Yixin Ma,Peizhe Cui,Yongkun Wang,Zhaoyou Zhu,Yinglong Wang,*,Jun Gao,*

1 College of Chemical and Environmental Engineering,Shandong University of Science and Technology,Qingdao 266590,China

2 College of Chemical Engineering,Qingdao University of Science and Technology,Qingdao 266042,China

ABSTRACT Extractive distillation is an effective method for separating azeotropic or close boiling point mixtures by adding a third component.Various technologies for performing the extractive distillation process have been explored to protect the environment and save resources.This paper focuses on the improvement of these advanced technologies in recent years.Extractive distillation is retrieved and analyzed from the view of phase equilibrium,selection of solvent in extractive distillation,process design,energy conservation,and dynamic control.The quantitative structure-property relationship used in extractive distillation is discussed,and the future development of extractive distillation is proposed to determine how the solvent affects the relative volatility of the separated mixture.In the steady state design,the relationship between the curvature of the residue curve and parameters of the optimal steady state is also highlighted as another field worthy of further study to simplify the distillation process.

Keywords:Thermodynamic Quantitative Structure-property relationship Solvent selection Process design Energy conservation Dynamic control☆ Supported by the National Natural Science Foundation of China (21676152).

1.Introduction

Exhaustion of energy,which is a resource problem,is a critical issue that should be considered in all fields.In the chemical industry,energy consumption is needed to ensure the production of enterprises.In the late 1990s,hard-working scientists promoted the vigorous development of the distillation industry in China[1].High energy consumption and serious pollution also appeared.Thus,the appropriate use of energy has attracted the attention of global researchers.Separation is an important process in the chemical industry,particularly in energy consumption during distillation,which accounts for 10%-15% [2]of the total energy consumption in chemical plants.Despite its high energy consumption,distillation is the preferred method for separating a mixture into pure components.These mixtures are typically azeotropic.The existence of azeotropes increases the difficulty of separating mixtures,and simple distillation cannot meet the separation requirements.The present study is a review of the development of the extractive distillation of azeotropes via dynamic control.The quantitative structural property relationship was proposed for solvent selection and to investigate the effect of a solvent on extractive distillation to improve synthesis.

Special distillations,such as azeotropic distillation [3],extractive distillation [4],pressure-swing distillation [5,6]and reactive distillation[7-9],are commonly used method to separate azeotropic mixtures.Azeotropic distillation consumes a great deal of energy because azeotrope mixtures consist of an entrainer,and a component needs to evaporate from the top of the column.Pressure-swing distillation is often used to separate azeotropic mixtures with pressure-sensitive properties.Reactive distillation is a complex process due to the separation of products by distillation while performing the reaction.Extractive distillation has been widely studied and applied in industry.To achieve extractive distillation,a sufficient change of relative volatility must occur within the separated components when the solvent is added.The effect of the solvent on azeotropic mixtures has been studied in theory and experimentally [10-12].

Extractive distillation was first proposed in the 1930s [13].Extractive distillation can be divided into three types according to the operating mode:continuous extractive distillation,batch extractive distillation and semi-continuous extractive distillation[4].Continuous extractive distillation utilizes the primary method according to the literature,as shown in Table 1.Continuous extractive distillation is a flowsheet containing two columns;the first column is called the extractive column,and the second column is called the solvent recovery column.The mixture to be separated is fed into the extractive column in the lower section,while the solvent is fed into the recovery column in the upper section.Two products are obtained at the top of the two columns,and the solvent is recycledonto the extractive column from the bottom of the solvent recovery column.Compared with continuous extractive distillation,batch extractive distillation[41-43]is used to separate mixtures with frequently changing compositions into several products.The mixture to be separated and the solvent are simultaneously charged in the reboiler,and products are obtained at the top of the column.To reach maximum productivity,it is crucial to determine the appropriate ratio of feed and solvent.Semi-continuous extractive distillation is another operating mode.The difference between batch extractive distillation and semi-continuous extractive distillation is the feed location of the solvent.The solvent is added in the boiler and at the middle and top [44,45]of the column in a semi-continuous mode.The two operating strategies in semi-continuous extractive distillation in terms of the charge of the initial feed are full charge and fractional charge[46].Full charge means that the feed is charged into the reboiler to reach the maximum capacity of the column at a fixed condenser vapor load;the column is then flooded without careful control of the reflux ratio and solvent feed rate.Fractional charge is a suitable operation mode to separate many azeotropic mixtures using a fixed column.

Table 1 List of the study of the azeotrope with extractive distillation separation

There is another classification method that is based on the type of solvent.Extractive distillation can be divided into four types:extractive distillation with a solid salt [47-50],extractive distillation with a liquid solvent [30,37,38,51,52],extractive distillation with a combination of a liquid solvent and a solid salt [51]and extractive distillation with an ionic liquid [52-56].A suitable solvent should decrease the consumption of solvent needed to meet the requirement of the product purity and be easy to operate.Extractive distillation with a liquid solvent is a traditional method compared with other types of solvent.The combination of a liquid solvent and solid salt used in extractive distillation has superior characteristics,such as ease of operation and high separation ability;this combined solvent is also suitable for both polar systems and non-polar systems.Furter [57]noted that although the use of salt as a solvent has the advantage of a high separation ability,salts corrode equipment,which limits their industrial application.Ionic liquids have increasingly been used as solvents in extractive distillation and have attracted the attention of researchers[54,58,59].Because ionic liquids have high costs and research is mainly limited to the laboratory,this review primarily focuses on liquid solvents without ionic liquids.

Although extractive distillation has some advantages over other special distillation strategies,there are disadvantages that limit its application in some mixtures.First,a third component needs to be introduced during extractive distillation,resulting in the appearance of impurities in the products.Second,the solvent used in extractive distillation can affect energy consumption and capital investment due to the different properties of the solvent used,such as its selectivity,capacity and boiling point.Third,the use of three components increases the difficulty of controlling the process[60].Fourth,it is difficult to select an effective solvent to separate mixtures to meet the separation requirement [61].

Selection of an appropriate solvent is a crucial factor and is the basis of extractive distillation.As Laroche reported in 1991 and Lelkes(1998)and Rodriguez-donis(2009)showed,the general feasibility criteria are based on the topology of the residue curve map and the location of the univolatility curve,which are both driven by the properties of the solvent and its interactions with the mixture to be separated.Studies on solvent selection are based on a computer-aided molecular design.A molecular simulation method is important to enhance the understanding of interaction between entrainer and separated mixtures.Li et al.[62]simulated binary mixtures of esters with alcohols or organic acids based on the TraPPE-UA force and investigated DFT as molecular simulation method to optimize geometries of ILs and molecules in the next year[63].Yin et al.[64]studied the selection of co-solvents to separate a benzene-cyclohexane system.Van Dyk and Nieuwoudt[65]proposed a method that combined a genetic algorithm and the UNIFAC model for computer-aided molecular design for extractive distillation.Recent studies have examined the characteristics of mixtures based on a model established according to the quantitative structure-property relationship (QSPR) [66-69].QSPR has been used in extractive distillation to predict the suitability of solvents.Kang et al.[70,71]established a quantitative structure relative volatility relationship based on a multiple linear regression model and artificial neural network model to predict the suitability of ethylbenzene as a solvent in a p-xylene system.

The theoretical design of extractive distillation can be improved using QSPR.The molecular information includes many aspects,such as topological descriptors,dipoles,jurs descriptors of the strain energy,and so on.The ability of the solvent to change the relative volatility of the separated mixture and the relatedness between the molecular structure of the solvent and the relative volatility are important to study.The nature of the solvent can be determined using some descriptors.The ternary phase diagram is a common tool that is used to determine the feasibility of extractive distillation,but there are no studies focusing on the inner link between the curvature change of the residue curve in the ternary phase diagram and the parameters of the separation process.The difficulty in performing extractive distillation is quantitatively analyzing the nature of the solvent when the relative volatility changes and describing the relationship between the curvature change of the residue curve and the parameters of the separating process using QSPR.

Some reviews and books have studied extractive distillation for separating azeotropes in terms of theoretical calculation of the column and application of extractive distillation[72,73].For example,Luyben and Chien[74]contributed to the knowledge of extraction by analyzing a phase diagram,the operating mode,and dynamic control.Lei et al.[4]reviewed extractive distillation and classified extractive distillation according to several aspects,such as the number of columns and types of solvents used.This analysis provided a detailed illustration of a specific case.Gerbaud V et al.[75]studied extractive distillation using a mathematical calculation of the equipment phase diagram to theoretically account for the establishment of extractive distillation in a published book.Mahdi et al.[76]reviewed the type of solvent used in extractive distillation and analyzed the superiority of different solvents.The aim of the present study is to review each step of extractive distillation in recently published papers and the future development of extractive distillation according to thermodynamic analysis,solvent selection,process design,and dynamic control.Fig.1 shows the classification and domain of extractive distillation.

2.Phase Equilibrium

The azeotropic phenomenon is when the composition of a separated mixture in the vapor and liquid phases is the same.Relative volatility is typically used to describe the azeotropic phenomenon and is a criterion to assess the feasibility of extractive distillation.The expression of relative volatility is defined as Eq.(1)

where x is the molar fraction in the liquid phase,y is the molar fraction in the vapor phase,i,j represent the components in the separated mixture,γ is the activity coefficient,and P0is the vapor pressure of the pure component.At the azeotropic point,αijis equal to 1,and this equation is unable to meet the requirement of separation of a mixture via single distillation [77].

Vapor liquid equilibrium (VLE) data are used to judge whether the property package used in a simulation is correct.Before the simulation and modeling of extractive distillation,the first step is to obtain VLE data.Although thermodynamic models,such as UNIQUAC,Wilson and NRTL,can be used to regress VLE data,the results typically show some deviation from experimental data.Thus,thermodynamic regression is necessary to determine the binary interaction parameters and reduce the gap between the predicted data and experimental data.The root-mean-square deviation(RMSD) is usually used as a criterion for judging the gap between experimental data and predicted data by thermodynamic regression.When the RMSD value is generally considered to be less than 1%,the thermodynamic model is considered to be of high accuracy and the binary interaction parameters obtained by the regression of the model are also considered to be reliable.COSMO-RS is a model that is based on quantum computation of the molecular structure,and compared with other models,this model requires fewer parameters and,as a promising model,is used to screen solvents in extractive distillation [78].The published literature studying the VLE of separated mixtures with solvents and the thermodynamic models are listed in Table 1.According to the separation scale of extractive distillation,the binary azeotrope mixtures,which can be separated with extractive distillation,are divided into three types in terms of the T-x y diagrams,and Fig.2 shows the three types.Examples of extractive distillation focusing on the separation of azeotropes are listed in Table 1.

Fig.1.The classification and study content of extractive distillation.

Fig.2.T-xy diagrams of different types of binary separated by extractive distillation:(a) minimum boiling azeotrope;(b) maximum boiling azeotrope;(c) low relative volatility mixture.

Azeotropes have been analyzed in various studies,and analysis of the azeotrope type can lead to the development of extractive distillation.The minimum boiling azeotropes and maximum boiling azeotropes are shown in Fig.2(a) and (b),respectively;these azeotropes are often used in industry and have been broadly studied using extractive distillation.Separation of mixtures with similar boiling relies on extractive distillation.The remarkable characteristics of a mixture is that the vapor line and liquid line in the T-x diagram become closer as the component composition changes;the mixture cannot form an azeotrope,but the relative volatility is very close to 1.The relevant T-x diagrams are shown in Fig.2(c).The mixtures shown in Fig.2(c) can theoretically be separated via single distillation,but a large column is required to achieve separation,so extractive distillation is the first choice for separating the type of mixture described above.

3.Selection of Solvent in Extractive Distillation

The solvent is the third component added to the separated mixture and is used to break the azeotropic phenomenon and change the relative volatility of the mixture to be separated.The selected solvent must possess certain qualities,such as it being stable,noncorrosive,inexpensive,selective,having a low boiling point and toxicity,and so on,and these qualities have been previously discussed in the literature[14].It is difficult to select a suitable solvent from a large number of candidates.The initial method of solvent selection is to examine the primary properties of the solvent and separated mixture,such as the distribution coefficients,selectivities,solvent losses,and so on.Research on solvent selection has attracted the attention of many scholars,and Computer-Aided-Molecular-Design(CAMD)and QSPR are widely used for solvent selection[61].

3.1.CAMD

The CAMD method was proposed in the 1980s because the introduction of CAMD provided a new method for solvent selection,and this method has slowly expanded to other processes,such as gas absorption,liquid-liquid extraction,extractive distillation and so on[79,80].CAMD is used to generate molecular structures with certain rules,and the newly formed molecule is tested and its molecular properties are matched,which is essential in extractive distillation.Kossack et al.[81]selected a solvent for separating acetone and methanol by using CAMD;after primary selection,the solvent was selected based on its selectivity at an infinite dilution via the rectification body method,and then,a rigorous mixed-integer optimization of the entire extractive flowsheet for the remaining entrainer candidates was performed to fix the remaining design degrees of freedom and determine the best entrainer.The application of CAMD is based on the contribution of UNIFAC,and all of the work on CAMD has been conducted either by using software or mathematic programming using special solution techniques,such as the support grid methodology [82]or branch-and-bound algorithm [83],to improve the accuracy of the CAMD method.A single solvent is commonly used in extractive distillation,and many systems have been separated with the use of a solvent containing a single component[24,84,85],but consideringthe cost,solvents containing two or more components are now considered for use in extractive distillation[86-88].In batch extractive distillation,researchers have tried toanalyze the feasibility of using light and intermediate solvents from the point of thermodynamics,and the type of the solvents used in extractive distillation are listed in Table 2.

Table 2 List of different type of binary mixture separated with different type of solvent

3.2.QSPR

Dong et al.[100]conducted a multiscale study on ionic liquids and found that the scale prompted the organic solvent,improving the development of extractive distillation.The change in the relative volatility of a binary mixture typically depends on the solvent added to the mixture;using the separation of isobutyl acetate/isobutyl alcohol as an example [101],the components N,Ndimethylformamide,1-hexanol,and butyl propionate were selected as alternative solvents.According to the x-y diagram drawn from the VLE data,the change of the relative volatility is different based on the different solvents used;butyl propionate and 1-hexanol increased the relative volatility of isobutyl acetate and isobutyl alcohol,but the effect of butyl propionate was superior to that of the 1-hexanol.By contrast,N,N-dimethylformamide decreased the relative volatility of isobutyl acetate and isobutyl alcohol.What are the differences among the alternative solvents?Further research on the component structure and component property is needed to provide a clear explanation.QSPR is used to study and determine the quantitative relationship between the activity of a compound and its molecular structure or its physical and chemical properties based on the method of mathematical statistics.The outcome of the method is the generation of a model that connects the material properties and material structure;the established model is used to predict the molecular properties of a new component or a nonexistent component with specific molecular groups.The process of model building typically occurs through four steps:data collection,calculation and selection of molecular descriptors,model building,and validation of the model.The crucial step in building the model is to select molecular descriptors related to the material properties from various alternative molecular descriptors.The method of descriptor selection occurs via three types[102]:classical methods[103,104],artificial intelligence-based methods[105-107]and miscellaneous methods[108-112].Although the methods described above can provide a fast strategy for molecular descriptor selection,every method has disadvantages;for example,the disadvantage of the forward selection method included in the classical methods is that several descriptors will lead to a good prediction,while the use of single descriptors leads to poor prediction;in this case,none of the molecular descriptors will be selected.The existing modeling method has two parts:a linear modeling method,such as multiple linear regression,and a nonlinear modeling method,such as artificial neural network.Validation of the QSPR model is typically via internal and external cross validation.

Early in the development of QSPR technology,QSPR was used in drug research [113,114],and with further development,QSPR has been used to study the material properties of single components[115,116]and binary component [117-119].Recently,QSPR has been used in extractive distillation to predict the relative volatility of a separated mixture with a solvent added [70,71].The models study a specific mixture and provide a relatively accurate prediction of the relative volatility.Nevertheless,no papers have been published on general factors that can explain the changes in the relative volatility applied to all systems.Fig.3 shows the contents of the current study and the future development of extractive distillation.QSPR might be able to explain the nature of the effect of the solvent on the mixture to be separated and to select a solvent for extractive distillation.Asuitablesolventcanpromotetheseparationofthemixture via extractive distillation.If a relationship between the molecular descriptors and solvent is built,then this model will provide a solid foundation for developing the extractive distillation process.

Fig.3.The current and future study domain of extractive distillation.

4.Process Design of Extractive Distillation

4.1.Pressure selection

Pressure is the primary factor that should be confirmed before beginning extractive distillation.Increasing the relative volatility in the extractivedistillationcolumncan facilitateseparationand reduce TAC.Many scholars have conducted extensive research on pressure selection for the extractive distillation column[20,120,121].Luyben[122]increased the pressure of the two columns in extractive distillation to separate acetone and chloroform,and the aim of that paper was to guarantee the use of cooling water in the condenser.In some cases,the boiling point of the component distilled in the top of the column is lower,so inexpensive cooling water cannot be used as the heat sink.The pressure of the column should be increased to ensure that the temperature of the reflux-drum is approximately 323 K.To reduce the consumption of steam,it is necessary to reduce the pressure of the column.However,there are some limitations that restrict pressure selection in some systems.If the components to be separated are thermally unstable,there will be a maximum temperature in the column based on the purity of the product.The pressure of the column can be fixed if the pressure drop is known;then,the pressure of the condenser can also be obtained by considering the purity of the distillate,and the temperature can then also be fixed.If the temperature is lower than the temperature needed to use water as a cold sink,then expensive refrigeration is required.Another limitation focuses on the light component;if the pressure needed to meet the requirement of 323 K in the reflux-drum is near the critical pressure of the separated component and the density difference of the vapor and fluid phase is small,then hydraulic problems restrict the high pressure in the column;therefore,refrigeration or vapor recompression is essential for these systems.

4.2.Determination of the distillation sequence for continuous extractive distillation

The separating configuration has a fatal effect on the subsequent economics and dynamic control.The type of solvent plays an essential role in confirming the separating configuration.The ternary phase diagram is a useful tool for analyzing phase behavior [123-125].Kiva et al.[123]used a ternary phase diagram to classify different ternary mixtures,and using ternary phase diagram can improve extractive distillation by determining solvent selection and the feasibility of the separation process.Rodriguez-Donis et al.[92]noted that the 0.0-1,1.0-1a,1.0-1b and 1.0-2 classes can meet all of the situations that occur in extractive distillation:the 0.0-1 class represents a low relative volatility mixture,the 1.0-1a class explains the separation of a minimum boiling temperature(maximum boiling temperature)azeotropic mixture with a heavy(light)solvent,the 1.0-2 class can explain the separation of a maximum boiling temperature(minimum boiling temperature)azeotropic mixture with a heavy(light)solvent,and the 1.0-1b classcan explaineither the maximum orminimum boiling temperature of an azeotropic mixture with an intermediate solvent.Fien and Liu[126]stated that the split strategy can be divided into direct and indirect methods.The direct method indicates that the low-boilingcomponent is separated at the top of the first column.By contrast,the indirect method indicates that the high-boiling component is initially separated at the bottom of the first column.Many studies have examined the common situationof the 0.0-1 class[61,94].Separation of a minimum boiling point azeotropic mixture with a light solvent has rarely been studied and is not considered in this review.In Fig.4,separation of a minimum boiling temperature azeotropic mixture with an intermediate boiling point solvent is selected as an example,and the solid line in the diagram represents the material balance line,which graphically shows the separation process.Two separating configurations can be used to separate the mixture.The first split configuration is called direct split;the configuration uses two columns that are called the extractive and solvent recovery columns.Feed stream F1 is fed into the extractive distillation column,and the recycled solvent from the top of the solvent recovery column is fed into the extractive distillation at a higher location.Light-boiling component A is distilled at the top of the extractive column,the mixture at the bottom of the column with trace A is fed into the solvent recovery column,the solvent reaches the top of the solvent recovery column,and pure B is obtained at the bottom of the recovery column.Another configuration used is that described above,except that component B is obtained at the bottom of the first column,component A is obtained at the top of the second column,and the solvent is recycled from the bottom of the second column to the first column.The final decision of which separation configuration to use depends on the minimal TAC.The residue curveis an effective tool for identifying the separation tendency,analyzing the feasibility of the technology and predicting the separation efficiency.TAC is widely used in the distillation process[127,128].Timoshenko et al.[129]designed an extractive distillation process by analyzing the feature of the residue curve.Wahnschafft et al.[130]confirmed the separation and products purities by analyzing the direction of the residue curve.In a previous study [131]of triple-column pressure-swing distillation,we showed that a change of the residue curve at different pressures can lead to a change of the separation configuration and parameters.In extractive distillation,introduction of the solvent can obviously change the direction of the residuecurve.Thus,if the relationship between the curvature of the residue curve and separation parameters is determined,it can decrease the time required and increase the efficiency of the distillation process,promoting the development of extractive distillation.

5.Process Intensification

For extractive distillation,optimization of the flowsheet and hybridization of extractive distillation with other technologies are the two main methods used to reduce energy consumption and capital costs.Fig.5 shows a schematic diagram of the optimization and improvement of extractive distillation.

5.1.Process optimization

Fig.4.Ternary phase diagram of extractive distillation and the split method.

Fig.5.Optimization and extractive dividing-wall column of extractive distillation.

Optimization involves the identification of the optimal values of the final variables,and the final result is to maximize or minimize the objective function while meeting the specified constraints.Under normal conditions,the steady state of extractive distillation is simulated using process simulators,such as Aspen Plus,PRO//II,and ChemCAD,among others.The equilibrium model and ratebased model are the two most commonly used models for simulation;the relevant publications are listed in Table 1.The variables,such as the feed location,number of theoretical plates,quantity of the solvent,and reflux ratio,should be optimized to obtain the minimal total annual cost(TAC).Douglas[132]provided a description of the parameters and an equation for the capital and energy costs in TAC.To reduce the optimization time and provide a clear picture of optimization,optimization methods were proposed in some previously published papers to adjust the variables and obtain the minimum energy consumption and equipment costs[133-135].The sequential iterative optimization method is widely used in continuous extractive distillation to obtain the optimal parameters,and some published papers have accepted the method and provided the optimization steps [23,136].The simulated annealing algorithm (SAA) and colony optimization method can be used to optimize the distillation process [137].Introducing these methods to extractive distillation can also lead to optimization of the process.In the optimized process,the quantity of solvent and the number of theoretical plates are the outer iterative loop variables,enabling the optimization of the number of stages.The locations of the feed and the reflux ratio can be selected as inner iterative loop variables.The minimal TAC can be obtained after a continually iterative calculation.In the simulation of recovering of organic solvents from aqueous waste mixtures,the genetic algorithm is used to optimize the process[135].Kossack et al.used the SR-MINLP technique to further optimize extractive distillation after primary optimization.This technique was efficient for optimizing extractive distillation according to both continuous and discrete variables [81].The main optimization factors of the batch extractive distillation concentrate on the number of separated components,and the influence of some parameters,such as the operation policy and quantity of the solvent,should be accounted for in the final TAC;optimization of batch extractive distillation,as has been previously studied [98,133].

5.2.Improvement of ED

Optimization is used to reduce the TAC by adjusting the parameters of the column settings.To obtain a more effective reduction of the TAC,itis necessary to improve extractivedistillationand simplify the separation process.Some methods have been implied in extractive distillation,which can be clarified into different aspects.Some techniques rely on the reuse of heat,such as thermal coupling,while some methods focus on improving the separating method,such as by using an extractive dividing wall,and Table 3 lists some published papers that studied various extractive distillation methods to reduce the cost of the separating process.Due to the energy and capital advantages of using a dividing wall column,this technique began to be used in industrial application a long time ago,and the first dividing wall column was used in 1985[157];due to the energy saving in extractive distillation [86,158,159],extractive dividing-wall columns(EDWCs)were further studied[99,141,148,150].

In recent studies,EDWCs have been shown to have two different schemes:single column and double columns.The single column scheme has the right and left sections divided by a wall;mixture A/B to be separated is fed into the column,and solvent S is fed into a higher location of the column than the mixture.Solvent S interacts with the component to be separated and reduces the vapor pressure of the heavy component B;hence,the volatility of B decreases and the light component A is obtained at the top of the left part.Then,the heavy component B is distilled from the top of the right part and pure solvent S is recycled back onto the column[141,160,161].The obvious characteristic of EDWCs is that the two columns used for conventional continuous extractive distillation are integrated into one column and the left and right part operate in the same section and use a single reboiler,which avoids the use of unnecessary columns and reboilers and directly reduces the capital costs and consumption of steam.The two columns scheme consists of a main column and a rectifying column;the rectifying column does not use a reboiler.Xia et al.[148]studied the separation of methylal and methanol and compared the TAC of the optimum EDWC scheme with conventional continuous extractive distillation.EDWC saved 11.6% of energy,which indicated that EDWCs are superior to other columns.

6.Dynamic Control

Extractive distillation is a process that contains various input and output variables [162].Some characteristics involving distillation control make it difficult to meet the requirements of industry.The interaction between the input and output variables makes extractive distillation difficult to control,and many researchers have studied the dynamic control of extractive distillation to generate an optimal control strategy for different systems using dynamic simulation software[163,164].The selection of the pairing should satisfy two criteria.First,the interaction between manipulated variables and control variables should be large enough.Second,the pairing should be able to reduce the time constants of the controllers.

In view of the point discussed above,some methods have been developed based on arithmetical operation to select the pairing.Bristol[165]proposed the relative gain array for multivariable control to solve the theoretical and practical deficiencies of matrix representation for the design of the system.Hovd and Skogestad[166]extended the relative gain array method to open-loop unstable plants to choose pairings,but the method neglected realistic dynamic information and the interaction loop pairing was often incorrect.The dynamic relative gain array method was proposed to solve the problem mentioned above by Witcher and McAvoy[167];the method used the transfer function model to calculate the relative gain array instead of the steady-state gain matrix,but this method is tedious and difficult to apply.An effective relative gain array method was presented by Xiong et al.[168],and the method combined the two means.The dynamic information of the distillation process can be reflected by the effective relative gain array method using a simple and effective method.

Fig.6 shows a case that involved the control scheme of extractive distillation,tuning condition and disturbance response curve,which are criteria that can be used to assess the control effect.The conventional control method in extractive distillation is the proportionalintegral-derivative(PID)control method.A case describing the optimal control scheme and controllers is shown in Fig.6.In the basic control scheme,the flow rate controllers,proportion controllers of the solvent to the fresh feed,liquid level controllers,pressure controllers and temperature controllers are involved;the basic control scheme can achieve optimal control under disturbances in most cases and is widely used in industrial production.For some separation systems,more complex controllers,such as a compositiontemperature cascade control,the reboiler heat duty/fresh feed flow rate (QR/F) scheme,and the fixed reflux ratio (RR) scheme,are needed to improve the control strategy.

The tuning method is used to obtain the tuning parameters,which have an important effect on the controllers;the result of tuning within the composition and temperature is crucial for setting the control strategy.The Ziegler-Nichols [169]and Tyreus-Luyben[170]methods are widely used for tuning,but they are limited for calculating the controller,so the tuning method of PID controllers has attracted the attention of researchers [171-174].PID controllers based on various models,such as the genetic algorithm[175]and the so-called n-th order lag process[176],were proposed to achieve the optimal control scheme.

A key point that determines the control result is the final robust performance of the process when facing a feed disturbance,such as feed flow and feed composition.Robust performance is typicallydemonstrated by the purity of the product,temperature of the sensitive stage,distillation and bottom flow rate.Fig.6 shows a common trace of a robust performance.To determine whether a trace is robust,the standards used are traces of the temperature and composition,which show a slight deviation and take little to time return to the steady state.

The PID controller is a widely used method in extractive distillation and is more suitable for industrial processes because of its simplicity and low requirements.However,the conventional PID controller has some disadvantages on some occasions,and some PID controller improvements have been made.Most chemical processes,such as multi-input variable and multi-output variable processes,are multivariable,and the interaction between them is nonnegligible.The decoupling PID controller can solve the problem of multi-input and multi-output [162].The fuzzy-PID controller method has also been studied because of its fast execution compared to the conventional control method and because it can be used to establish a real-time control system [177].

Dynamic control in extractive distillation is complex due to the addition of a third component.The solvent flow rate/fresh feed flow rate(S/F)scheme is used to ensure that there is sufficient solvent in extractive distillation,but the optimal control scheme is difficult to obtain.The third component can occasionally be the key point in process control.Luyben [178]investigated the effect of three solvents on the controllability of the separation of acetone and methanol using extractive distillation,and the results showed that the three solvents all achieved stabilization,but the product purity was poor with chlorobenzene because the solvent was facing a feed disturbance.To solve the problem of the lack of stable dynamic control with a specific solvent,Wang et al.[179]presented a method to improve the amount of solvent used and sacrifice the TAC to obtain robust control.This method is also discussed in the separation of ethanol and water,and the same conclusion was drawn[180].

7.Conclusions

Fig.6.Flowsheet of the typical control system for continuous extractive distillation.

Six aspects of extractive distillation were reviewed herein:thermodynamic analysis,QSPR,the solvent used,process design,process intensification,and dynamic control.Extractive distillation is a common technology for separating azeotropes,which are nonpressure-sensitive or close-to-boiling-point mixtures.Finding a suitable solvent is the key aspect of extractive distillation.The role of QSPR in extractive distillation is becoming more important.A published paper reported the QSPR model,which relates the molecular structure and melting point of compounds,from which we can confirm that the information of microscopic molecular structure can be used to represent and predict some properties of matters.Some established models can predict the ternary azeotropic temperature with highly accuracy;in other words,it is feasible to identify the properties of a mixture with molecular descriptors.So,the phenomenon that it can be easier to separate an azeotropic mixture with the addition of solvent can be explained from the microcosmic aspect;in addition,QSPR is a common technology to explain the quantitative relationship between a compound structure and its various physicochemical properties;therefore,QSPR can be a useful method to select a solvent in extractive distillation.Due to its high energy consumption and high investment,optimization of extractive distillation and improvements based on traditional extractive distillation should be further explored.Process intensification can help to reduce energy consumption and capital costs while protecting the environment.Dynamic control could promote the development of industry in the direction of intelligence and security.We propose that the most important aspects of extractive distillation to explore are the natural relationship between the solvent and separated mixture and the factors that lead to changes in relative volatility.

Nomenclature

QR/F reboiler heat duty/fresh feed flow rate

S/F solvent flow rate/fresh feed flow rate

αijrelative volatility

γi,γjactivity coefficient