Controlofreactive dividing wallcolumn for selective hydrogenation and separation of C3 stream☆

Xing Qian,Shengkun Jia,Yiqing Luo,Xigang Yuan*,Kuo-Tsong Yu

SchoolofChemicalEngineering and Technology,ChemicalEngineering Research Center,Collaborative Innovation Center ofChemicalScience and Engineering(Tianjin),State Key Laboratory of ChemicalEngineering,Tianjin University,Tianjin 300072,China

1.Introduction

Distillation is one of the mostwidely applied separation technologies in chemicalindustries.However,distillation is stilla capital-and energyintensive operation.Dividing wallcolumn(DWC)provides a promising alternative for distillation process intensi fication[1–4].Compared with conventionalcon figurations,the energy saving of DWCs can be up to 30%reported in literatures[5,6].Furthermore,DWCs can be applied to implement azeotropic,extractive and reactive distillations,which lead to azeotropic dividing wallcolumns(ADWC),extractive dividing wall columns(EDWC)and reactive dividing wallcolumns(RDWC)[7–20].

In our previous work[21],a novelprocess for C3 streamselective hydrogenation in ethylene plant was proposed.The proposed process combines the propylene catalytic reactive distillation for the hydrogenation with the column of depropanization to achieve a thermally coupled distillation column con figuration,which was thermal dynamically equivalent to reactive dividing wallcolumn(RDWC).The proposed RDWC con figuration was proved to be a promising and economically attractive process for selective hydrogenation and separation of C3 stream because ofits lower totalannualcost(TAC)compared to the corresponding conventionalreaction-and-distillation process.The reactive distillation improved the selectivity of the hydrogenation reaction with a lower investmentcost for the equipmentand the thermalcoupling of the distillation columns.

The dif ficulties in the controlof DWC are due to its inner complex structure and interactions among different control loops.Skogestad et al.[22–26]studied the controlofthree-product dividing wallcolumn and four-product dividing wallcolumn.Ling and Luyben[27]proposed the temperature difference controlfor dividing wallcolumn,and Huang etal.[28]proposed the double temperature difference controlscheme.Kiss et al.[29–31]investigated traditional PID control and advanced MPC(model predictive control).Chien et al.[12,17]investigated the design and controlofazeotropic dividing wallcolumns and extractive dividing wallcolumns.Xu etal.[11,32]studied the different controlstructures for extractive dividing wallcolumns.Controlofreactive distillation has also been addressed in the literatures[33–35].However,little research has been done on reactive dividing wallcolumns.The objective ofthis paper is to investigate multi-loop PID controlschemes ofreactive dividing wallcolumn for selective hydrogenation and separation of C3 stream.Four differentcontrolschemes are proposed in this paper for reactive dividing wallcolumn.The first two controlstructures are for the composition control,while the last two are for temperature control.For the RDWC studied in the present paper,there are 8 controlled variables(product butadiene composition,product propane composition,product propylene composition,base levelof the depropanization column,base levelof the propylene distillation column,re flux tank levelof the propylene distillation column,pressure of the depropanization column and pressure of the propylene distillation column)and 8 manipulated variables(the reboiler duty of the depropanization column,reboiler duty of the propylene distillation column,re flux flow rate of the propylene distillation column,bottom product flow rate of the depropanization column,bottom product flow rate of the propylene distillation column,top product flow rate of the propylene distillation column,compressor frequency,condenser duty of the propylene distillation column).Theoretically,there are 8!=40320 combinations of them forcontrolschemes.However,only a fewcontrolschemes have been known as effective[11,12,32].Instead of searching for the optimalones among allthe possible combinations,in the presentpaper4 controlschemesthathave been proved to be effective on the controlofpartially thermally coupled distillations are used[11,12,32],and their behaviors are investigated in controlling the RDWC for the selective hydrogenation and separation of C3 stream.The performances of the control structures are tested in terms of the product flow rate time pro files and the productcomposition time pro files with±20%persistent disturbances in either feed flow rate or feed compositions.

2.Steady-state Design

The proposed process is based on the traditional front-end deethanization process,which is shown in Fig.1.In such a process,the deethanizer's bottom stream,ofwhich the major compositions are simpli fied as three main components(Light:C3H6,Middle:C3H8,Heavy:C4H6)and MAPD as given in Table 1,is fed into the depropanizer.The top distillate stream of the depropanizer goes into a selective hydrogenation reactor,and the out flow of the reactor feeds into the propylene distillation column to get a propylene product of 99.60%in mole fraction.Before the simulation of the processes,a short-cut method based on the Winn–Underwood–Gilliland approach is used to provide initial designs for the distillation columns.Then,with the initialvalues given by the short-cut method,the traditionalreaction-and-distillation process as shown in Fig.1 is simulated rigorously employing Aspen Plus?software(property method:RK-Soave).

Fig.1.Conventionalreaction-and-distillation process.

Table 1 Feed data

Fig.2.Thermally coupled reactive distillation column.

The proposed thermally coupled reactive distillation process is shown in Fig.2(a)and it is thermodynamically equivalent to the RDWC con figuration as shown in Fig.2(b).

The selective hydrogenation reactions include hydrogenation of methylacetylene to propylene,hydrogenation ofpropadiene to propylene and hydrogenation ofpropylene to propane and can be represented as the following.

The kinetics come from the work done by Yu[36]and are shown as follows:

where rMAPDand rC3H6are the ratesofhydrogenation ofMAPDto propylene and of propylene to propane,respectively,and CMAPDand CH2are concentrations of MAPD and hydrogen,respectively.

The productpurity speci fications of the three products in this paper are 99.5%mole fraction ofbutadiene,93%mole fraction ofpropane and 99.6%mole fraction ofpropylene,respectively.The decision variables of the RDWC are shown in Table 2.

Table 2 Decision variables of the RDWC

Before the Aspen Plus steady state simulation is exported to Aspen Dynamics,the Tray Sizing section of the distillation column in Aspen Plus is used to determine the needed size of the column.The re flux drum and the sump ofeach column and the decanter are sized to provide 10 min holdup with 50%liquid level[37,38].

Fig.3.Illustration of the impurity composition controlstructure CS1.

Fig.4.Dynamic responses to feed flow rate disturbances for CS1.

3.Composition Control Structure

In this section,the dynamics and controlof the proposed RDWC process in Fig.2 willbe examined by using the pressure-driven simulation function provided in Aspen Dynamics.

3.1.Impurity composition controlstructure CS1

The impurity composition controlis a controlstructure,in which the controlled variable is the impurity composition in the corresponding product stream.The impurity composition in the bottom butadiene product stream in Fig.2 is the propane composition and those in the bottom propane product and the top propylene product streams are the propylene and the propane compositions,respectively.The overallcontrolstrategy of the impurity composition controlstructure CS1 is illustrated in Fig.3.PIcontrollers are used in this paper,except the P controllers for the controls of the levels.Relayfeedback testfunction provided in Aspen Dynamics is applied in regulating the parameters in allthe controllers,and Tyreus–Luyben settings are used to obtain the values for the gains and integraltimes.The tuning results for the impurity composition control structure CS1 are listed in Table 3.

As shown in Fig.3,the main schemes of the impurity composition control structure CS1 include composition controlof three impurity compositions in the corresponding product streams.In the bottom butadiene product stream of the depropanization column,the impurity propane composition(1.54‰mole fraction)is controlled by manipulating the corresponding reboiler duty(directacting).According to the pioneers'contributions[11,27,39],the feed forward control(QR/F controller,for example)is able to shorten the dynamic response time,reduce the maximum deviations and offer an immediate adjustment when the fresh feed flow rate changes.In this paper,the reboiler duty of the propylene distillation column is proportionated to the fresh feed flow rate(QR/F controller).The impurity propylene composition(6.78%mole fraction)in the bottom propane productstream of the propylene distillation column is controlled by manipulating the ratio of the corresponding reboiler duty to the fresh feed flow rate(direct acting).The liquid re flux of the propylene distillation column is proportionated to the fresh feed flow rate(R/F controller)as well.The impurity propane composition(2.85‰mole fraction)in the top propylene product stream of the propylene distillation column is controlled by manipulating the ratio of the liquid re flux to the fresh feed flow rate(direct acting).

In the investigation ofcontrolof the reactive distillation,the concentration of reactant on the feed stage should be controlled by manipulating the feed flow rate.The hydrogen composition on the feed stage of hydrogen is controlled by feed flow rate of hydrogen(reverse acting).

Fig.5.Dynamic responses to feed composition disturbances for CS1.

Fig.5(continued).

Chien etal.[12]assumed thatthe actualliquid splitratio may notbe the same as the designed ratio.Therefore,the liquid split ratio in the RDWC is set to be fixed other than using as a manipulated variable as suggested by Chien.In this paper,the liquid thermalcoupling stream flow rate is proportionated to the liquid out flow rate of stage 112,which means the liquid split ratio(LSR)is fixed.In this case,the liquid thermalcoupling stream flow rate is on cascade because it receives its set-pointsignalfrom the LSR controller.

In order to test the proposed control structure performance,two types of disturbances are used,namely,±20%changes in fresh feed flow rate and±20%changes in fresh feed composition.The changes for each of the three components:butadiene(H),propane(M)and propylene(L)are used in the feed composition change disturbances.Figs.4 and 5 give dynamic responses to disturbances in feed flow rate and feed composition,respectively.

It is shown in Fig.4 that the impurity composition controlstructure CS1 is able to handle the feed flow rate disturbance well.Although there are some oscillations in the propane product flow rate,the three product flow rates are able to move to a new steady state within 10 h.Consequently,the three product compositions are able to move to a new steady state very quickly,attributed to the use of the feed forward control(R/F controller and QR/F controller).

As the stage number of propylene distillation column is normally considerably large,the dynamic response of the propylene distillation column is usually very slow.The feed forward controllers are employed to shorten the dynamic response time,reduce the maximum deviations and offer an immediate adjustment.Feed forward control is faster than feedback control,as the former acts while the fresh feed changes,whereas the later has no action until deviations between the set point and the process variable occur.

Fig.5 sequentially shows results for the 20%changes of the compositions of light component propylene,middle component propane and heavy component butadiene.As for the three kinds of feed composition disturbances,the flow rates of the butadiene product,the propane product and the propylene product move to a new steady state within 5 h,10 h and 15 h,respectively.The controlled impurity compositions are returned to the set points as shown in the middle columns of Figs.4 and 5.However,some product compositions are not totally returned to the original values as shown in the right columns of Figs.4 and 5.This is because there is usually more than one impurity component in a product stream.However,the deviations are less than 1.5%mole fraction in the dynamic responses of the impurity composition controlstructure CS1.This may be due to two important manipulated variables,the reboiler duties of the corresponding columns,are retained in the RDWC con figuration compared with the traditional two-column process.

Fig.5(continued).

3.2.Productcomposition controlstructure CS2

In the multicomponent distillation,there is more than one impurity in a product stream.Thus,even though the controlled impurities are returned to the set points,the product compositions may not return to the originalpoints.To overcome this disadvantage,a product composition controlstructure CS2 is proposed.In CS2,the compositions ofbutadiene,propane and propylene are controlled in the butadiene,propane and propylene productstreams,respectively.The overallcontrolstrategy of the product composition controlstructure CS2 is shown in Fig.6.The tuning results for CS2 are listed in Table 4.In CS1,the set point of CCB2 is 6.78%,while the set points of CCB1 and CCD2 are 1.54‰and 2.85‰,respectively.This means the impurity fraction ofB2(6.78%)is 44 times higher than that of B1 and 24 times higher than that of D2.During the relay-feedback test,the amplitude is 5%of the controller.Therefore,the variations of the tuning parameters of CCB2 are relatively larger than those for the other loops comparing Table 4 with Table 3.

Fig.6.Illustration of the product composition controlstructure CS2.

Table 4 Controller tuning parameters of CS2

Fig.7.Dynamic responses to feed flow rate disturbances for CS2.

The CS2 structure differs from CS1 in that the three productcompositions other than the three main impurity compositions are controlled.In the bottombutadiene productstream of the depropanization column,the product butadiene composition is controlled by manipulating the corresponding reboiler duty(reverse acting).The reboiler duty of the propylene distillation column is proportionated to the fresh feed flow rate(QR/F controller).The productpropane composition in the bottom propane product stream of the propylene distillation column is controlled by manipulating the ratio of the corresponding reboiler duty to the fresh feed flow rate(reverse acting).The liquid re flux of the propylene distillation column is proportionated to the fresh feed flow rate(R/F controller)as well.The productpropylene composition in the top propylene productstreamof the propylene distillation column is controlled by manipulating the ratio of the liquid re flux to the fresh feed flow rate(reverse acting).

Figs.7 and 8 give dynamic responses of both CS1 and CS2 to disturbances in feed flow rate and feed compositions,respectively.Comparing with CS1,CS2 is able to let the product compositions return exactly to the set points.There are small maximum deviations and a few numbers of oscillations in the product flow rates and the product purities.However,the new steady state values of the product streams are similar in the feed flow rate disturbances of CS1 and CS2.The oscillations are mainly due to the multiple impurities,and may also be attributed to the improper values of the controllers'tuning parameters,even though they have been regulated for many different tests.

Fig.8.Dynamic responses to feed composition disturbances for CS2.

Fig.8(continued).

The product purity speci fications of the three products in this paper are 99.5%mole fraction ofbutadiene,93%mole fraction ofpropane and 99.6%mole fraction ofpropylene,respectively.As the product purity speci fication of propane is not as high as propylene and butadiene,the verticalcoordinates of dynamic response pro file of propane concentration is different from that of propylene and butadiene.

4.Temperature Control Structure

4.1.Three point temperature controlstructure CS3

Comparing with composition control,temperature controlbrings in shorter dead time and costs less in both capitalinvestmentand maintenance.There are several methods to select temperature control locations.Sensitivity criterion is used in this paper.Fig.9 shows the temperature pro files of the depropanization column and the propylene distillation column and the open-loop gains between tray temperature and the corresponding reboiler duty.The curves show that the 28th stage of the depropanization column,and the 101th stage and the 132th stage in the propylene distillation column are sensitive to changes in heat input.

The overall control strategy of the three point temperature control structure CS3 is illustrated in Fig.10.The tuning results for the three point temperature control structure CS3 are listed in Table 5.

CS3 is differentfrom CS2 in thatitcontrols the three temperatures of the sensitive trays other than controlling the three product compositions.The temperature of the 28th stage of the depropanization column is controlled by manipulating the corresponding reboiler duty(reverse acting).The temperature of the 132th stage of the propylene distillation column and that of the 101th stage of the propylene distillation column are controlled by manipulating respectively the ratio of the corresponding reboiler duty to the fresh feed flow rate(reverse acting)and the ratio of the liquid re flux to the fresh feed flow rate(reverse acting).

Figs.11 and 12 give the dynamic responses to disturbances in the feed flow rate and feed composition,respectively.Besides being economically attractive,the three point temperature control structure CS3 shows better performance in response speed than the impurity composition control structure CS1 and the product composition controlstructure CS2.CS3 substantially reduces the number ofoscillations and moves to a new steady state signi ficantly quickly.It takes less than 10 h for product stream flow rates to move to the new steady states.Notably,CS3 can suppress the peak maximum deviations of the butadiene composition when fresh feed flow rate changes.

Fig.9.Open-loop sensitivity analyses.

4.2.Two pointtemperature controlstructure CS4

The controlof DWC is more dif ficult than the controlof a conventionaltwo-column separation process,as there are more interactions among controlloops.Therefore,a single temperature controlloop of the propylene distillation column is employed in CS4 to validate the one temperature control for the propylene distillation column.The 132th stage is more sensitive than the 101th stage in the propylene distillation column as shown in Fig.9,and thus the temperature of the 132th stage is controlled.

In CS4 the TC3 controller that controls the temperature of the 101th stage in the propylene distillation column in CS3 is deleted.The overall controlstrategy of the two pointtemperature controlstructure CS4 is illustrated in Fig.13.The tuning results for the two point temperature controlstructure CS4 are listed in Table 6.

Figs.14 and 15 give dynamic responses of both CS3 and CS4 to disturbances in feed flow rate and feed composition,respectively.The dynamic responses of the product streams flow rates are similar in CS3 and CS4.Comparing with the three point temperature control CS3,the two point temperature control CS4 presents a little smaller maximum deviations of the butadiene composition while a little higher deviations of the propylene composition when fresh feed flow rate disturbances occur.As for the feed composition disturbances,the performance of CS3 is better than that of CS4 when the heavy component feed composition changes.The propylene product composition is controlled better using CS3 when light component feed composition disturbs,while it is controlled better using CS4 when middle component feed composition disturbs.However,the differences of the final product composition in CS3 and CS4 are less than 1%.Therefore,both CS3 and CS4 are able to provide effective control performance for RDWC.

Fig.10.Illustration of the three point temperature controlstructure CS3.

Table 5 Controller tuning parameters of CS3

5.Remarks and Conclusions

For the reactive dividing wallcolumn(RDWC)developed in our previous work,four controlstructures are proposed and tested.CS1 and CS2 are composition control structures,while CS3 and CS4 are temperature control structures.All of them are able to provide effective control of RDWC.This may be due to that two important manipulated variables,the reboiler duties of the corresponding columns,are retained in the DWC con figuration like the traditional two-column process[12].Manipulating the reboiler duty is a direct way of changing vapor rates,which can be achieved fairly rapidly.This can avoid manipulating liquid rate change(via manipulation of the re flux,for example)to cause signi ficant delay due to the hydraulic lags[37].

Tables 7 to 10 give the comparisons among the control schemes for differentdisturbances.Table 7 shows thatfor the feed flow rate disturbances,the maximum deviations can be reduced by temperature control,but the settling time increases compared to the composition controlofCS1.Tables 8,9 and 10 show thatforthe feed composition disturbances,the maximum deviations are allless than 1.3%with the settling times of within 20 h.The composition controlschemes perform better than the temperature control ones in terms of the maximum deviation.

As the controlled variables are the product compositions,the product compositions return exactly to the set points in CS2.CS3 and CS4 are able to signi ficantly reduce the number ofoscillations and shorten the settle time,comparing with CS2.The stripping section in the propylene distillation column is more sensitive than the rectifying section.Thus,the sensitive tray in the stripping section is preferably used ifonly one temperature is controlled in the propylene distillation column.When heavy component feed composition changes frequently,CS3 could be a better choice than CS4.

Fig.11.Dynamic responses to feed flow rate disturbances for CS3.

The feed forward controllers are employed in the four control strategies to shorten the dynamic response time,reduce the maximum deviations and offer an immediate adjustment.As the RDWC combines reactive distillation and dividing wall columns,two variables(the concentration ofreactant on the feed stage and the liquid splitratio)should be considered during the controlstructure design.In a reactive distillation process,the concentration ofreactant on the feed stage should be controlled by manipulating the feed flow rate.Besides,a three-product DWC is different from conventionaltwocolumn process in that it has a liquid split ratio and/or a vapor split ratio as extra freedoms.As the actual liquid split ratio may not be the same as the designed ratio,it should not be used as a manipulated variable according to Chien et al.[12]for the consideration of practicalindustrialapplication.Compared with composition control,temperature controloffers shorter drag time,less equipment investment and less maintenance costs.Therefore,CS3 and CS4 should be more practicalthan CS1 and CS2 in industrialapplications.

It should be noted that for the temperature controlschemes,the temperature on the sensitive tray may not correlate well with the product composition for multicomponent mixture,especially when the concentration for the impurity components is high,like in the case for product of propane,which is as high as 7%in the present paper.This is the main reason why the propane in the product composition cannot return to the original level(with an error of 2%)as shown in Figs.14 and 15.Besides,the steady state deviation may also be attributed to the limitation by employing PI(and P)controllers in the RDWC,which is complex in terms of control.Even though conventional PID control strategies have been well adopted in the previous studies of DWC control,the advanced control strategies such as model predictive control(MPC)is worthy of investigations in the future.

Acknowledgments

The authors appreciate the valuable suggestions from Dr.Ming Xia and the assistances from the staffat State Key Laboratory of Chemical Engineering(Tianjin University).

Fig.12.Dynamic responses to feed composition disturbances for CS3.

Fig.12(continued).

Table 6 Controller tuning parameters of CS4

Table 7 Controlperformance comparison offeed flow rate disturbances

Table 8 Controlperformance comparison offeed composition disturbances of L

Table 9 Controlperformance comparison of feed composition disturbances of M

Table 10 Controlperformance comparison of feed composition disturbances of H

Fig.12(continued).

Fig.13.Illustration of the two point temperature controlstructure CS4.

Fig.14.Dynamic responses to feed flow rate disturbances for CS4.

Fig.15.Dynamic responses to feed composition disturbances for CS4.

Fig.15(continued).

[1]I.Dejanovic,L.Matijasevic,Z.Olujic,Dividing wallcolumn-a breakthrough towards sustainable distilling,Chem.Eng.Process.49(2010)559–580.

[2]O.Yildirim,A.A.Kiss,E.Y.Kenig,Dividing wallcolumns in chemicalprocess industry:A review on current activities,Sep.Purif.Technol.80(2011)403–417.

[3]D.Staak,T.Gruetzner,B.Schwegler,D.Roederer,Dividing wallcolumn for industrial multi purpose use,Chem.Eng.Process.75(2014)48–57.

[4]K.T.Chu,L.Cadoret,C.C.Yu,J.D.Ward,A new shortcut design method and economic analysis of divided wallcolumns,Ind.Eng.Chem.Res.50(2011)9221–9235.

[5]M.Emtir,E.Rev,Z.Fonyo,Rigorous simulation of energy integrated and thermally coupled distillation schemes for ternary mixture,Appl.Therm.Eng.21(2001)1299–1317.

[6]S.Hernandez,S.Pereira-Pech,A.Jimenez,V.Rico-Ramirez,Energy ef ficiency of an indirect thermally coupled distillation sequence,Can.J.Chem.Eng.81(2003)1087–1091.

[7]A.A.Kiss,J.J.Pragt,C.J.G.van Strien,Reactive dividing-wallcolumn show to getmore with less resources?Chem.Eng.Commun.196(2009)1366–1374.

[8]L.Y.Sun,X.W.Chang,C.X.Qi,Q.S.Li,Implementation of ethanol dehydration using dividing-wall heterogeneous azeotropic distillation column,Sep.Sci.Technol.46(2011)1365–1375.

[9]A.A.Kiss,R.M.Ignat,Innovative single step bioethanol dehydration in an extractive dividing-wallcolumn,Sep.Purif.Technol.98(2012)290–297.

[10]A.A.Kiss,D.Suszwalak,Enhanced bioethanol dehydration by extractive and azeotropic distillation in dividing-wallcolumns,Sep.Purif.Technol.86(2012)70–78.[11]M.Xia,B.R.Yu,Q.Y.Wang,H.P.Jiao,C.J.Xu,Design and control of extractive dividing-wall column for separating methylal–methanol mixture,Ind.Eng.Chem.Res.51(2012)16016–16033.

[12]Y.C.Wu,H.Y.Lee,H.P.Huang,I.L.Chien,Energy-saving dividing-wallcolumn design and controlfor heterogeneous azeotropic distillation systems,Ind.Eng.Chem.Res.53(2014)1537–1552.

[13]S.-J.Wang,S.-H.Cheng,P.-H.Chiu,K.Huang,Design and control of a thermally coupled reactive distillation process synthesizing diethyl carbonate,Ind.Eng.Chem.Res.53(2014)5982–5995.

[14]R.M.Ignat,A.A.Kiss,Optimaldesign,dynamics and controlof a reactive DWC for biodieselproduction,Chem.Eng.Res.Des.91(2013)1760–1767.

[15]R.Delgado-Delgado,S.Hernandez,F.O.Barroso-Munoz,J.G.Segovia-Hernandez,A.J.Castro-Montoya,From simulation studies to experimentaltests in a reactive dividing walldistillation column,Chem.Eng.Res.Des.90(2012)855–862.

[16]Y.Tavan,S.Shahhosseini,S.H.Hosseini,Design and simulation of ethane recovery process in an extractive dividing wallcolumn,J.Clean.Prod.72(2014)222–229.

[17]Y.C.Wu,P.H.C.Hsu,I.L.Chien,Criticalassessment of the energy-saving potential of an extractive dividing-wallcolumn,Ind.Eng.Chem.Res.52(2013)5384–5399.

[18]H.-Y.Lee,I.K.Lai,H.-P.Huang,I.L.Chien,Design and controlof thermally coupled reactive distillation for the production of isopropyl acetate,Ind.Eng.Chem.Res.51(2012)11753–11763.

[19]Y.-C.Wu,H.-Y.Lee,C.-H.Lee,H.-P.Huang,I.L.Chien,Design and control of thermally-coupled reactive distillation system for esteri fication of an alcohol mixture containing n-amylalcohol and n-hexanol,Ind.Eng.Chem.Res.52(2013)17184–17197.

[20]H.Zhang,Q.Ye,J.W.Qin,H.Xu,N.Li,Design and controlofextractive dividing-wall column for separating ethyl acetate-isopropylalcohol mixture,Ind.Eng.Chem.Res.53(2014)1189–1205.

[21]X.Qian,S.Jia,Y.Luo,X.Yuan,K.-T.Yu,Selective hydrogenation and separation ofc3 stream by thermally coupled reactive distillation,Chem.Eng.Res.Des.99(2015)176–184.

[22]I.J.Halvorsen,S.Skogestad,Optimal operation of petlyuk distillation:Steady-state behavior,J.Process Control9(1999)407–424.

[23]D.Dwivedi,J.P.Strandberg,I.J.Halvorsen,H.A.Preisig,S.Skogestad,Active vapor splitcontrolfor dividing-wallcolumns,Ind.Eng.Chem.Res.51(2012)15176–15183.

[24]D.Dwivedi,J.P.Strandberg,I.J.Halvorsen,S.Skogestad,Steady state and dynamic operation of four-product dividing-wall(kaibel)columns:Experimentalveri fication,Ind.Eng.Chem.Res.51(2012)15696–15709.

[25]D.Dwivedi,I.J.Halvorsen,S.Skogestad,Controlstructure selection for three-product petlyuk(dividing-wall)column,Chem.Eng.Process.64(2013)57–67.

[26]D.Dwivedi,I.J.Halvorsen,S.Skogestad,Controlstructure selection for four-product petlyuk column,Chem.Eng.Process.67(2013)49–59.

[27]H.Ling,W.L.Luyben,Temperature controlof the btx divided-wallcolumn,Ind.Eng.Chem.Res.49(2010)189–203.

[28]N.Wu,K.J.Huang,S.J.Luan,Operation of dividing-wall distillation columns.2.A double temperature difference control scheme,Ind.Eng.Chem.Res.52(2013)5365–5383.

[29]A.A.Kiss,C.S.Bildea,A control perspective on process intensi fication in dividingwallcolumns,Chem.Eng.Process.50(2011)281–292.

[30]A.A.Kiss,R.R.Rewagad,Energy ef ficient control of a BTX dividing-wall column,Comput.Chem.Eng.35(2011)2896–2904.

[31]R.R.Rewagad,A.A.Kiss,Dynamic optimization of a dividing-wall column using modelpredictive control,Chem.Eng.Sci.68(2012)132–142.

[32]M.Xia,Y.P.Xin,J.W.Luo,W.S.Li,L.Shi,Y.Min,C.J.Xu,Temperature controlfor extractive dividing-wall column with an adjustable vapor split:Methylal/methanol azeotrope separation,Ind.Eng.Chem.Res.52(2013)17996–18013.

[33]M.Al-Arfaj,W.L.Luyben,Comparison of alternative controlstructures for an ideal two-product reactive distillation column,Ind.Eng.Chem.Res.39(2000)3298–3307.

[34]M.A.Al-Arfaj,W.L.Luyben,Design and control of an ole fin metathesis reactive distillation column,Chem.Eng.Sci.57(2002)715–733.

[35]M.A.Al-Arfaj,W.L.Luyben,Plantwide control for tame production using reactive distillation,AIChE J.50(2004)1462–1473.

[36]Z.Q.Yu,B.L.Gao,J.Y.Zhang,Kinetic study ofc3 cut liquid-phase selective hydrogenation over supported metalcatalyst,Ind.Catal.(China)10(2002)11–16.

[37]W.L.Luyben,Distillation design and control using aspen simulation,John Wiley&Sons,New York,2013.

[38]W.L.Luyben,I.L.Chien,Design and control of distillation systems for separating azeotropes,John Wiley&Sons,New York,2011.

[39]L.Sun,Q.Wang,L.Li,J.Zhai,Y.Liu,Design and control of extractive dividing wall column for separating benzene/cyclohexane mixtures,Ind.Eng.Chem.Res.53(2014)8120–8131.