S e l e c t i o n o f a S e p a r a t i o n P r o c e s s
HAROLD R. NULL
Monsanto Company
St. Louis, Missouri
22.1 INTRODUCTION
There is no rote procedure by which separation processes are chosen. However, there are certain guidelines that usually are followed in arriving at the separation process sequence to be adopted in the plant design. The degree to which the optimum separation scheme is approached depends on the time and money allotted to development and analysis and on the skill of the process design engineer in applying guidelines to the selection of sequences for detailed evaluation.
It is safe to say that the ultimate criterion is economics. However, the economic criterion is subject to a number of intangible constraints. These constraints may include corporate attitude toward market strategy and timing, reliability, risks associated with innovation, and capital allocation. In this regard, separation processes are no different than any other type of process. To illustrate the influence of the intangible constraints, we might consider two extreme cases.
1. The product is of high unit value with a short market life expectancy.
2. The product is a high-volume chemical with many producers in a highly competitive market.
In Case 1 the procedure would probably be to choose the first successful separation method found. The ultimate overall economics of this situation are determined by getting into the market ahead of the com-petition and selling as much of the product as possible while the market lasts. However, if the market turns out to have a sustained life, it would be foolish in the extreme to continue to follow this philosophy on second- and third-generation plants. As time passes and competitors move in, products ultimately take on the nature of Case 2.
Even in Case 2, there are time and money constraints on the process development and design teams. However, the ultimate economic viability of the plant dictates that, within those constraints, the team do the most thorough job possible in the development and evaluation of various process schemes to make the closest approach to the economic optimum design possible. In its broadest meaning, this development and evaluation activity constitutes process synthesis. In a narrower context, process synthesis is a rigorous methodology for choosing the optimum flowsheet. In still narrower context process synthesis refers to the use of a computer program to develop the flowsheet.
22.2 INITIALSCREENING
The first step in the selection of a separation process is to define the problem. Usually, this entails estab-lishing product purity and recovery specifications. Product purity specifications are established ultimately
by customers, although they may not be stated explicitly by the customer. Recovery specifications are set to assure an economic process. Ideally, the recovery specification should be a variable to be optimized by the process design. In practice, there is usually insufficient time to consider recovery as a variable, and it is specified either arbitrarily or by consensus.
22.2-1 Property Differences
Having established the products to be separated and the purity and recovery desired for each, the next essential step is to determine which separation methods are capable of accomplishing the separation. In order for two components to be separable, there must be some difference in properties between them. The objective of design of the separation process is to exploit the property differences in the most economical manner to accomplish the separation. The following properties are used as bases for separation processes:
Equilibrium properties Vapor pressure Solubility
Distribution between immiscible liquid phases Melting point
Chemical reaction equilibrium Electric charge (isoelectric point) Surface sorption
Rate properties
Diffusivity
Ionic mobility Molecular size Molecular shape
22.2-2 Classification of Unit Operations
While separation unit operations often depend on several property differences for their overall success, there is usually one property difference that forms the primary basis for separation. The commonly used methods of separation and the primary basis for separation of each follows:
Distillation—vapor pressure
Extraction—distribution between immiscible liquid phases Crystallization—melting point or solubility
Adsorption—surface sorption
Reverse osmosis—diffusivity and solubility Membrane gas separation—difrusivity and solubility Ultrafiltration—molecular size
Ion exchange—chemical reaction equilibrium Dialysis—difrusivity
Electrodialysis—electric charge and ionic mobility Liquid membranes—difrusivity and reaction equilibrium Electrophoresis—electric charge and ionic mobility
Chromatographic separations—depends on type of stationary phase Gel filtration—molecular size and shape
22.2-3 Scale of Operation
The scale of operation often is the determining economic factor in the selection between alternative sepa-ration processes. For example, air sepasepa-rations on a very large scale are accomplished most economically by cryogenic distillation processes. Small-scale air separations, however, often are accomplished more economically by other means such as pressure swing adsorption or hollow-fiber gas separation membranes. The scale of operation at which the choice switches is case dependent and includes such factors as location and oxygen and nitrogen purity specifications. However, cryogenic distillation rarely is challenged by other methods for high-purity separations at air feed rates exceeding 100,000 standard cubic feet per hour (2832 m3/h).
TABLE 22.2-1 Approximate Single-Line Maximum Capacities for Selected Separation Processes
Separation method Upper commercial scale for single-line operationDistillation No limit
Extraction No limit (certain types of column have been used only up to 6 ft (1.83 m) diameter.
Crystallization 25-150 x 106 lbm/yr (10-70 x 106 kg/yr)
Adsorption No limit
Reverse osmosis 1 x 106 lbm/yr (0.45 x 106 kg/yr) water flow per module, with
many modules per unit
Membrane gas separation 2 x 106 lbm/yr (0.9 x 106 kg/yr) gas permeated per module, with
many modules per unit
Ultrafiltration 1 x 106 lbm/yr (0.45 x 106 kg/yr) water flow per module, with
many modules per unit
Ion exchange 1 x 109 lbm/yr (450 x 106 kg/yr) water flow
Electrodialysis 1 X 106 lbm/yr (0.45 X 106 kg/yr) water flow
Electrophoresis 1000 kg/yr carrier-free basis
Chromatographic separations 0.5 x 106 lbm/yr (0.24 x 106 kg/yr) gas, 0.2 x 106 lbm/yr (0.09 X
106 kg/yr) liquid
Gel filtration 100 kg/yr carrier-free basis
Any separation process chosen must be compatible with the scale of operation of the commercial plant. While it is common to see two or three parallel lines of operation in a commercial installation, plant operation becomes unwieldy when more than three parallel lines must be operated. In the case of very-high value products, perhaps as many as ten parallel lines might be considered; however, it is extremely rare that more than ten parallel lines would be considered. One must be careful how one interprets the term "parallel lines." For example, a membrane separation process may require many identical "modules" piped together in a single unit with common feed and product headers and a single instrumentation package which monitors the entire package. In such a case, the entire installation acts as a single unit; operators and control systems treat it as a single unit or line, regardless of the number of individual modules that may be included in the unit. On the other hand, a number of identical chromatographic columns operating in parallel will have to be monitored individually by the operators or have individual monitoring and control instrumentation. In this case, each column is effectively a separate line.
Many separation operations have an upper limit on the capacity which can be handled by a single unit, and this in turn limits the production scale for which that operation will be considered. In some cases, the upper limit represents a true limitation imposed by a physical phenomenon, while in other cases, the upper limit is simply the maximum scale on which the commercial equipment presently is manufactured. In either case, future developments are apt to make any tabulation of "single-line" limits subject to future obso-lescence.
At the risk of some entries being obsolete by the time of printing, a list of single-line upper limits for the various types of common separation process is given in Table 22.2-1. An entry of "no limit" in the table implies that some other equipment, either upstream or downstream of the separation, is almost certain to determine the single-line bottleneck.
22.2-4 Design Reliability
On all factors influencing the decision to choose one process in preference to another, design reliability is the most important. Regardless of any other considerations, the plant, when constructed, must work properly to produce an acceptable product that can be sold at a profit. The economic consequences of a design failure are too dire to accept an unworkable separation process design. Design reliability is not really definable in quantitative terms because it actually relates to the amount of testing and demonstration that must be done before a suitable commercial scale design is produced. A general statement regarding design reliability for several separation processes follows.
DISTILLATION
EXTRACTION
The design of extraction processes requires a knowledge of the phase-equilibrium relationships between the components to be separated and the extraction solvents to be considered. With this information and a knowledge of the pertinent physical properties (densities* interfacial tension, viscosities, etc.). one usually can project a preliminary design that is reliable enough to determine the choice of solvents and operating conditions to be used and the type of extraction equipment to use. However, the designs generated in this manner are not nearly so reliable as those for distillation. Consequently, small-scale testing on the type of equipment to be used in the plant is required. Given such tests, the vendors of proprietary extraction equipment can provide a reliable scale-up to the commercial-size equipment. For nonproprietary equipment, reliable scale-up can be done for mixer-settlers, sieve tray columns, and columns packed with Raschig rings. Vendors of proprietary packings also claim to be able to scale-up reliably, but there is little if any public data available to assess the reliability of scale-up methods.
CRYSTALLIZATION
The design of crystallization equipment is quite difficult. Knowledge of the phase equilibrium and physical properties does not allow one to predict the behavior of the crystallization process. Bench-scale experiments on the actual stream always are required, and equipment vendors usually require a pilot-plant test in small commercial equipment before designing the equipment. Even with such testing, operating adjustments usually must be made on the actual commercial installation before the equipment will operate to yield an acceptable product. Even with such testing, there are occasional failures at the commercial scale. However, since crystallization often can offer the potential for the purest product attainable by any separation process at considerably less energy consumption than other methods, we can expect to see its use increase sub-stantially in the future. Many of the design reliability deficiencies are likely to diminish as more widespread use and large-scale testing occur.
ADSORPTION
Adsorption equipment and operating cycles usually can be designed reliably on the basis of measured adsorption isotherm data and a reasonable number of relatively small-scale experiments to determine the mass transfer characteristics of the stream to be treated in the chosen adsorbant bed. If the stream treated contains several adsorbed components it is generally necessary to run experiments on the actual mixture to be treated, since multicomponent isotherm behavior cannot be predicted in general from the individual isotherms.
REVERSE OSMOSIS
The design of reverse osmosis units usually requires that a single module or a smaller-scale test unit be operated using the actual feed stream to be treated. The tests are conducted not only to determine the operating characteristics of the membrane, but also to determine what pretreatment of the stream may be necessary to protect the membrane from fouling and structural damage. Once the appropriate tests have been made, however, the larger-scale design can proceed with confidence since most large-scale reverse osmosis installations consist of a number of modular units operating in parallel.
MEMBRANE GAS SEPARATION
The comments on reverse osmosis apply equally well to gas separation with selective membranes.
ULTRAFILTRATION
The evolution of a design for an ultrafiltration process requires extensive testing on the bench scale and normally requires pilot-plant operation. Not only must the equipment design be determined by testing, but also the operating cycle and recirculation flow rates must be resolved to prevent excessive fouling or "concentration polarization." There are a number of ultrafiltration configurations from which to choose (e.g., hollow fiber, plate-and-frame, and spiral-wound). If an optimum choice is to be made, all the configurations to be considered must be tested, since it is not possible to predict the characteristics of one configuration or vendor source from the data obtained on another. One also must determine by test the passage and retention of the components to be filtered and how this changes with time and operating conditions. The rated molecular weight cutoff associated with a particular membrane is only a nominal rating, and the actual fractional retention may vary considerably between components of the same molecular weight.
I O N EXCHANGE
DIALYSIS AND ELECTRODJALYSIS
The comments on reverse osmosis apply equally well to dialysis and electrodialysis with selective mem-branes.
ELECTROPHORESIS
Electrophoresis, at this time, is an inherently small-scale process. While there is considerable development underway to increase the capacity of electrophoresis equipment, the only way to determine the separation characteristics at present is to test the stream to be treated in the actual equipment to be used in the commercial installation. Thus, the only reliable scale-up possible consists of using multiple parallel units of the largest type tested.
CHROMATOGRAPHIC SEPARATIONS
Scale-up of chromatographic separations usually is done by testing at several sizes following a sequence such as: analytical column, 1 in. (245 mm) diameter preparative column, 6 in. (15 cm) diameter column, 3 ft (1 m) diameter column, with optimum injection-elution policy determined at each step. Each scale of operation has its own distribution and packing method problems associated with it, which can be solved only partially by testing at a smaller scale.
GEL FILTRATION
Gel filtration is a chromatographic operation, and all the comments regarding chromatography also apply to gel filtration. There are additional problems associated with gel filtration, however, because often the packing consists of a swollen gel with little structural strength. Thus, the maximum packing height is an additional factor to be determined at each scale of operation. Since such gels depend largely on wall friction for their support, the maximum height of packing decreases with increasing column diameter.
22.2-5 Necessity for Pilot Plant
The need for pilot-plant operation to test separation process design is discussed under the various unit operations in the section on design reliability. Aside from the considerations of separation design reliability, it may be necessary to pilot plant a process to test the integrated operation of all the units. If pilot plant operation is needed for that purpose, then all the units must be included in the pilot plant whether or not their individual testing is required for design reliability.
22.2-6 Number of Steps Required
A strong variable in the cost of separation processes is the number of processing steps required to accomplish the desired separation. However, the number of steps is not by any means an absolute determinant of the cost of a separation process. There are many instances in which it is more economical to install a separation sequence incorporating two or more steps when it would be possible to accomplish the separation with a single-step process. A specific example is heavy water separation, which can be accomplished in a single distillation column with very many stages and very high reflux but is accomplished more economically in a multicolumn chemical-exchange process involving hydrogen sulfide. Nevertheless, in the majority of cases, a single-step process will prove more economical than a multistep process. A number of separation methods are discussed with regard to their ability to accomplish essentially complete separations in a single step.
DISTILLATION
Unless there is an azeotrope formed between the components to be separated, it is possible to separate completely a binary mixture into its components in a single distillation column. In some cases it is even possible, with the use of sidestreams, to obtain more than two pure components from a single distillation column. In general, a multicomponent mixture in which there are no azeotropes and no sidestreams with-drawn can be separated completely in a distillation train using one column less than the number of product streams required.
EXTRACTION
CRYSTALLIZATION
In those cases in which complete solid solution with no eutectic formation occurs, it is possible to separate completely a binary mixture into its components via countercurrent multistage melt crystallization. How-ever, such systems are the rare exception, whereas eutectic formation is the general rule in solid-liquid equilibria. Therefore, in the overwhelming majority of cases, the degree of separation attainable by crys-tallization is limited by the eutectic composition. To accomplish a complete separation, it is necessary to couple an auxiliary separation step to the crystallization process to break the eutectic.
In the case of solution crystallization, one product is usually the pure crystal with occluded mother liquor and the other is either the mother liquor (a saturated solution of the crystalline material in the solvent) or a vapor or condensate of the solvent. The solid product stream must be subjected to a subsequent wash and filtration step (often with centrifugation) to recover the crystalline material in pure form.
ADSORPTION
Adsorption processes deposit one or more components selectively onto a granular solid material. If the adsorbing material is to be reused or the adsorbed material is to be recovered as a product, the adsorption process must be coupled with a regeneration step. Thus, adsorption is inherently a two-step process.
REVERSE OSMOSIS
The products from a reverse osmosis separation step consist of a purified solvent stream and a solution somewhat more concentrated than the feed stream. Therefore, it is not a complete separation process and must be accompanied by additional processing steps if complete separation or high recovery of solvent is desired. The degree of recovery or extent of concentration of the reject stream is limited by the osmotic pressure of the concentrated solution. The membrane must maintain its integrity and permit reasonable flux of solvent with a pressure difference across the membrane exceeding the osmotic pressure. In general, excessive osmotic pressures are encountered when the solute mole fraction in the concentrated solution exceeds 0.03-0.06.
MEMBRANE GAS SEPARATION
It is theoretically possible to accomplish a complete separation of a pair of gases by membrane separation, either by very highly selective membranes combined with very low partial pressure of the permeating component on the permeate side of the membrane, or by multistage cascades. However, it usually is not economical to accomplish high purity and high recovery of both components of a binary by membrane separation alone. Therefore, it usually is necessary to couple membrane gas separation with other processing steps when nearly complete separation is required.
ULTRAFILTRATION
Theoretically, ultrafiltration can separate two high-molecular-weight solutes from each other completely if the difference in molecular size is sufficiently large (differing by a factor of 10). When the molecular weight-size difference is not quite so great, it still may be possible to accomplish the complete separation by cascading. In any case, the two product streams from ultrafiltration consist of very dilute solutions and additional processing steps will be required if pure products are required. If the desired product form is a more concentrated solution, ultrafiltration may be used to accomplish the concentration.
I O N EXCHANGE
Ion exchange is similar to adsorption in that a regeneration step must be included to recover the product and/or reuse the bed. In addition, the product streams from ion exchange are usually aqueous solutions and additional processing steps must be included if the products are desired in pure form. In the case of water purification, probably the most widely used application of ion exchange, the deionized water effluent from the ion exchanger is the desired product and further processing is unnecessary.
DIALYSIS A N D ELECTRODIALYSIS
Dialysis processes, like extraction, are selective transfer processes rather than complete separation pro-cesses. Therefore, if the transferred components are desired in pure form, additional processing steps are required. Unlike extraction, the solvent does not contaminate the retained stream since the solvent on both sides of the membrane is the same.
ELECTROPHORESIS, CHROMATOGRAPHIC SEPARATION, A N D G E L FILTRATION
Electrophoresis, chromatography, and gel filtration all separate components in dilute solution. They must be followed by concentration steps if a dissolved product is satisfactory, and additional processing steps if pure products are desired.
22.2-7 Capital and Capital Related Costs
discuss the various factors in the selection of a separation process and, qualitatively, their effect on the capital required. When we compare two alternative separation processes, we choose the one with greater capital only if the revenue generated (savings in expense) is great enough to justify the expenditure of the difference in capital requirements between the alternatives. It is pertinent to recognize that there are certain expenses generated by or correlated to the capital itself. Such capital related expenses (e.g., depreciation and maintenance) generate annual expenses in the range of 20-25% of the capital expenditure. Among the factors that affect capital are the following.
SCALE OF OPERATION
A widely used rule for estimating the effect of scale of operation on capital requirement is to make capital proportional to the 0.6 power of capacity. While this usually works quite well when applied to the total cost of large plants and to individual operations that can be carried out in a single unit regardless of the capacity required, it must be modified considerably for small-scale operations and for operations that must be carried out in a number of parallel lines. For very small-scale operations, the required capacity may be below that of the minimum feasible size of certain critical equipment items, in which case the capital cost is virtually independent of the scale of operation. An example of this is a situation in which instrumentation is the dominant cost and the same instrumentation can be used regardless of the scale of operation. On the other hand, when multiple parallel lines must be used because the capacity of the plant is greater than the maximum single line size for the method of separation chosen, the capacity exponent tends to be between 0.9 and 1, rather than the conventional 0.6. For those separation methods using a single line of operation composed of multiple modules, the exponent tends to lie between 0.6 and 0.9. The scale of operation applicable to a number of separation methods is discussed in Section 22.2-3.
DESIGN RELIABILITY
Capital requirements are affected indirectly by design reliability in that larger safety factors are applied when design methods are less reliable. There is also a tendency to install more extensive control systems and more spare backup equipment when there is uncertainty in the reliability of the design methods. Such practices significantly increase the capital requirement for those methods with low design reliability. The design reliability of the various separation processes is discussed in Section 22.2-4.
NECESSITY OF A PILOT PLANT
Pilot plants normally are considered part of the research and development expense and usually do not represent a part of the capital of a specific plant. In some cases, especially where design reliability is low, the pilot plant may continue to be maintained as part of the plant and used for trouble-shooting or to test proposed plant modifications. In such cases, the cost of the pilot plant becomes a component of the plant capital. Section 22.2-5 discussese the need for pilot plants for the various methods of separation.
NUMBER OF STEPS REQUIRED
A crude method of very preliminary capital estimation is to count the number of processing steps required and multiply by the capacity and a cost factor. While such a method is not recommended for anything other than very preliminary qualitative thinking, it does point out the fact that capital is related directly to the number of steps in the complete process. A method of separation that can accomplish a complete separation in a single step is usually a lower capital process than one that requires a number of steps. The number of steps associated with the various separation methods is discussed in Section 22.2-6.
22.2-8 Energy Requirements
One of the most important considerations in choosing among separation processes is the energy requirement of the process. Energy cost is often the dominant operating expense, and energy requirements also have a direct influence on a portion of the capital in the form of heat exchangers, compressors, pumps, and so on. "Energy requirement'* can be a deceptive term, however, because our real concern regarding energy costs relates not to the quantity of energy that flows through the process but rather to the amount of fuel that must be burned to induce that flow of energy. If we had only the first law of thermodynamics to consider, these two quantities would be in direct proportion. Second law effects distort the relationship between energy flow and fuel burned, thereby forcing us to consider not only the quantity of energy flow but also the quality of the energy that flows through a process. As always, the ultimate choice is based on an economic comparison of the specific cases considered rather than generalizations. However, it is possible to draw some general conclusions regarding comparisons between separation processes which allow one to reduce the number of alternatives to be given the full economic analysis. The discussion that follows is based on the paper by Null.1
above the ambient air temperature. Generally, the heating value of the fuel burned to deliver this work will be three to four times the quantity of work delivered via this route, depending on the efficiency of the utility steam system. To supply the heating requirements of the plant, steam must be diverted from this expansion path at a sufficiently high temperature that it can be transferred into the process at the point of use. As a result, the amount of work delivered by that steam will be reduced by the following amount:
(22.2-1)
Since the diverted power must be replaced by additional steam expanded to the ultimate condensing pressure, the fuel heating value required to deliver this heat is given by
(22.2-2)
When refrigeration is required to move heat from a lower temperature to a higher temperature, work must be supplied by burning fuel, giving
(22.2-3)
(22.2-4)
where
Energy requirements are based on the ultimate fuel required as given by Eqs. (22.2-2) and (22.2-4) with the recognition that the temperatures Ts and Tr do not represent a continuum of temperatures, but rather
only the discrete temperatures at which the steam and refrigeration utilities are available.
We can make generalizations only if we make extensive approximations. While not ail methods of separation can be readily generalized, some are summarized in the following discussions. To compare the fuel efficiencies of different separation processes, we derive the approximate equation for fuel equivalent for each and compare the values.
DISTILLATION
Since distillation historically has been the separation method of choice whenever it could be used, it represents a suitable base case against which to compare the energy economy of other separation processes. We can approximate the heat requirement of distillation processes by
(22.2-5)
with approximately the same quantity of cooling required for condensation. The fuel requirements for the reboiler are
(22.2-6)
If the condenser is cooled by air or water, the fuel requirement is negligible compared to the reboiler, but if refrigeration is required a fuel penalty accrues to the condenser as well as the reboiler:
(22.2-7)
On the other hand, if useful heat is recovered from the condenser, a fuel credit accrues to the condenser:
(22.2-8)
EXTRACTION
If we examine only the typical extraction process, the only direct requirement appears to be the pumping energy to bring the feed and solvent together and the power for internal agitation if internal agitation is used. These energy requirements are certainly negligible in comparison to the energy requirements for distillation. However, the extraction column does not accomplish the desired separation alone; it merely selectively transfers one or more components into the solvent. To complete the separation it is necessary to separate the extracted product from the solvent. In most cases it is also necessary to remove the small amount of dissolved solvent from the raffinate. These downstream separations are accomplished most frequently by distillation, and the energy requirements of the downstream separations determine the true energy requirements associated with the extraction process.
If the solvent is less volatile than the components of the feed stream, the fuel equivalent for the recovery of the extracted product from the solvent is approximated by
(22.2-9)
and the ruel equivalent for recovery of solvent from the raffinate is given by
(22.2-10)
If the solvent is more volatile than the feed components, the fuel equivalent for the recovery of extract product is given by
(22.2-11)
and the fuel equivalent for solvent recovery from raffinate is given by
(22.2-12)
It is possible, of course, that the solvent might be less volatile than the raffinate and more volatile than the extract, or vice versa. In any event, the total fuel equivalent is determined by adding the appropriate subset of Eqs. (22.2-9)-(22.2-12). These equations assume that the overhead condensers in the recovery distil-lations accrue neither penalty (refrigeration) nor credit (heat recovery). If this is not the case, the appropriate adjustments may be made in the manner described in the paragraphs on distillation.
While we cannot make generalizations to cover every case, we can draw some conclusions regarding the energy efficiency of extraction compared to a distillation process for accomplishing the same separation. For example, Eqs. (22.2-9) and (22.2-10) compared with Eq. (23.2-6) lead to the conclusion that extraction is more efficient for the nonvolatile solvent whenever
For the case of the volatile solvent, a similar comparison gives
CRYSTALLIZATION
The heat requirement for continuous countercurrent crystallization can be approximated in a manner similar to that used for distillation:
(22.2-13)
with about the same amount of cooling required at the cold end of the cascade. The equivalent fuel associated with the melting is given by
(22.2-14)
The freezing or crystallization end of the cascade usually requires some degree of refrigeration with the associated fuel equivalent given by
(22.2-15)
Comparison of the above equations with Eq. (22.2-6) yields the criterion that crystallization is more energy efficient than distillation when
Since the value of AHf/AHv is typically about 0.2 and the required reflux Rc « Rd, the above criterion
usually shows a potential for a substantial energy advantage for melt crystallization whenever it can be used to accomplish the same separation as distillation. This conclusion holds true even when refrigeration is required to a fairly deep level. Discussions in previous sections regarding design reliability, scale of operation, and number of steps required indicate why crystallization is not used more widely in spite of its inherent energy advantages.
GAS ADSORPTION
The energy requirement associated with gas adsorption is the energy needed to regenerate the spent bed. This is accomplished most frequently by heating the entire bed to such a temperature that the adsorbed material is desorbed and leaves as a vapor. The heat of regeneration includes the sensible heat of the entire bed and the vessel containing it plus the heat of adsorption of the adsorbed material. The heat of adsorption is greater than but of the same order of magnitude as the heat of vaporization for physically bound adsorbates, and much greater than the heat of vaporization for chemisorption. A similar analysis to that used above shows that adsorption is more energy efficient than distillation whenever
Since the least-volatile component is usually the material most strongly adsorbed, in cases in which ad-sorption or distillation could be used, B — Nads. Thus, the criterion above indicates that gas adsorption is
preferable to distillation in those cases in which trace quantities of nonvolatile material are to be removed from the bulk of the feed stream. These are the instances in which adsorption normally is considered regardless of energy requirements.
REVERSE OSMOSIS
Reverse osmosis uses a membrane to separate a virtually pure solvent from a solution while simultaneously concentrating the solutes in the solution. The motive force for the separation is the pressure drop across the membrane, which must be considerably greater than the osmotic pressure of the solvent in the concen-trated effluent to generate an acceptable flux through the membrane. The fuel equivalent is related to the pumping work required by the expression
(22.2-16)
at solute concentrations exceeding 5 mol %. Thus, the appl cation of reverse osmosis is limited to those cases in which membranes have been developed with the desired selectivity and with sufficient structural integrity to withstand the required pressure drop.
OTHER M E T H O D S OF SEPARATION
The energy requirements for other methods of separation are not reduced so readily to simple approxima-tions. For example, energy requirements for gas separation by membranes are strongly dependent on the pressure at which feed is available and the required pressures of the permeate and permeant products, as well as the number of cascade stages required for the separation. Ion-exchange energy requirements are indirect inasmuch as they are associated with the energy required to produce the chemicals used for regeneration. Separation methods that require dilute liquid solutions, such as chromatographic separations of all kinds and electrophoresis, have energy requirements related to the handling and purification of the large quantities of water or carrier gases inherently associated with the process. While the direct energy requirements may be small at times, the indirect energy requirements are usually very large when compared to more conventional bulk separation methods such as distillation and extraction.
22.3 CHOOSING THE BASE CASE
A great deal of research has been reported in the synthesis of process flowsheets in recent years, and attempts have been made to develop systematic algorithms whereby the best separation sequence can be synthesized with a minimum of input to a computer program. In practice, separation processes rarely are chosen in this manner. Neither the computational algorithms nor the quantitative descriptions of all the separation methods is developed sufficiently well at this time to allow such automatic choice between separation methods. However, current developments in process synthesis can be of value in choosing a near-optimum sequence of steps all involving a single type of separation process, such as distillation.
Current practice is to make a qualitative judgment, based on the several criteria described in Section 22.2 of the methods or combination of methods which have the greatest probability of providing a process to accomplish the desired separation. Out of such a list one is chosen as a base case. This choice will represent the qualitative judgment of the process designer as to which of the processes listed is likely to be the most economical method of separation that can be developed within the time and money constraints of the project. If distillation is among the list of possible methods, it almost always will be the first base case chosen. A rereading of Sections 22.2-1-22.2-6 will make it quite evident why distillation is favored when it can accomplish the desired separation. Following distillation, the next most likely base case would use extraction when possible. Crystallization, adsorption, and ion exchange would be the next most likely base case choices, with all the others following in no particular order.
Once the base case has been chosen, process designers will propose a design and determine capital and operating costs in as much detail as the current knowledge of the properties of the feed stream mixture and its components and operating characteristics of the chosen unit operation warrants. At this stage, assump-tions are made and noted when needed data are not available. Determination of these needed data becomes the subject of subsequent research and development. After the base case has been evaluated, a judgment is made regarding the next most likely economic competitor process. The challenger process then is evaluated with regard to capital and operating costs with assumptions noted in the same manner as the base case. If the challenger process has lower capital and operating costs, it then becomes the new base case. In case the process with the lower operating cost requires the greatest capital, the saving in operating cost is compared to the difference in capital to determine whether the return generated by the savings meets the corporate criterion for return on capital. If it does, the lower operating cost process becomes the base case. If not, the lower capital process becomes the base case.
This sequential procedure is continued until all the candidate processes have been evaluated. However, the final choice has not been made necessarily at this point. In the early stages of process development, many assumptions have to be made regarding unavailable data. The assumptions made in the latest base case must be either verified or modified as additional experimental data are obtained. As additional infor-mation is obtained, both capital and operating costs usually rise compared to earlier estimates. This may force a reevaluation of the base case and some of the closest alternatives. The reevaluation may dictate a change in the base case choice. In addition, initial evaluations are done on process schemes that have not been optimized. If the evaluations indicated a large difference between alternatives, optimization of the processes usually will not change the choice. However, in cases where the economic differences between alternatives is not large, optimization of the processes can change the choice of a separation process.
When the continuing activities of evaluation, obtaining data, reevaluation, optimization, and demon-stration are completed, the final base case will be the actual process installed.
22.4 PROCESSSIMULATION
for complete flowsheets as well as individual unit operations. Not all separation methods are covered in these software packages, but most include at least distillation, absorption, and extraction. In cases where subroutines are not available for the specific separation method of interest, there is usually a "generic separation'* subroutine available that at least will allow the computation of the overall material and energy balance of the operation and its effect on the complete flowsheet. Some simulation programs also have cost estimation capability built into the program.
Process simulation programs represent a very powerful tool for the process engineer in the selection, optimization, and design of separation processes. If the procedures suggested in this chapter are to be carried to fulfillment, many evaluations will have to be made before a final choice is possible. When properly used (as a powerful tool and not as a substitute for engineering judgment), process simulation can decrease greatly the time and personnel required to reach a decision among separation processes. It even can be used in the early stages of process development to guide the course of the research.
There is often a hesitancy to use simulation programs in the early stages of process development because many of the parameters required as input to the program are not well known. In such cases, it is proper to use estimated values that bracket the range of probable expected values and to test the simulation program's sensitivity to parameters. In this way, we can decide which properties must be determined experimentally in order for a valid choice to be made. It is also appropriate, in the early stages of process selection and development, to reduce the computation time required in the simulations by using short-cut methods in the unit operation subroutines if reasonably accurate short-cut subroutines are available. In so doing, however, the engineer should be aware of the assumptions implicit in the short-cut computations and should use rigorous simulations where the short-cut assumptions can result in serious errors. The process designer also should recognize that the early simulation results are tentative and as the development or design proceeds toward the final selection or design, the simulations will be repeated a number of times with a shift from assumed ranges of parameter values and short-cut methods to firm values of parameters and rigorous simulations.
22.5 PROCESSSYNTHESIS
In recent years a considerable amount of literature has appeared in chemical engineering journals on the subject of "process synthesis." Nishida et al.2 reviewed the current state of the art as it applies to chemical
processes. If process synthesis methodology were developed fully, one could invoke a computer program which, with minimal input, would select not only the separation method of choice but also would determine how the various unit operations should be connected together to form the entire flowsheet. At the present time, however, process synthesis methodology for separation processes is in the early stages of research.
Very little process synthesis literature has dealt successfully with the problem of selection between alternative methods of separation. Indeed, most process synthesis literature has dealt with the selection of the flowsheet sequences for a single separation method with only simple, sharp separations between com-ponents of adjacent selectivity, without recycle. Furthermore, most of this literature has used only distil-lation in illustrative applications. In practice, much use is made of recycle, nonsharp splits, and complex columns (in distillation); and, of course, distillation is not always the separation method of choice.
While process synthesis research has not yet produced the solution to the problem of generation of the best flowsheet, a number of heuristic rules have been stated by various process synthesis researchers which can be of considerable aid in guiding the selection of the near optimum or base case and for systematically choosing those alternate sequences to be investigated. Some of the heuristic rules developed will be com-peting and in some cases even contradictory, but nevertheless they do serve to reduce the enormous amount of work that would be involved in evaluating all possible separation sequences.
Among the more widely accepted heuristics for selection of separation processes are the following:
1. Favor distillation unless relative volatilities are less than 1.05. 2. Do the easiest separations first.
3. When selectivities do not vary widely, remove the component having the highest mole fraction first.
4. Remove the component with the highest volatility first when distillation is feasible.
5. When mass separating agents are used (i.e., extractive distillation, or extraction), perform the recovery steps for the mass separating agent and/or dissolved products in the immediately following step.
6. Do no use a second mass separating agent to remove or recover a mass separating agent.
While these rules sometimes do not yield the optimum flowsheet, when applied with reasonable judgment they can lead to an acceptable base case against which alternative flowsheets can be prepared.
optimum case. Such methods are still under development and no generally accepted strategy has emerged yet. However, two useful methods which have been proposed are (1) to challenge one by one the heuristics which led to the base case selection and (2) some form of branch and bound technique. The many branch and bound strategies that have been proposed use the common principle that, once a base case has been established* an alternate sequence need be evaluated only through the step in which its cost or other objective function exceeds the base case.
Process simulation is now a powerful tool for process design. As process synthesis methodology ma-tures, it is reasonable to expect process simulation to be a powerful subset of process synthesis. Before that becomes a reality, however, much must be done to currently available process simulation packages. More separation operations must be incorporated into the simulation packages with their associated cost calculations. The most serious deficiency in the simulation software, however, is the absence of short-cut computations of sufficient accuracy to lead to valid conclusions.
In any process synthesis algorithm, simulations must be done so many times that only very rapid simulation computations are possible. Usually, this means short-cut computations even with high-speed computers. Unfortunately, short-cut calculations available today have implicit in them assumptions that can lead to completely erroneous conclusions. Distillation calculations provide a ready example. Usually, process synthesis investigations of distillation sequences have made use of Fenski-Underwood equations to simulate the individual columns. These equations carry the assumption of constant relative volatility of all the components. In many cases, this assumption merely affects the magnitude of the results but does not alter the conclusions reached. In other very commonly occurring cases, an unworkable flowsheet will result. Any computation method that assumes constant relative volatility cannot simulate an azeotrope, a very common occurrence in petrochemical processes. The constant relative volatility assumption will lead to the conclusion that the key components can be separated in a single distillation column to the degree of purity and recovery specified, although in some cases an unacceptable number of stages may be required. In actual practice an azeotropic system will require a minimum of two distillation columns although the number of total stages might be modest. Unfortunately, at the present time, only rigorous methods have the capability to simulate the azeotropic system sufficiently well to produce results of the required quality. It is currently more fashionable to spend research effort on the structure of the process synthesis methodology than on the development of high-quality, rapid computation algorithms for the individual unit operations. However, the ultimate utility of the process synthesis software to the process designer will depend on the availability of the rapid, high-quality simulation of the individual separation operations.
Process synthesis, at the present time, falls far short of its goal of providing a systematic procedure that assures us of the selection of the optimum methods of separation and the optimum sequence of those methods. Therefore, for some time to come, we will continue to have to rely on the more subjective and qualitative methods described in previous sections. However, we can begin to adapt some of the method-ology being generated by process synthesis research to some phases of the selection of separation processes. We also can continue to encourage the development of process synthesis methodology. In the author's view, however, this development will be slow to come regarding selection between methods of separation. We should expect more rapid progress in the optimization of the separation sequence for a given method of separation such as distillation.
NOTATION
molten product molar flow rate from high-temperature end of a melt crystallizer molar flow rate of distillation overhead product
molar flow rate of extract stream from extractor
combustion efficiency of the steam plant, that is, ratio of heat actually transferred to steam cycle to the heating value of the fuel
combined combustion and Carnot efficiency of the steam plant operating across its maximum temperature span
power cycle efficiency, that is, ratio of the actual power delivered to that of a Carnot engine across the same temperature span
refrigeration cycle efficiency, that is, ratio of power required by a reverse Carnot cycle to the actual power required by the refrigeration cycle across the same temperature span
molar feed rate to a reverse osmosis unit heat of fusion of crystallizer product
heat of regeneration per mole of adsorbed material in a gas adsorption process heat of vaporization of distillation overhead
heat of vaporization of extraction solvent
rate of heat transfer
heat requirement in a melt crystallizer heat requirement for distillation
fuel equivalent, that is, heating value of fuel required to deliver a power or heat requirement fuel equivalent for distillation condenser
fuel equivalent for heating in melt crystallizer fuel equivalent for refrigeration in melt crystallizer
fuel equivalent for recovery of extract product in an extraction process fuel equivalent for distillation reboiler
fuel equivalent for solvent recovery from the raffinate in an extraction process fuel equivalent of a reverse osmosis process
molar flow rate of raffinate from extractor reflux ratio of melt crystallization process reflux ratio of distillation process
reflux ratio required for extract product recovery gas law constant
reflux ratio required for solvent recovery from raffinate molar flow rate of extraction solvent
heat rejection temperature (absolute) of power cycle temperature (absolute) of refrigeration utility requirement
temperature (absolute) of refrigeration utility for distillation condenser temperature (absolute) of reverse osmosis process
temperature (absolute) of utility steam required for distillation reboiler
temperature (absolute) of utility steam replaced by heat recovered from high-temperature distillation condenser
temperature (absolute) of utility steam required for melt crystallization
temperature (absolute) of utility steam required for recovery of extract product from solvent maximum temperature (absolute) of steam in the power cycle
temperature (absolute) of utility steam required for recovery of solvent from raffinate temperature (absolute) of utility steam required for regeneration of adsorption bed work
power cycle work equivalent to a heat requirement work required for reverse osmosis process
mole fraction of solvent in the retentate from a reverse osmosis process Mole fraction of extraction solvent in the raffinate from an extractor activity coefficient of solvent in the retentate from a reverse osmosis process osmotic pressure of the solvent in the retentate from a reverse osmosis process
REFERENCES
22.2-1 H. R. Null, Energy Economy in Separation Processes, Chem. Eng. Prog., 76(8), 42-49 (1980). 22.2-2 N. Nishida, G. Stephanopoulos, and A. W. Westerberg, A Review of Process Synthesis, AIChE