of equip-ment,and A is the capacity/size parameter of the

equipment.

Figure 4.RCM and material balance lines for Scheme 3.

Table 1.Stream Summary Based on Simulation Results for Scheme 3

stream number 1234567pressure (atm) 1.20 1.00 1.10 1.10 1.05 1.00 1.00temperature (°C)25.025.082.782.065.040.040.0vapor fraction 0010.99998100mole frac

CyH 0100.0304530.5346710.5346710.865440IPA 0.500.6103200.5862160.2602520.2602520.128140H2O 0.500.3896800.3833310.2050770.2050770.006420total flow (mol/min)1666.67 1.34×10?41364.913516.486698.696698.694014.73stream number 891011121314pressure (atm) 1.00 1.10 1.16 1.20 1.16 1.20 1.20temperature (°C)40.081.6103.586.9103.740.040.0vapor fraction 0100000mole frac

CyH 0.0398990.0497710.000000trace trace 0.8654400.039899IPA 0.4578750.5709250.0010000.9999990.0010000.1281400.457875H2O 0.5022270.3793040.9990000.0000010.9990000.0064200.502227total flow (mol/min)2683.972151.57301.76832.52532.394014.732683.97

The bare module cost factor (F BM ),a function of operating pressure and the construction materials,is calculated as =+=F B B F F F for vessels and heat exchanger 1for sieve trays in the column BM M P BM 12(4)where B 1and B 2are constants depending on the equipment type.F M is material factor,and F P is pressure factor.Carbon steel is chosen as the material of construction throughout,since the process under study is operating near ambient pressure and no corrosive chemicals are present.The quantity factor (f q )based on the number of trays (N )in the column,is calculated as =+?<=≥f N N N f N log 0.47710.08516log 0.3473(log )for 201for 20q q 1010102(4a)The bare module costs (C BM )for all of the major equipment units are shown in Table 2.The total bare module cost for each scheme is also presented in Table 2.In addition,the unit cost of 13.28($/GJ)for low-pressure steam and the unit cost of 0.354($/GJ)for cooling water are used according to Truton et al.9The annual steam and cooling water costs,together with the total annual utility cost for each scheme,are given in Table 3.It should be noted that the utility cost here,which is equivalent to the operating cost in Arifin and Chien,6is the sum of the steam cost and the cooling water cost.On the other hand,the total annual cost (TAC)is employed here as a measure of economic potential,which does not involve sales revenues for products and is used for preliminary cost estimates when comparing alternative flowsheets during process synthesis.According to Seider et al.,10the TAC in this study,with which several alternative distillation sequences are examined,can be calculated as =+×i TAC Annual utility cost Total capital cost min (5)The term i min is a minimum acceptable (or threshold)return on investment.The total capital cost is,in fact,the total bare-module costs for all columns and their auxiliaries based on the current Chemical Engineering Plant Cost Index (CEPCI).The bare module cost (C BM )shown in Table 2is based on the CEPCI of 397,which is an average value during the period of May to September of 2001.9The current CEPCI on December of 2010is found to be 560.4.Thus,the total capital cost for each scheme can be

calculated as

=×Total capital cost

Total bare module cost 560.4/397(6)

Furthermore,Seider et al.10recommend that i min be taken as 0.2.Based on a three-year payback period,Arifin and Chien 6adopt i min =1/3.Olujic et al.11consider that i min =0.1,according to an assumed plant lifetime of 10years.A comparison of all the calculated TACs as well as cost ratios (determined with reference to the cost of Scheme 1)is summarized in Table 4for the three alternative schemes using

various values of i min .Pump costs however are not included in

the TAC,like those in the work of Arifin and Chien,6as they

are relatively small when compared with the costs of other equipment.It is found from Table 4that the required TAC for

Scheme 1and Scheme 2are almost the same at different i min values.Scheme 3,however,requires much less TAC than the other two schemes.Despite the fact that Scheme 3has one more column than Scheme 2,Scheme 3consumes less energy than Scheme 2.On the other hand,Scheme 3has the least total capital cost among all the alternative schemes.

4.DYNAMIC SIMULATION AND BASIC CONTROL STRUCTURE

Dynamic simulation of Scheme 3is established with a program written in FORTRAN language.The dynamic simulation program is built on an equilibrium-stage model of Franks,12and Figure 5shows the computational flowchart associated with the model for stage j .A cubic spline interpolation (Rovaglio and Doherty 13)can be used to fit the data,and a binodal curve can thus be generated,with which liquid composition on each tray will be checked to see whether it is located inside the phase envelope or not (i.e.,phase splitting).If phase splitting takes

place on a tray,the VLLE computation mode will be adopted.

Otherwise,the VLE computation mode is used.Given a stage pressure and initial liquid composition of component i ,the vapor composition of component i leaving stage j and the tem-perature on stage j can both be computed based on a bubble-point temperature calculation.Once the stage temperature is known,the density of liquid mixture at bubble-point condition

Table 2.Constants of Bare Module Equipment Cost

equipment a type K 1K 2K 3B 1B 2F m F p

E101HEX multiple-pipe 2.76520.72820.0783 1.74 1.55 1.0 1.0E201HEX fixed-tube 4.3247?0.3030.1634 1.63 1.66 1.0 1.0E301HEX multiple-pipe 2.76520.72820.0783 1.74 1.55 1.0 1.0E102HEX kettle reboiler 4.4646?0.52770.3955 1.63 1.66 1.0 1.0E202HEX kettle reboiler 4.4646?0.52770.3955 1.63 1.66 1.0 1.0E302HEX kettle reboiler 4.4646?0.52770.3955 1.63 1.66 1.0 1.0V101DC horizontal vessel 3.55650.37760.0905 1.49 1.52 1.0 1.0C101COL vertical vessel 3.49740.44850.1074 2.25 1.82 1.0 1.0

tray sieve 2.99490.44650.3961----

C201COL vertical vessel 3.49740.44850.1074 2.25 1.82 1.0 1.0

tray sieve 2.99490.44650.3961----

C301COL vertical vessel 3.49740.44850.1074 2.25 1.82 1.0 1.0

tray sieve 2.99490.44650.3961----

a HEX:heat exchanger,DC:decanter,COL:column.

is estimated by modified Rackett equation(Reid et al.).14The liquid flow rate leaving stage j can then be calculated by Francis weir formula.The liquid and vapor mixture enthalpies at the stage temperature can be estimated by assuming ideal mixing. By energy balance on the stage,the flow rate of leaving vapor stream is calculated.Through the use of overall and component material balances,liquid holdup and liquid composition on the stage are integrated with fourth-order Runge?Kutta method with a time interval of0.01min.

A dynamic simulation run is done using the FORTRAN program,and the operating conditions of Scheme3at steady state are computed until the conditions of the system do not change with time.Also,the steady state results of all streams from the dynamic simulation in FORTRAN are about the same as the results from the design study using Aspen Plus. Physical properties and model parameters used in the dynamic simulation,such as critical properties,heat capacity coefficients,and Antoine constants,are taken from the Aspen Plus data bank and binary parameters of NRTL model are obtained from Wang et al.9Their values are the same as those used in the steady-state simulation for Scheme3using Aspen Plus.On the basis of80%flooding and a tray spacing of0.6m, Columns C101,C201,and C301are designed to have dia-meters of0.75,1.73,and0.92m,respectively.The weir height of each tray in every column is assumed to be0.05m.In addition,the column base volumes for the three columns are all sized for a liquid hold-up of10min.Pressure drops per tray for the three columns are set at0.007bar.A holdup time of 20min in the decanter is assumed to allow for two liquid phases to separate.

The basic control for Scheme3is investigated,and Figure6 depicts its plant-wide control structure.In order to maintain the column head pressures,three pressure control loops have been employed.The bottom product qualities of the three columns are maintained by adjusting reboiler duty,and a tray tem-perature control loop is used for each column.Open-loop sensitivity analysis for the tray temperature control loop in each column is carried out using Aspen Plus.Figure7shows open-loop sensitivity analysis results of the three columns by changing±0.5%of reboiler duties.Tray#7of the C101 column,tray#10of the C201column,and tray#8of the C301 column are chosen as three temperature control points,in view of the high sensitivity of tray temperature with respect to variations in reboiler duty.PID control is used in these three tray temperature control loops.The tuning constants are K c (proportional gain)=0.8,τI(integral time)=8.0,andτD (derivative time)=0.125for these three tray temperature control loops.A ratio control scheme is implemented to reject feed rate disturbance of the C201column,and the ratio of the organic reflux flow to the feed flow rate of the C201column is kept constant.The organic-phase liquid level of the decanter is controlled by manipulating entrainer makeup flow,and the aqueous-phase level of the decanter is controlled by manipula-ting aqueous flow rate.For the organic-phase level loop and the aqueous-phase level loop of the decanter,the P-only controllers are used and K c=10is used as in Arifin and Chien.6The bottom liquid levels of the three columns are controlled by manipulating the bottom product flow.The column bottom level control loops are considered to be ideal.As the overhead vapor of Column C201is condensed into liquid,the con-densate temperature is controlled at40°C by manipulating cooling water flow.The top temperature control loop of the C201column is also assumed to be ideal control.

Two types of disturbances are used to test the proposed plant-wide control structure,namely±20%changes in both fresh feed rate and fresh feed H2O composition.Figure8shows the dynamic responses for±20%step change in fresh feed rate. The plots in the first and second rows show responses of top

https://www.wendangku.net/doc/753037258.html,parison of Equipment Specifications and Costs of the Schemes

Scheme1Scheme2Scheme3 column1(C101)

total no.of trays14-14 diameter(m)0.83-0.75 height(m)10.80-10.8

C BM($)61,900-56,000 condenser(E101)

heat transfer area(m2)27.94--

C BM($)44,500--reboiler(E102)

heat transfer area(m2)21.98-19.43

C BM($)137,000-128,000 column2(C102)

total no.of trays343438 diameter(m) 1.63 1.61 1.73 height(m)22.8022.8025.2

C BM($)277,000272,000333,000 condenser(E201)

heat transfer area(m2)191.78186.66243.12

C BM($)142,000140,000158,000 reboiler(E202)

heat transfer area(m2)42.4342.0324.42

C BM($)209,000208,000145,000 column3(C103)

total no.of trays161616 diameter(m)0.90 1.250.91 height(m)12.0012.0012

C BM($)73,400109,20074,300 condenser(E301)

heat transfer area(m2)39.1072.59-

C BM($)61,700114,000-reboiler(E302)

heat transfer area(m2)26.2750.0928.95

C BM($)151,000238,000160,000 decanter

total flow rate(m3/min)0.4300.4180.552 diameter(m) 1.94 1.92 2.11 height(m) 5.82 5.76 6.33

C BM($)63,00061,60075,200 total capital cost($)1,220,5001,142,8001,054,300 annual steam cost($/year)2,464,5002,474,3001,843,000 annual cooling water cost($/year)63,20063,60047,000 annual utility cost($/year)2,527,7002,537,9001,890,000 https://www.wendangku.net/doc/753037258.html,parison of TACs and Cost Ratios of the Schemes

Scheme1Scheme2Scheme3 i min=0.1TAC($)2,649,7502,652,1801,995,430

cost ratio 1.000(base) 1.0010.753

i min=0.2TAC($)2,771,8002,766,4602,100,860

cost ratio 1.000(base)0.9980.758

i min=1/3TAC($)2,934,5332,918,8332,241,433

cost ratio 1.000(base)0.9950.764

vapor flow and bottom product flow of the three columns.Due to the ±20%change in the fresh feed rate,the vapor flow and the product flow,for the most part,increase or decrease over time before reaching their steady-state values.The water product flows (from C101and C301)and IPA product flow also increase (or decrease)and reach their new values in 80min with a +20%(or ?20%)change in fresh feed rate.For +20%change in fresh feed rate,the total water product flow

rate

Figure 6.Plant-wide control structure for Scheme

3.

Figure https://www.wendangku.net/doc/753037258.html,putational flowchart of equilibrium stage j .

changes from 834.20to 1001.03mol/min,and the IPA product flow rate changes from 832.49to 998.65mol/min.The plots in the third row show that the three tray temperatures are brought back to their desired set-points.The plots in the last row

show

Figure 7.Open-loop sensitivity analysis for the three columns in Scheme 3(a)C101,(b)C201,and (c)

C301.

Figure 8.Closed-loop responses with ±20%fresh feed rate changes (dashed lines,?20%;solid lines,+20%).

that the stabilized product compositions of the three columns are very near their purity specifications.The water product compositions are back to 99.9mol %,and the IPA product composition is also greater than 99.99989mol %.Figure 9,on the other hand,shows the dynamic responses for ±20%changes of water composition in the feed.The plots in the first row show that top vapor flows of the three columns all generally decrease over time with an increase in the water composition in the feed.However,as shown in the second-row plots,the response to a positive step change in the H 2O com-position is such that the product flow of Column C101generally increases over time before finally approaching a steady-state value,whereas the steady-state product flows of Columns C201and C301decrease with the same step change.The plots in the third row also show that the three tray temperatures can be effectively brought back to their set-point values.The plots in the last row show that the product compositions of the three columns are still in the ultrapure region.The water product compositions of the C101column are greater than 99.88mol %,and the water product compositions of the C301column are very near to 99.9mol %.The IPA composition in the product stream is greater than 99.99987mol %.5.CONCLUSIONS This article presents a study of designing a three-column heterogeneous azeotropic distillation configuration to separate IPA and water using cyclohexane as the entrainer,featuring energy saving and cost-effective.Due to low reflux ratios,a pre-concentrator column and an entrainer recovery column employed in the previous works are replaced by stripping columns.Our proposed scheme,essentially a three-column sequence using strip-ping columns in place of conventional distillation columns,saves more energy and is more cost-effective than other schemes in the literature.Regardless of the minimum acceptable return on investment being used (i.e.,1/3,0.2,or 0.1),the total annual cost of the proposed scheme is less than that of the other schemes.The basic control for the proposed scheme has also been looked into.A tray temperature control loop implemented in each of these three columns can be used to maintain the bottom product compositions.In addition,ratio control of the organic reflux flow to the feed flow of the IPA purification column is capable of rejecting feed rate disturbances.Further-more,if fresh feed rate or water composition in the feed is subject to a ±20%change,the simulation results of the closed-

loop system reveal that the control strategy for the proposed

scheme can yield good control performance.■AUTHOR INFORMATION Corresponding Author *Phone:+886-4-2359-0262.Fax:+886-4-2359-0009.E-mail:shcheng@https://www.wendangku.net/doc/753037258.html,.tw.■REFERENCES (1)Widagdo,S.;Seider,W.D.Azeotropic distillation.AIChE J.1996,

42,96.(2)Pham,H.N.;Doherty,M. F.Design and Synthesis of Heterogeneous Azeotropic Distillation ?III.Column Sequences.Chem.Eng.Sci.1990,45,1845.(3)Ryan,P.J.;Doherty,M.F.Design/Optimization of Ternary Heterogeneous Azeotropic Distillation Sequences.AIChE J.1989,35,1592.(4)Luyben,W.L.Control of a Multiunit Heterogeneous Azeotropic Distillation Process.AIChE J.2006,52,623.(5)Chien,I.L.;Zeng,K.L.;Chao,H.Y.Design and Control of a Complete Heterogeneous Azeotropic Distillation Column System.Ind.Eng.Chem.Res.

2004

,43,2160.

(6)Arifin,S.;Chien,https://www.wendangku.net/doc/753037258.html,bined Preconcentrator/Recovery Column Design for Isopropyl Alcohol Dehydration Process.Ind.Eng.Chem.Res.2007,46,

2535.

Figure 9.Closed-loop responses with ±20%feed H 2O composition changes (dashed lines,?20%;solid lines,+20%).

(7)Wu,Y.C.;Chien,I.L.Design and Control of Heterogeneous Azeotropic Column System for Separation of Pyridine and Water.Ind. Eng.Chem.Res.2009,48,10564.

(8)Wang,C.J.;Wong,D.S.H.;Chien,I.-L.;Shih,R.F.;Liu,W.T.; Tsai,C.S.Critical Reflux,Parametric Sensitivity,and Hysteresis in Azeotropic Distillation of Isopropyl Alcohol+Water+Cyclohexane.Ind. Eng.Chem.Res.1998,37,2835.

(9)Turton,R.;Bailie,R.C.;Whiting,W.B.;Shaeiwitz,J.A.Analysis, Synthesis,and Design of Chemical Processes,3rd ed.;Pearson Education: Boston,2009.

(10)Seider,W.D.;Seader,J.D.;Lewin,D.R.Product and Process Design Principles:Synthesis,Analysis,and Evaluation,2nd ed.;John Wiley and Sons:New York,2004.

(11)Olujic,Z.;Sun,L.;de Rijke,A.;Jansens,P.J.Conceptual Design of an Internally Heat Integrated Propylene-Propane Splitter.Energy 2006,31,3083.

(12)Franks,R.G.E.Modeling and Simulation in Chemical Engineering; John Wiley and Sons:Canada,1972.

(13)Rovaglio,M.;Doherty,M.F.Dynamics of Heterogeneous Azeotropic Distillation Columns.AIChE J.1990,36,39.

(14)Reid,R.C.;Prausnitz,J.M.;Poling,B.E.Properties of Gases and Liquids,4th ed.;McGraw-Hill Book Company:New York,1987.