A New Catalog of the Most Promising Separation Sequences for

Ind. Eng. Chem. Res. 2001, 40, 5795-5809

5795

A New Catalog of the Most Promising Separation Sequences for Homogeneous Azeotropic Mixtures I. Systems without Boundary Crossing Zainuddin A. Manan and Rene´ Ban ˜ ares-Alca´ ntara* Universitat Rovira I Virgili, Departament d’Enginyeria Quimica, ETSEQ, Carretera de Salou s/n, 43006 Tarragona, Catalonia, Spain

This paper presents a systematic procedure for generating a catalog of the most promising separation sequences for homogeneous azeotropic mixtures without boundary crossing. First, the azeotropic mixtures are classified into classes of azeotropic systems. This is followed by the generation of feasible distillation sequences for each class of azeotropic system. The sequences generated are screened using new heuristics developed here by means of geometric reasoning, as well as some well-established sequencing heuristics, to give the most promising distillation sequence for each class of azeotropic system. The procedure results in a new catalog of classes of azeotropic systems with their corresponding most promising separation sequences. 1. Introduction Decision making during the synthesis and design of distillation sequences for ideal and slightly nonideal mixtures is influenced by variables such as the feed composition, volatility of the components to be separated, required product purities, column reflux ratios, and number of stages. The presence of azeotropes introduce new variables such as the choice of entrainer; the type of azeotropes formed; the shape and locations of distillation boundaries; and, in the case of heterogeneous systems, the properties of a liquid-liquid equilibrium region. This set of variables, which affects the number and connectivity of the separation units, the flow rate of entrainer, and the column design, vary to such an extent from one mixture to another as to make it impossible to formulate and solve a general numerical model for the optimal synthesis of separation sequences. For ternary mixtures, we can represent most of the azeotrope-related features on ternary diagrams known as residue curve maps (RCMs). Schreinemakers1 first defined RCMs as ternary diagrams displaying traces of the liquid composition remaining in a single-stage batch still as a result of vaporization over a period of time. The liquid traces are better known as residue curves. We limit our discussions to the synthesis of separation sequences for azeotropic mixtures with a maximum of three components. Unless stated otherwise, we use the following conventions for any given mixture: (1) Every ternary diagram follows the convention for the location of pure components with respect to their boiling points shown in Figure 1, that is, low-boiling component at the top left corner, medium-boiling component at the bottom left, and high-boiling component at the bottom right. (2) The fractional composition of any given component ranges between 0 and 1. Reading any two compositions from the horizontal and vertical edges of the right triangular diagram gives enough information to calculate the composition of the third component. (3) Components are listed in the text according to their boiling

Figure 1. Typical residue curve map for homogeneous azeotropic mixtures.

points: light-medium-high-boiling component, with the boiling characteristics of the entrainer shown in parentheses. For example, ethanol-water-ethylene glycol (high-boiling entrainer). (4) The components that originally exist as a binary azeotropic mixture that is to be separated into two essentially pure components are referred to as either “binary feed constituents” or “binary constituents”. (5) “Azeotropic mixture” refers to an individual mixture, whereas “azeotropic system” refers to a collection of similar azeotropic mixtures. In the next section, we discuss some of the important work related to this study and the issues that have led to the development of the catalog of the most promising sequences described in this work. 2. Related Work

* Author to whom correspondence should be addressed. E-mail: [email protected].

To date, research related to the optimization of distillation sequences for azeotropic mixtures is very

10.1021/ie990215l CCC: $20.00 © 2001 American Chemical Society Published on Web 11/01/2001

5796

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001

limited when compared to the extensive work done for zeotropic mixtures. Knight and Doherty,2 Ryan and Doherty,3 Pham and Doherty,4-6 Stichlmair et al.,7,8 Laroche et al.,9-11 Wahnschafft et al.,12 and Rooks et al.13 provide contributions in the area of the synthesis of separation sequences for azeotropic mixtures. Knight and Doherty formulated a systematic procedure involving the numerical optimization of the entrainer-to-feed ratio for a homogeneous azeotropic mixture. They generalized that the procedure is applicable to nonideal and azeotropic separations even though it was applied only for the extractive distillation of the ethanol-water-ethylene glycol (high-boiling entrainer) mixture. The procedure was later used with slight modifications in ref 13 during the heat integration of the distillation columns for separating homogeneous azeotropic mixtures. However, those case studies were also restricted to extractive distillation processes. Laroche et al., who used the optimum entrainer-to-feed ratio as one of the criteria for entrainer selection, extended the use of the entrainer optimization procedure to other homogeneous azeotropic mixtures that do not introduce new azeotropes.11 Ryan and Doherty3 proposed a useful technique involving numerical optimization of the design and operating parameters for several alternative separation sequences for the ethanol-water-benzene (high-boiling entrainer) heterogeneous azeotropic mixture. Stichlmair and Herguijuela8 focused their discussions on the basic concepts concerning the separation regions, entrainer selection, and resulting separation sequences for azeotropic mixtures without residue curve boundary crossing. Laroche et al.9-11 provided some useful guidelines for selecting the entrainers for the separation of homogeneous azeotropic mixtures that do not introduce new azeotropes and reported how some unusual behaviors shown by these mixtures might affect entrainer selection. In another study, they also stated the necessary conditions for separability of homogeneous azeotropic mixtures that use one, two, or three columns and discussed the synthesis of separation sequences for these mixtures. Their studies discussed separation synthesis in the context of entrainer selection but did not provide guidelines for screening between the alternative separation sequences. Among others, Wahnschafft et al.12 assumed that azeotropic mixtures generally result in a large number of separation sequences. They proposed an automated approach for the synthesis of complex separation processes including heterogeneous azeotropic distillation. Rooks et al.13 used an equationbased geometric methods and rapid generation of process alternatives for nonideal and azeotropic distillation system. The method is useful for mixtures with more than four components but critically relies on vaporliquid equilibrium models. From the studies associated with the synthesis of distillation sequences for azeotropic mixtures mentioned, a number of issues remain unresolved: (1) Other than by conducting rigorous economic assessments, no guidelines are available for eliminating inferior column sequences from alternative separation trains of homogeneous and heterogeneous azeotropic mixtures. As a result, designers often have to evaluate the economics of every conceivable sequence to find the most promising one. Such a task can prove very timeconsuming, for instance, when a number of different

entrainers are being evaluated and a few separation options exist for each type of entrainer. (2) We also question the need to automate the synthesis of some complex separation processes as proposed by Wahnschafft et al.12 We also show that it is possible to generate every desirable separation sequence for some common classes of homogeneous azeotropic mixtures and to identify the most promising sequence mainly through geometric reasoning. This enables us to produce a catalog of classes of homogeneous mixtures with their corresponding most promising separation sequences. Thus, it could be expensive and inefficient to search for the entire solution space and to automate the synthesis of sequences for homogeneous azeotropic systems for which only a fixed and limited number of alternative sequences exist. (3) Our analysis reveals that the procedures for optimization of distillation sequences for azeotropic mixtures have so far been applied to homogeneous systems that do not introduce new azeotropes and to only the class of heterogeneous mixtures that is similar to the ethanol-water-benzene system. From our literature survey, we found no evidence to support or disclaim that results from the optimization studies performed by Knight and Doherty,2 as well as by Ryan and Doherty,3 are applicable to other types of homogeneous and heterogeneous mixtures. In view of the peculiarities of azeotropic mixtures, it is considered important to demonstrate whether these procedures can indeed be extended to more complex homogeneous and other heterogeneous systems. Here, we provide an explanation based on the RCM geometry of how these optimization results can be applied to more complex homogeneous systems. Our proposed catalog of the most promising distillation sequences for azeotropic systems with and without boundary crossing attempts to resolve the issues mentioned. These predefined sequences, which could be incorporated in an intelligent system, can help a designer tremendously. In the next section, we distinguish between two classes of synthesis problems that emerge from two distinct classes of homogeneous azeotropic mixtures. The proposed procedure is described in sections 4-6. It involves three key steps: (1) azeotropic mixture classification (section 4), in which the mixtures are first grouped into classes according to the boiling point (temperature) of the entrainer relative to that of the pure components and azeotropes (e.g., a system with a high-boiling entrainer that does not introduce new azeotropes that is used to break a minimum-boiling binary azeotrope); (2) generation of every feasible distillation sequence for the azeotropic systems formed, where, for the first time, distillation sequences arising from feed compositions from every distillation region are systematically considered (section 5); and (3) sequence screening, in which the alternative sequences are then screened using new heuristics that we have developed through geometric reasoning and some wellestablished heuristics for screening separation sequences (section 6). 3. Classification of the Synthesis Problem The notion of residue curve boundaries as limits of the range of feasible separation has been noted by Van Dongen and Doherty,14 as well as by Doherty and Caldarola.15 Van Dongen and Doherty rationalized that

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001 5797

the composition profile of a distillation column, which approximates the residue curves when the column operates at infinite reflux, cannot cross the residue curve boundaries. Doherty and Caldarola on the other hand used as a practical working assumption that the mass balance line linking the feed, distillate, and bottoms compositions of a continuous distillation column must be located in the same distillation region no matter what the column operating conditions are. Laroche et al.10 described that the notion of residue curve boundaries has meaning only for columns operating at infinite reflux and that finite reflux columns are almost always able to cross these boundaries, contrary to what had been reported by Doherty and co-workers. They added that, even when operated at infinite reflux, the mass balance line of a column can still cross a residue curve boundary that is sufficiently curved. We begin our analysis by distinguishing between two types of separation synthesis problems that emerge from two distinct classes of homogeneous azeotropic mixtures that are formed as a result of mixing a binary azeotropeforming mixture with a desired entrainer: (1) A type I synthesis problem deals with a homogeneous azeotropic mixture with at least one of the binary feed constituents being neither the origin nor the terminus of a residue curve. This implies that at least one of the binary feed constituents is medium-boiling (a saddle) in relation to the entrainer and the azeotrope(s) (see any system in Figures 2 and 3). A homogeneous azeotropic mixture in this class either has no distillation boundary or has a distillation boundary that cannot be crossed by the composition profile of a continuous distillation column. Visually, such a distillation boundary appears straight even though in the mathematical sense all distillation boundaries are curved to a certain extent. Manan17 proposed a boundary curvature test to determine the extent of distillation boundary curvature. When a distillation boundary exists, the binary feed constituents are located in the same distillation region so that separation is feasible. A type I problem is referred to as synthesis of separation sequences for homogeneous azeotropic systems without boundary crossing. (2) A type II synthesis problem deals with a homogeneous azeotropic mixture whose binary feed constituents are either the origins or the termini of the residue curves. This implies that the boiling points of the binary feed constituents are either lower or higher than the boiling points of the azeotrope(s) and the entrainer. A homogeneous azeotropic mixture in this class exists with the binary feed constituents located in two separate distillation regions. It has a distillation boundary that is curved and, therefore, can be crossed by the composition profile of a continuous distillation column. A type II problem is referred to as synthesis of separation sequences for homogeneous azeotropic systems with boundary crossing. A detailed treatment of this problem is presented in ref 17. The distinction between the type I and type II problems depends heavily on the accuracy of the method used to determine the distillation boundary curvature. Here, the thermodynamics would be crucial if we were to apply the simulation approach like the one proposed by Manan.17 Fortunately, alternative methods for rapid tracing of RCMs and accurate identification

of interaction parameters are now available to reliably distinguish between the two types of synthesis problems. It is well-known that those homogeneous azeotropic systems with or without boundary crossing exhibit distinctly different behaviors with respect to synthesis of their separation sequences. We expect a similar situation to arise during screening and optimization of the separation sequences for the two types of systems. Optimization of systems without boundary crossing has been studied by Knight and Doherty2 and Knapp and Doherty.14 In contrast, almost nothing is available on the screening of sequences for homogeneous azeotropic systems with and without boundary crossing. This issue is addressed in detail here and in Manan.17 The next section describes the first step of our proposed sequencing approach. 4. Classification of Homogeneous Azeotropic Mixtures without Boundary Crossing The ternary mixtures are grouped into classes of homogeneous azeotropic systems on the basis of the type of entrainer used to break their azeotropes. Our survey of the work done by Doherty and co-workers and Stichlmair and co-workers7,18 indicates that the two most widely used criteria for the classification purposes are as follows: (1) The boiling point of the entrainer relative to those of the binary feed constituents and the azeotropes, e.g., high-, low-, or medium-boiling. To guarantee a feasible separation sequence, the entrainer cannot form a homogeneous azeotropic mixture with a straight distillation boundary that divides the binary feed constituents into two separate regions.16 This restriction, however, does not apply to entrainer-forming liquidphase heterogeneous regions and those that result in homogeneous mixtures with curved boundaries between the components to be separated. Liquid-phase immiscibility makes it possible to move from one region to another by using liquid-liquid phase separation. On the other hand, it is possible for a material balance line of a desired separation to cross a curved boundary by using appropriate combinations of mixing and splitting. (2) The type of azeotrope introduced by the entrainer (either maximum- or minimum-boiling), if any. The systems that are more common commercially are the ones with minimum-boiling binary azeotropes and an entrainer that does not add a new azeotrope (i.e., systems WOBC-A1 and WOBC-A2 in Table 1, with system WOBC-A2 being the most common). The less common are those with maximum-boiling binary azeotropes (i.e., systems WOBC-B1, WOBC-B2, and WOBCB3 in Table 1, with system WOBC-B3 being the least common). We have so far come across very few examples dealing with the separation of binary mixtures with maximum-boiling azeotropes. Laroche et al., who investigated over 400 homogeneous mixtures, report that binary mixtures with minimum-boiling azeotropes are far more common than those with maximum-boiling azeotropes.11 This explains why most homogeneous azeotropic distillations separate minimum-boiling azeotropes. Binary mixtures with maximum-boiling azeotrope are nonetheless listed in Table 1 and Figure 3 as type WOBC-B mixtures, bearing in mind that they are of less importance commercially.

5798


Figure 2. RCMs and the possible separation sequences for breaking minimum-boiling azeotropes for classes of ternary homogeneous mixtures without boundary crossing

Entrainers for mixtures WOBC-A1, WOBC-A2, WOBCB1, and WOBC-B2 in Table 1 result in feasible separation sequences because they neither add new azeotropes nor form distillation boundaries. The others add either a maximum- or a minimum-boiling azeotrope and form a straight distillation boundary, which divides each of the RCMs into two distillation regions. Notwithstanding the distillation boundary, separation is still feasible because the binary feed constituents lie in the same distillation region. Figures 2 and 3 present pictorial representations of the ternary systems listed in Table 1, i.e., of the WOBC-A and WOBC-B systems, respectively. These systems have been analyzed by Foucher and Doherty,18

as well as by Stichlmair et al.19 during the selection of feasible entrainers. In section 5, we generate the alternative separation sequences for the homogeneous azeotropic systems mentioned. The sequences are later screened using some new heuristics developed here by means of a geometric approach (section 6) to produce a catalog pairing a typical homogeneous azeotropic system RCM with the most promising separation sequence (see Figures 11 and 12). Such a catalog provides a quick an essential guide for selecting the best separation sequence for a given homogeneous mixture during the early stages of flowsheet design. Figure 2 presents a pictorial version of the WOBC-A from Table 1 obtained from Foucher and Doherty.18 We have


Figure 3. RCMs and the possible separation sequences for breaking maximum-boiling azeotropes for classes of ternary homogeneous mixtures without boundary crossing. Table 1. Classes of Ternary Homogeneous Azeotropic Systems without Boundary Crossing system

entrainer for breaking the azeotrope18

WOBC-A. Binary mixture with a minimum-boiling azeotrope

WOBC-A1. Medium boiler forming no new azeotrope WOBC-A2. High boiler forming no new azeotrope WOBC-A3. Low or medium boiler forming a maximum-boiling azeotrope with either the medium- or low-boiling binary feed constituent, respectively WOBC-B1. Low boiler forming no new azeotrope WOBC-B2. Medium boiler forming no new azeotrope WOBC-B3. High or medium boiler forming a minimum-boiling azeotrope with either the medium- or high-boiling binary feed constituent, respectively

WOBC-B. Binary mixture with a maximum-boiling azeotrope

extended this pictorial representation to WOBC-B systems (see Figure 3). 5. Alternative Column Sequences The synthesis of separation sequences for ideal mixtures can result in a combinatorially large number of feasible distillation sequences that can be narrowed down by using some well-established heuristics, such

as the rank-ordered heuristics developed by Nadgir and Liu.20 In contrast, the presence of distillation boundaries and azeotropes prevent many splits that would have been feasible otherwise, drastically limiting them to a few feasible ones, particularly in the case of homogeneous mixtures. The separation of a binary homogeneous azeotrope into two essentially pure constituent components is

5800


Figure 4. Typical binary feed compositions for synthesis of separation sequences for homogeneous azeotropic mixtures.

normally performed in at least two distillation units consisting of the extractive (or homogeneous azeotropic distillation) and entrainer recovery columns.9,20 Laroche et al. also demonstrated the possibility of separating certain binary azeotropic mixtures into two pure products in only one column when both of the desired product compositions lie on the same residue curve.10 This case, however, is too restrictive to the particular type of mixture mentioned. Moreover, because the entrainer goes through the azeotropic column in only a single pass, the entrainer requirement and the potential waste problems could prove expensive and detrimental in the long run, even though the entrainer requirement is said to be relatively small. Admittedly, the onecolumn scheme might be useful in the event that all of the entrainer is continuously available within the same process. In this study, we limit our discussion to the general case of azeotropic separation that includes a separate column for entrainer recovery, i.e., a minimum of two distillation columns. We hereby regard the twounit sequence as the minimum unit sequence. For the separation of a ternary azeotropic mixture, our observations show that the number and types of alternative distillation sequences that can be generated depend on the class of the azeotropic mixture and the location of the binary feed. Because of the peculiarities of azeotropic mixtures, grouping such mixtures into classes of homogeneous azeotropic systems as described in section 4 simplifies the sequencing task and enables a wide range of homogeneous azeotropic mixtures to be considered. In addition, how the location of the binary feed influences the separation structures has never been fully investigated. A synthesis study can begin with any one of the three “typical binary feeds” shown in Figure 4. Laroche et al.10 listed some of the alternative sequences for homogeneous azeotropic systems whose entrainers do not introduce new azeotropes. In the rest of this section, we generate every feasible distillationbased separation option for ternary homogeneous azeotropic systems without boundary crossing, including those that introduces new azeotropes [see, for example,

cases WOBC-A3(i) and WOBC-A3(ii) in Figure 2]. The separation sequencing procedure for azeotropic mixtures described by Malone21 is implemented. As will be shown, each class results in a fixed number of feasible separation sequences. At this stage, we exclude heat integration. 5.1. Alternative Sequences for Case WOBC-A2, Table 1 (High-Boiling Entrainer). The ethanolwater-ethylene glycol (high-boiling entrainer) mixture is used as an example. Figure 5 shows that the separation of such a system requires at least two columns. Except in the case of an azeotropic feed, any given binary feed should result in three possible sequences that include a direct sequence (option 1.1), a direct sequence with a preconcentrator (option 1.2), and a nonsharp split sequence (option 1.3); Figure 6 illustrates the difference between a direct and an indirect sequence. In the context of azeotropic distillation, a preconcentrator is a specific type of nonsharp split that partly removes one or more of the binary azeotrope constituents and brings the feed to the azeotropic composition before it is sent to an azeotropic distillation column. The direct sequence (option 1.1) removes ethanol overhead, water and ethylene glycol under flow. In this case, an indirect sequence is not feasible because water alone cannot be removed from the bottom of the extractive column as it is a medium-boiling species. In the exceptional case when a homogeneous azeotropic mixture demonstrates the curious behavior of reversed volatility between the binary feed constituents, a direct sequence can recover the medium boiler overhead and the light and heavy boilers under flow. Acetone-isopropyl ether-dimethysulfoxide (DMSO) (high-boiling entrainer) is an example of such a system. Berg and Yeh reported that, upon distillation, nearly pure isopropyl ether is found overhead, while an acetone-DMSO mixture exits at the bottom of the extractive column.22 An important final note concerns the separability and operability of extractive distillation. Separability of this system is highly dependent on the location of the entrainer feed in relation to the azeotropic feed. The high-boiling entrainer feed is best placed near the top of an extractive column, above the azeotropic feed, to ensure that a sufficient concentration of entrainer is available in the extractive section (i.e., between the entrainer and the azeotropic feed) to break the azeotrope. If a single feed is used instead, separation becomes infeasible because this eliminates the column extractive section.11 5.2. Alternative Sequences for Cases WOBC-A1, WOBC-A3(i), WOBC-A3(ii), WOBC-B2, and WOBCB3 (Medium-Boiling Pure or Pseudo Entrainer). We might encounter one of the following ternary homogeneous azeotropic systems under this classification, i.e., the homogeneous azeotropic systems with (1) a medium-boiling entrainer introducing no new azeotrope (the binary azeotrope can either be maximum- or minimum-boiling; cases WOBC-A1 and WOBC-B2 shown in Figures 2 and 3, respectively), (2) a medium-boiling entrainer that forms a maximum azeotrope with the light boiler [case WOBC-A3(i) in Figure 2], (3) a lowboiling entrainer that forms a maximum azeotrope with the medium boiler [case WOBC-A3(ii) in Figure 2], (4) a medium-boiling entrainer that forms a minimum azeotrope with the high boiler [case WOBC-B3(i) in Figure 3], or (5) a high-boiling entrainer that forms a


Figure 5. Separation options for the ethanol-water-ethylene glycol mixture (case WOBC-A2, Table 1).

minimum azeotrope with the medium boiler [case WOBC-B3(ii) in Figure 3]. It is quite well-known that an entrainer does not have to be pure to enable separation. We define a pseudo entrainer as an entrainer with impurities in excess of about 10% (this figure should be regarded as a mere estimate and is chosen simply because this level of impurity is not normally acceptable in finished products). In this case, the pseudo entrainers are either the maximum or minimum azeotropes formed between the entrainer and one of the binary azeotrope constituents, and they result in distillation boundaries that separate the pure entrainer and one of the binary feed constituents into different distillation regions. Note that the first of the aforementioned systems, cases WOBCA1 and WOBC-B2, use a pure medium-boiling entrainer, whereaswhile the rest form medium-boiling pseudo entrainers [e.g., see the A/E azeo stream in cases WOBC-A3(i) and (ii), Figure 2]. The behavior of the systems with medium-boiling pseudo entrainers is equivalent to that of the system with a medium-boiling pure entrainer. Laroche et al.9 listed the possible sequences only for the cases that employ pure medium-boiling entrainers. Our screening procedure includes the systems with pseudo entrainers that also lead to a few other

feasible separation sequences. These systems are among the common feasible homogeneous azeotropic mixtures,18 and are important to consider as they give designers more options in finding, for example, an internal entrainer (referring species already present in the process). Figure 6 shows that, except in the case of an azeotropic feed, a maximum of five alternative separation sequences can be generated for a homogeneous azeotropic mixture that employs a pure medium-boiling entrainer that does not introduce new azeotropes. These include sequences with either direct or indirect sharp first splits, leading to two-column designs (option 2.1 or 2.2, respectively), or sequences with a preconcentrator (options 2.3 and 2.4) or a nonsharp first split (option 2.5), leading to three-column designs. Manan17 has shown that systems with medium-boiling pseudo entrainers also result in the same types of sequences as shown in Figure 6. The preceding analysis indicates that the available options for separating homogeneous azeotropic mixtures are fixed and limited. In particular, we learn that both the direct and indirect separation sequences exist only when the medium-boiling pure or pseudo entrainer is used. In ref 17, we show that this finding can also be generalized to homogeneous azeotropic systems with

5802


Figure 6. Separation options for case WOBC-A1 (and similar systems WOBC-A3, WOBC-B2, and WOBC-B3), Table 1. Mixture example: acetone-benzene-heptane (medium boiling entrainer) system.

boundary crossing. The observation leads us to the following heuristic: Heuristic 1. For a given binary feed composition, only the direct separation sequence is feasible for the lowboiling entrainer case, and only the indirect separation sequence is feasible for the high-boiling entrainer case. Both the direct and indirect separation sequences exist only when the medium-boiling pure or pseudo entrainer is used. Statements from Wahnschafft et al.12 and Laroche et al. claiming that azeotropic mixtures can result in a large number of possible separation sequences might be misleading because homogeneous azeotropic mixtures

do not normally lead to a large number of possible sequences, as generally assumed. Therefore, it should be realized that the proposed automation of separation sequences in Wahnschafft et al.23 might prove to be computationally expensive and inefficient, at least as far as homogeneous mixtures are concerned. Because the number of separation options is naturally limited, it is possible to produce a catalog of the most promising separation sequences for a wide range of homogeneous azeotropic mixtures commonly encountered in the process industry. Such “off-the-shelf” information would be valuable during the presynthesis stage, as it eliminates potentially repetitive screening and optimization tasks.


In the next section, the alternative separation sequences generated are screened by means of geometric reasoning to produce the most promising separation sequence for each class of homogeneous azeotropic system listed in Table 1 and Figures 2 and 3. 6. Screening the Alternative Sequences Each of the common classes of azeotropic system discussed in section 5 can yield up to a maximum of five different types of sequences consisting of two or three units. Each class might include any combination of the direct sequence, the indirect sequence, the direct sequence with a feed preconcentrator, the indirect sequence with a feed preconcentrator, and the nonsharp split sequence. Our strategy for screening is to eliminate the infeasible or inferior sequences in three sequential steps: Step 1. Determine whether direct or indirect sequence is feasible (apply heuristic 1 from section 5). Step 2. Determine whether preconcentration is costeffective (apply heuristic 2 from this section) Step 3. Choose between the direct and indirect sequences when a medium-boiling entrainer is used. Step 1 was described in section 5 during sequence generation. Steps 2 and 3 are explained in detail in sections 6.1 and 6.2, respectively. 6.1. Step 2. Determine whether Feed Preconcentration Is Cost-Effective. The two-unit sequence is naturally more attractive than the three-unit sequence because of the high capital cost incurred by an extra column. This is true as long as the amount of entrainer required is sufficiently small. Clearly, a lower entrainer flow rate will lead to a smaller column and, hence, lower capital and operating costs. Knight and Doherty2 reported a 400% savings in the total annual cost when a preconcentrator was included in the extractive distillation of an ethanol-water system with ethylene glycol as the entrainer, concluding that it is always economical to preconcentrate a dilute feed. Ryan and Doherty, who studied several sequences for the heterogeneous separation of ethanol-water-benzene (with benzene as the entrainer), however, state that preconcentration managed to save only energy costs (but not capital costs) when the feed contains less than 4% ethanol. This last study reported a marginal 7% net savings in the annual energy cost after the capital investment for a preconcentrator column was deducted. The savings tended to diminish at a slightly higher ethanol percentages. The discrepancy between the two findings related to the advantages of preconcentration can be attributed to the peculiarities of azeotropic mixtures. Therefore, it might be necessary to resort to numerical optimization to determine whether preconcentration can indeed be cost-effective. We have observed that it is possible to establish this through geometric reasoning. In general, preconcentration is economical if it fulfills the following criteria: Criterion 1. Preconcentration is economical if it can reduce the entrainer requirement per unit of binary feed (i.e., the specific entrainer requirement, SER). Note that, even though preconcentration always reduces the total amount of entrainer required in an azeotropic column (because of the reduced binary feed load), the SER might increase or decrease. For example, assume that, before preconcentration, 100 kmol/h of entrainer is required to separate 100 kmol/h of a binary azeotropic mixture (SER ) 1.0). If only 80 kmol/h

entrainer is required to separate 70 kmol/h of the preconcentrated mixture, then the SER has increased to 1.14 even though the amount of entrainer has decreased. To calculate the SER, we need only the binary and ternary feed compositions and the lengths of the segments of the azeotropic column mixing line. The reduction in the amount of entrainer can be estimated if the flow rate of either the entrainer or the binary feed before preconcentration is known. Criterion 1 suggests that a reduction in the SER essentially guarantees that preconcentration will be economical. However, it is important to note that, in a specific case, preconcentration is economical even with an increase in the SER. Criterion 2 describes this special case. Criterion 2. Preconcentration is economical if it can minimize separation difficulty. Preconcentration also leads to substantial savings if it can minimize separation difficulty, e.g., through the removal of a heavy or more plentiful component as the bottoms product of a preconcentrator, thus preventing its removal as an overhead product of an azeotropic or an entrainer recovery column, or through a reduction of the size of an overhead recycle stream, particularly the one containing a heavy or more plentiful component. Criterion 2 should be used when the SER increases upon preconcentration. It can be concluded that preconcentration can only be excluded if both criteria 1 and 2 are not fulfilled. Example 1 illustrates a simple procedure that can be used to calculate the SER from the geometry of a RCM. Example 1. Determination of the SER. Figure 7 shows that, for a constant azeotropic column bottoms composition B1, the ratio of the lengths of the azeotropic column mixing line segments, |(M - F)/(E - M)|, decreases when the binary feed composition changes along the AB edge of the ternary diagram in the direction of pure A (e.g., from F to F1). By the lever rule, this ratio represents the SER (i.e., the entrainer-to-binary feed ratio); thus, it can be concluded that the SER tends to decrease in the same direction. This relation is true even though the ternary diagram in Figure 7 is a 90° triangle and distorts the actual distance between points, i.e., the distance measured in a 60° or equilateral triangle. The preceding analysis leads to the key heuristic for screening a homogeneous azeotropic distillation sequence: Heuristic 2. Favor the sequence that reduces the specific entrainer requirement or the one that minimizes separation difficulty. Otherwise, choose the minimumunit sequence. 6.2. Step 3. Choose between the Direct and Indirect Sequences when a Medium-Boiling Entrianer Is Used. In screening the direct and indirect sequences, there are apparent contradictions among findings of various researchers. Laroche et al.,9 Buell and Boatright,24 and Berg and Yeh22 state that “ the light boiler is often recovered first, but not always”. Foucher and Doherty,18 on the other hand, imply that the indirect sequence is preferable because it allows for the use of a single-feed column and eliminates the problems of trace-component accumulation. An azeotropic system can be problematic in terms of its separability and operability. These constraints can be used as criteria to choose between the direct and indirect sequences. A direct sequence can be sensitive to the number of column feed points. It normally

5804


Figure 7. SER decreases as the binary feed changes along the AB edge of the ternary diagram in the direction of pure A.

Figure 8. Effect of preconcentration for homogeneous systems without boundary crossing when a high-boiling entrainer is employed with the binary feed rich in the (a) light boiler A and (b) medium boiler B.

requires multiple feed points for optimal operation.10 In addition, a direct sequence is also more likely to cause trace components and light impurities to accumulate through use of an overhead entrainer recycle loop. The indirect sequence, on the other hand is flexible enough to accept single or multiple feed points without much effect on separability. To overcome these constraints,

the indirect sequence should be favored. Note that the operability criterion applies to all feed stream compositions, including an azeotropic feed. Heuristic 3. Choose the indirect sequence for the medium-boiling entrainer case. The three key screening steps described previously are applied in the next subsections to identify the most


Figure 9. Effect of preconcentration for homogeneous systems without boundary crossing when a medium-boiling entrainer is employed. (a) and (b) are the direct sequences with low- and high-boiling-component-rich feeds, respectively.

promising separation sequences for systems WOBC-A1, WOBC-A2, WOBCA3(i), and WOBC-A3(ii). The procedure leads to a catalog pairing the RCMs and the corresponding most promising separation sequence for each system analyzed (see Figures 11 and 12). 6.3. The Most Promising Sequence for Case WOBC-A2 (High-Boiling Entrainer/Extractive Distillation). An example of this system is an ethanolwater mixture with ethylene glycol (EG) as the highboiling entrainer. A binary feed that is rich in the medium boiler is shown in Figure 8a. The indirect sequence is infeasible by heuristic 1. The sequence with a nonsharp first split from option 1.3 is clearly inferior because it leads to impure products and more than the minimum number of units. The two-column sequence (option 1.1 of Figure 5), on the other hand, requires the medium boiler (water in this example) to be completely removed overhead of the entrainer recovery column. In contrast, the sequence with a preconcentrator (option 1.3 of Figure 5) has two advantages: (i) it reduces both the SER and the amount of entrainer required in the extractive column, and (ii) it allows some water to be removed at the bottom of the preconcentrator column, thereby partly avoiding the more difficult overhead removal of water in the entrainer recovery column. Figure 8a shows that, as the overall feed composition of the extractive column moves away from the entrainer tip, i.e., from M1 to M1′, less entrainer is required per unit flow rate of binary feed (the SER is reduced).

For a feed that is rich in water, these changes could result in a substantial savings in energy costs in excess of the capital investment required for the preconcentrator column. Knight and Doherty5 show that this is indeed the case for the ethanol-water-EG (highboiling entrainer) system. In this case, preconcentration fulfills heuristic 2. In conclusion, for extractive distillation, preconcentration is generally advantageous when the binary feed is rich in the medium-boiling component. Figure 8b presents the same homogeneous azeotropic system but now with a feed rich in the low-boiling component, A. In this case, a preconcentrator that removes part of the low-boiling component (ethanol in the example) changes the overall extractive column composition from M1 to M1′. M1′ is slightly leaner in A and richer in B and the entrainer. The flow rate of component B at the top of C-2 is the same for the sequences with and without a preconcentrator, but the amount of entrainer per unit of binary feed (SER) has slightly increased (see Figure 8b). Note also that, in this case, there is no overhead recycle stream. Apart from increasing the overall capital costs, a preconcentrator in this case increases the SER. Thus, there is no advantage to adding a feed preconcentrator for extractive distillation when the feed is rich in the low-boiling binary feed constituent. By heuristic 2, the minimumunit sequence should be favored in this case. The conclusion of the preceding geometric analysis is supported by the simulation results presented in ref 1.

5806


Figure 10. Effect of preconcentration for homogeneous systems without boundary crossing when a medium-boiling entrainer is employed. (a) and (b) are the indirect sequences with low- and high-boiling-component-rich feeds, respectively.

6.4. The Most Promising Sequence for Cases WOBC-A1 and Similar Systems (Medium-Boiling Pure or Pseudo Entrainer). Both the direct and indirect sequences are feasible for these systems according to heuristic 1. Except in the case of an azeotropic feed for which preconcentration is not applicable, there are five possible sequences altogether, as shown in Figure 6. As usual, we eliminate the sequence with a nonsharp first split (option 2.5 of Figure 6). In the case of an azeotropic feed, we are left with the direct and indirect minimum-unit sequences to choose from. By heuristic 3, the indirect sequence (option 2.2 of Figure 6) should be chosen. This is in order to have a feed flexibility and to eliminate the possible problem of trace accumulation. For nonazeotropic feed, one must establish whether a preconcentrator can reduce the SER. Figure 9 shows the direct separation sequences for a homogeneous azeotropic mixture with a medium-boiling entrainer and either (a) a low- or (b) a high-boiling-component-rich feed. Both sequences result in overhead removal and recycle of the medium-boiling entrainer, making separation in the entrainer recovery column difficult. Figure 9a and b shows that feed preconcentration increase the SER for a low-boiling-component-rich feed but decreases the SER for a high-boiling-component-rich feed, respectively. Preconcentration is clearly economical for the high-boiling-component-rich feed as SER has been reduced. For the low-boiling-component-rich feed, even

though the SER has increased upon preconcentration, the entrainer flow rate has been reduced through a reduction in the binary feed. A feed preconcentrator in this case can minimize a difficult separation by reducing the size of an overhead entrainer recycle. Thus, feed preconcentration is very desirable for both cases a and b of Figure 9. The conclusions from this geometric analysis are supported by the simulation results presented in ref 17. The indirect sequences shown in Figure 10a and 10b, on the other hand, are faced with difficult azeotropic separation tasks as a result of the overhead recovery of the entrainer-component A mixture. Feed preconcentration is also desirable in these cases because of the reduction in the total entrainer flow rate and, hence, the much less difficult azeotropic separation. An indirect sequence that removes the entrainer and the low-boiling component overhead is preferred over a direct sequence by heuristic 3. 7. Conclusions A general procedure for the synthesis of optimal distillation sequences for azeotropic mixtures is currently unavailable. The results of an optimization study2 conducted for a specific type of homogeneous mixture cannot be generalized to other types of homogeneous mixtures because of the peculiarities of azeotropic mixtures. In the absence of general guidelines for


Figure 11. RCMs and most promising separation sequences for breaking minimum-boiling homogeneous azeotropes for cases WOBCA1 and WOBC-A2.

screening the alternative column sequences, it might be necessary to optimize and evaluate the economics of

every sequence generated to determine the most promising sequence. Such an approach becomes cumbersome

5808


Figure 12. RCMs and most promising separation sequences for breaking minimum-boiling homogeneous azeotropes for cases WOBCA3(i) and WOBC-A3(ii).

when a number of different entrainers are being evaluated and when a few separation options exist for each type of entrainer.

In this paper, we have proposed a systematic procedure for generating a catalog of the most promising separation sequences for homogeneous azeotropic mix-


tures without boundary crossing. The major developments and new insights in this paper include (1) geometric methods for the synthesis and screening of the alternative separation sequences for azeotropic systems without boundary crossing that exploit information such as the boiling points of pure components and azeotropes, the binary feed composition, and the specific entrainer requirements, most of which, can be extracted from the RCM; three key heuristics for screening the separation sequences for homogeneous azeotropic mixtures without boundary crossing; and a catalog pairing the RCMs of the ternary systems with the most promising separation sequences. We foresee that these predefined sequences, if incorporated in an intelligent system, can help a designer tremendously. Acknowledgment The authors thank Professors Rafiqul Gani and Jack W. Ponton for their critical review of a previous version of this work. We acknowledge the financial support provided by the Association of Commonwealth Universities, U.K., and Universiti Teknologi Malaysia for this research. Finally, we also acknowledge Aspen Tech for the provision of Aspen Plus for research purposes. Literature Cited (1) Schreinemakers, F. Phys Chem., Stoechiom. Verwandtschaftsl. 1901, 36, 257. (2) Knight, J. R.; Doherty, M. F. Optimal Design and Synthesis of Homogeneous Azeotropic Distillation Sequences. Ind. Eng. Chem. Res. 1989, 28, 564-572. (3) Ryan, P. J.; Doherty, M. F. Design/Optimization of Ternary Heterogeneous Azeotropic Distillation Sequences. AIChE J. 1989, 35 (10), 1592-1601. (4) Pham, H. N.; Doherty, M. F. Design and Synthesis of Heterogeneous Azeotropic Distillation I. Heterogeneous Phase Diagrams. Chem. Eng. Sci. 1990, 45 (7), 1823-1836. (5) Pham, H. N.; Doherty, M. F. Design and Synthesis of Heterogeneous Azeotropic Distillation II. Residue Curve Maps. Chem. Eng. Sci. 1990, 45 (7), 1837-1843. (6) Pham, H. N.; Doherty, M. F. Design and Synthesis of Heterogeneous Azeotropic Distillation III. Column Sequences. Chem. Eng. Sci. 1990, 45 (7), 1845-1854. (7) Stichlmair, J.; Fair, J.; Brave, J. Separation of Azeotropic Mixtures via Enhanced Distillation. Chem. Eng. Prog. 1989, 85 (1), 63-69. (8) Stichlmair, J.; Herguijuela, J. Processes of Zeotropic and Azeotropic Ternary Distillation. AIChE J. 1992, 38, 1523-1535. (9) Laroche, L.; Bekiaris, N.; Andersen, H.; Morari, M. Homogeneous Azeotropic Distillation: Comparing Entrainers. Can. J. Chem. Eng. 1991, 69, 1302-1319.

(10) Laroche, L.; Bekiaris, N.; Andersen, H.; Morari, M. Homogeneous Azeotropic Distillation: Separability and Flowsheet Synthesis. Ind. Eng. Chem. Res, 1992, 31 (9), 2190-2209. (11) Laroche, L.; Bekiaris, N.; Andersen, H.; Morari, M. The Curious Behavior for Homogeneous Azeotropic Distillation: Implications for Entrainer Selection. AIChE J. 1992, 38 (9), 13091328. (12) Wahnschafft, O. M.; Rudulier, J.-P. L.; Westerberg, A. W. A Problem Decomposition Approach for the Synthesis of Complex Separation Processes with Recycles. Ind. Eng. Chem. Res. 1993, 32, 1121-1141. (13) Rooks, R. E.; Julka, V.; Doherty, M. F.; Malone, M. F. Structure of Distillation Regions for Multicomponent Azeotropic Mixtures. AIChE J. 1998, 44 (6), 1382-1391. (14) Knapp, J.; Doherty, M. Thermal Integration of Homogeneous Azeotropic Distillation Sequences. AIChE J. 1990, 36 (7), 969-984. (15) Dongen, D. V.; Doherty, M. Design and Synthesis of Homogeneous Azeotropic Distillation. 1. Problem Formulation for a Single Column. Ind. Eng. Chem. Fundam. 1985, 24, 454-463. (16) Doherty, M.; Caldarola, G. Design and Synthesis of Homogeneous Azeotropic Distillations. 3. The Sequencing of Columns for Azeotropic and Extractive Distillations. Ind. Eng. Chem. Fundam. 1985, 24, 474-485. (17) Abdul-Manan, Z. Process Synthesis for Waste Minimisation with Emphasis on the Synthesis of Cleaner and Cost-Effective Distillation Sequences for Azeotropic Mixtures. Ph.D. Thesis, Department of Chemical Engineering, University of Edinburgh, Edinburgh, Scotland, 1998. (18) Foucher, E. R.; Doherty, M. F.; Malone, M. F. Automatic Screening of Entrainers for Homogeneous Azeotropic Distillation. Ind. Eng. Chem. Res. 1991, 30, 760-772. (19) Stichlmair, J. Separation of Ternary Mixtures by Rectification. Int. Chem. Eng. 1991, 31, 423-433. (20) Nadgir, V.; Liu, Y. Studies in Chemical Process Design and Synthesis: 5. A Simple Heuristic Method for Systematic Synthesis of Initial Sequences for Multicomponent Separations. AIChE J. 1983, 29, 926-934. (21) Malone, M. F.; Doherty, M. F. Separation System Synthesis for Nonideal Liquid Mixtures; AIChE Symposium Series; AIChE: New York, 1995; Vol. 91, pp 9-18. (22) Berg, I,.; Yeh, A. The Unusual Behavior of Extractive DistillationsReversing the Volatility of Acetone-Isopropyl Ether System. AIChE J. 1992, 31 (3), 504-606. (23) Wahnschafft, O.; Jurain, T.; Westerberg, A. SPLIT: A Separation Process Designer. Comput. Chem. Eng. 1991, 15 (8), 565-581. (24) Buell, C.; Boatright, R. Furfural Extractive Distillation for Separating and Purification of C4 Hydrocarbons. Ind. Eng. Chem. Res. 1947, 39, 695-705.

Received for review March 23, 1999 Accepted August 2, 2001 IE990215L

A New Catalog of the Most Promising Separation Sequences for

Recommend Documents