Otto 1977 - Shortcuts For Distillation Design

You might also like

Download as pdf
Download as pdf
You are on page 1of 20
COMMONWEALTH OF AUSTRALIA Copyright Regulations 1969 WARNING This material has been provided to you pursuant to section 49 of the Copyright Act 1968 (the Act) for the purposes of research or study. The contents of the material may be subject to copyright protection under the Act. Further dealings by you with this material may be a copyright infringement. To determine whether such a communication would be an infringement, it is necessary to have regard to the criteria set out in Part 3, Division 3 of the Act. THE UNIVERSITY OF SYDNEY ArticleReach Citation information 20 June 2014 Journal, Chemical engineering (New York) Document ID: 958461 wu Article: SHORTCUTS FOR DISTILLATION DESIGN | Author: Frank, Otto ISSN: 0009-2460 6t ? ove HM vome 28 DS Issue 6 J \ Quarter. To: 205.227.91.137 Season = ner > TA Month: Ot Day: ot Year 1977 Patron's Library Information Address: University of Sydney, Fisher Library, FO3, ‘Sydney, NSW 2006, AUSTRALIA Pages: 110 -128 Emait: _illfish@library usyd edu.au Patron Note Fax. 011-61-2-9351-3689 Patron Information Staff Note: Patron Number: p1482676@9usyd Patron Type: ARLIR grad Paged _>>Gusyd<<, Qunew, 22006, ogeor, Smurd Patron Email; aani2456@uni sydney.edu.au Locations: Preference: Not Specified Holdings Information Location: Storage Call DST 10650 HOLDINGS: 53:8(1946)-68:2(1961),68:4(1961)-76:20(1969),76:23(1969)-78:2(197 1)-78:4(1971),78:6 (1971)-84:9(1977),84:11(1977)-93:20(1986),93:22(1986)-101:5(1994), 101:7(1994)-101:9(1994),101:11 (1994)-105:8(1998), 105: 10(1998)-1 14(2007) This article has been supplied by. The University of Sydney Library, Australia sydney.edu.aufibrary Ifyou have any enquiries or you would like assistance, please contact your library. | | For both batch and continuous distillations, the author offers shortcut techniques, as well as practical advice, for fixing the number of equilibrium stages, setting the operating conditions, sizing the column, and choosing the internals. * O Design of a distillation column progresses through four major steps: © Evaluation and regression of vapor-liquid equilib- rium (viz) data. © Calculation of the number of equilibrium stages. ® Determination of tray hydraulics. ™ Selection of tray or packing efficiencies OF these, only the second lends itself to a strictly theoretical approach. In contrast, thermodynamically Consistent equilibrium data are frequently difficult to come by, and their application to multicomponent systems can yield results of indifferent accuracy. Effi iency correlations disclosed in the open literature are Mill of a very rudimentary nature, with little advance- Tent recorded over the past 20 years. More often than ot, tray or column efficiencies reflect the uncertainties f the equilibrium data rather than the actual system characteristics or column configuration. Column hydraulics have been extensively investi- Sated. Reasonably good empirical data are available to describe the performance of the more popular types of trays and packing. These have been redefined in terms of acceptable operating ranges that are often no less curate than the published correlations. Presented below are selection guidelines and shortcut design procedures that will simplify the evaluation of Many standard distillation columns. The author has armarked cases where itis advisable not to cut comers, And where the development of a detailed, computerized “lution is justified. Where to use shortcut procedures The use of shorteu, hand calculation procedures uld be considered for several purposes: 1. Scoping studies suitable for preliminary costs. Parametric evaluation of operating variables. (ie: Separations having coarse purity requirements ‘e. contaminants >0.5 wt%). 4. Detailed designs for ideal and close-to-ideal sys- 3. Designs for systems for which equilibrium data are inavailable, eo the other hand, rigorous design procedures should applied if the following are true inl, The separation is a multicomponent one, requir- 8 Nigh product purity. a2: The system is highly nonideal but good equilib- Fur data are available : la i ltivevoatity between key component i u Otto Frank, Allied Chemical 4, One or more of the components is near the critical pressure. ‘A number of rules-of-thumb are helpful in judging whether or not mixture characteristics favor the use of simple vue relationships to design the column. These rules are listed in Table I according to the ideality of the mixture to which they refer. Selecting operating parameters Before any detailed distillation column design or evaluation can be undertaken, it is essential to define a number of operating parameters: ‘Feed composition—When the component ratio is large (5:1), variation in the feed composition can noticea- bly influence reflux ratio and number of trays. Itis less significant when the concentrations of key components in the feed are of about equal magnitude. Product purity—The specified concentration levels of high-boiling components in the distillate and low-boil- ing ones in the bottoms are the sole criteria for estab- lishing the number of trays and reflux ratio, The finer the split, the greater will be the number of stages or the required refiux ratio. Feed equilibrium—Whenever possible, the equilibrium relationship among system components should be es- tablished experimentally. The blind assumption of an ideal system is certainly not valid. In nonideal systems, deviation of activity coefficient from ideality is mostly positive. The relative volatility between components is thus curtailed at the top of the column, making it harder to obtain a pure overhead than assumption of ‘an ideal system would have predicted. ‘Thermal state of eed and reflux—Feed quality can have a noticeable effect on tray requirements. If the feed is subcooled below its bubble point, the number of trays in the rectifying section will decrease, whereas those in the stripping section will increase. More heat will be required in the reboiler and less cooling in the con- denser. The reverse is true for a feed containing vapor. Subcooled reflux will increase the molar ratio of liquid and vapor flows, and thereby increase internal reflux. The top tray will act as a partial condenser which, at the expense of efficiency, condenses vapor to reheat the external reflux to its equilibrium tempera ture. Usually there is no justification to subcool reflux before returning it to the column. Column pressure—Raising the column operating pres- sure will increase reboiler and condenser temperatures. auticle and several upcoming ones will be available in a reprint on Mt eet re of vapor-liquid mixtures poe ‘On the other hand, this change will also decrease vapor velocity, since the inerease in vapor density more than offsets @ corresponding temperature effect on volume. Often, relative volatility improves at lower pressures, making separations easier. In vacuum service, this effect, may be quite pronounced. ‘Sometimes the operating pressure may be dictated by separation requirements, making it necessary to operate outside the region of an azeotrope or below the critical pressure of a component. Column temperature—Column operating conditions are frequently selected to allow available plant utilities to be used as a heat source and heat sink. It is also impor- tant to select a reboiler temperature that is not so high that it will degrade the products. Given here are guide- lines regarding the approach temperatures of the heat sink and the heat source for different reboiler and con- denser services: Heat-sink approach, °C Refrigeration Cooling water Pressurized fluid Boiling water Air Heat-source approach, Process fluid 10-20 Steam 10-60 Hot oil 20-60 Energy utilization ‘The customary practice of adding heat into the bot- tom of a column and then abstracting an almost equal 110 quantity from the condenser at the top makes distill tion one of the highest energy consumers in a chemical or petrochemical plant. Energy conservation measures should be considered during the early stages of colum? design and layout. These measures include: @ Recuperative heat exchange between a cold feed and a hot bottoms stream. (The reboiler effluent ‘must be a large fraction of the feed.) Columns cascaded so that vapor from one is 60 densed in the reboiler of another. Insulation, An economic evaluation may justify insulation even if process temperatures are as 1o¥ as 50-60°C. Generation of low-pressure steam in condensers Reduction of refrigeration levels in low-temper®” ture separations by operating at the highest poss ble pressure. Recompression of overhead vapor to raise its et ‘ergy to a level where it can be used as a heat source for the reboiler, Worthwhile when the temperatut® drop across a column is small and a single com pression stage is sufficient to raise the vapor ef perature above that of the reboiler. Shortcut stage calculations ‘Two or more components in a liquid system can Pe separated by taking advantage of their different boiling points, In a boiling mixture, the coexisting vapor a liquid phases have different compositions; lower-boiling | constituents predominate in the vapor, whe! higher-boiling ones concentrate in the liquid. “The maximum concentration difference between the two phases is reached at equilibrium—assumed for 4 ideal stage. Since itis possible to calculate rationally number of successive equilibrium stages required fo" 2 tpecihed separation, it hat beoorme the practice #1 tablish first what degree of separation is theoretical possible and then to estimate how closely commer ‘equipment will approach this goal. a Graphical design procedures, as well as most short algebraic correlations [5, 10, 13, 14], usually deal Wi separations of binary mixtures. However, since few oe i binary systems are encountered in industry, calculatis procedures are usually applied to pseudo-binary ™y tures in which two principal components are design" as the light key and the heavy key. uct | ‘When it is necessary to produce high-purity proaug, streams, compounds boiling adjacent to each othel i, | the temperature scale become the key components. The ight key is the lowest boiling component present i? iy bottoms stream, whereas the heavy key is the his! boiling component present in the overhead. Wook kr bog beedcatica, wil ace te wey selected to yield a specified product mix. It is fy probable that the light and heavy keys do not adjacent to each other, but have an intermediate ing component between them—usually referred (0 distributed key. che Mixed-products are only occasionally specified ine chemical processing industries; consequent, the cent procedures outlined here will deal only wit keys Graphical stage calculations The simplest and most direct approach for analyzing inary distllations is still the graphical technique de- Vised by McCabe and Thiele in 1925, Despite its age, the xy diagram remains a highly useful tool for quick evaluation of a column (Fig. 1). The influence of reflux Tatios, product concentrations, and feed conditions can readily spotted by adjusting the slope and point of origin of the operating and q lines. The thermal condition of the feed is defined by the slope of the q line: Heat to convert 1 mole of feed to saturated vapor) ~~" Molar heat of vaporization Values of q that characterize a range of feed qualities are given here: Feed quality ‘ grline slope Saturated liquid 1 infinite Saturated vapor ° 20 Cold liquid feed > positive Superheated vapor 5:1). Although there will be few occasft for installing new bubble-cap columns, there are st! large ruanber of existing facilities that may require analSi= ‘A wide variety of bubble-cap shapes have bee? or ployed in columns, the majority of them being. oe round, bell-shaped type having vertical slots w Tye caper by W. Bll in {consi a proms resend fant Sol Popular types are inverted, rectangular boxes (tunnel ays), and inverted “teacups” without slots. Round caps are usually supplied in three sizes and fare selected according to column diameter: ‘Tower dia, ft Cap di 25-5 3 412 4 10 and up 6 J. and 2-in caps have sometimes been employed in smnall, low-temperature stills. Normal slot height varies between 0.25 and 1.0 in for 3- and 4-in caps, but may go as high as 1.5 in for 6-in aps, It is recommended that slots not be left open at the bottom, and that caps be equipped with a Strength-giving shroud ring to forestall deformation of the teeth. Experience indicates that cap configuration little effect on performance. Furthermore, there ppears to be no loss of efficiency if slots are omitted entirely, and all vapor passes under the lip of the cap. _ The critical dimension in bubble-cap design is the liquid seal—ie., the depth of liquid through which the Vapor must travel. Large slot seals enhance plate effi- iency at the expense of pressure drop, For design pur- Poses, a dynamic slot seal has been defined: Dynamic slot seal = hy, + hye + 4/2 (8) Where h,, is the static seal, in. liquid; hyg is the height of liquid over the weir, in; and \/2 is the average liquid Sradient, in (Fig. 9a). Practical limits for the dynamic Sct seal are’ given here: Pressure, psig Dynamic slot seal, in <0 05-15 0-50 10-25 50-200 prig 15-30 200-400 pig 20-40 ‘The weir height for bubble-cap trays is set by adding static seal to the slot height. In low-pressure and @tmospheric columns, the static seal usually runs be- ‘ween Y, and ¥, in. The pressure drop across a bubble- ap tray usually ranges between 2.5 and 5 in of liquid. Caps are generally arranged in an equilateral trian- Sular pattern, with rows normal to liquid flow. The “ps are usually spaced to have a pitch between 1.25 4nd 1.5, but never with less than 1-in clearance between Adjoining caps. Not all of the active area can be used for cap place- Ment. Stilling sections are customarily provided next to inlet downcomer and next to the overflow weir. is also considerable wasted area where, for struc iral reasons, it is impossible to place caps (Fig. 9b). A liquid gradient exists across trays between the point f liquid inflow and the overflow weir. This gradient Provides the necessary driving force to overcome fric- total resistance due to vapor flow, tray surface and ‘ubble caps. Because rows of caps offer a flow barrier ed much like rows of baffles, the gradient across a Pubble-cap tray can be quite pronounced. This is espe- Sally a concern in large-diameter columns (>5 ft for lerate and high-pressure systems; >7 ft for vacuum tems) and at high liquid rates. Too high a gradient = Stati seal ‘2. Important dimensions in cap design bs, Distribution of tray area Effect of excessive liquid gradient om tray stability Design and layout of bubble-cap tray should avert excessive liquid gradients SHRRTTCAT ENGINEERING MARGIT, TT 119 DISTILLATION DESIGN Pea et Fig. 10 mney J will induce blowing at one end of the tray, and dump- ing at the other (Fig, 9c). Sieve trays ‘The sieve tray is probably the most widely used contacting device found in columns today. It should be considered first in the design of new columns for several reasons: 'D Installed costs are lowest of all tray-type devices. 8 Design procedures .are well known, 9 Fouling tendency (with large holes) is low. © Capacity equals or exceeds that of other tray types. 5 Efficiency, with proper design, is good. jeve trays are not recommended whenever: 1 Pressure drop must be very low (<2.5 mm Hg/ tray). m= Turndown ratios are high (>3:1 at high pressure; >2:1 at low pressure). © Liquid rates are very low (<2 gal/ft of avg. flow idth). Ina sieve tray layout (Fig. 10), the active tray area can be defined as the entire tray deck from the inlet downcomer-skirt on one side, to the overflow weir on the other (i.c., the column cross-sectional area minus the area occupied by the two downcomers). Placement of holes in the active area is restricted only by the positioning of tray support rings and beams. Perforations can be made within 2 or 3 in. of the inlet downcomer or outlet weir. Holes are generally arranged in an equilateral triangular pattern, with rows normal quid flow. For optimum efficiency, pitch-to-diame- ter ratio of the hole should fall between 2 and 4.5. Hole sizes range from ¥, to 1 in, with Y,-in holes being the | Small holes (' and % in) lessen weeping if juid downflow has a high surface tension, and reduce entrainment in low-pressure systems. Large holes (9, and 1 in) should be used in fouling service. Pressure drop is lower in trays that are installed with perforations punched upwards to create a nozzle effect 120 GENIAL RGINEERING ARG 15 20 25 30 35 40 fig + oes in of liquid Weer Pe Rca accu in the direction of vapor flow. Despite this advantage tray panels are usually installed with downpunch holes in order to reduce the risks presented by jagged edges to personnel installing or inspecting tower inter nals Open hole area depends on vapor rate and the spec fied tray pressure-drop. In most cases, it ranges betwee” 4% and 16% of the active tray area. Because the rate vapor flow is inversely proportional to the square of pressure, large open areas are found in vacuum to¥” ers, whereas in pressurized towers open areas a smaller. ‘Outlet weir height on a sieve tray ranges between and 4 in, For low- and atmospher ; the height is usually 1 to 2 in. In where the mass flowing over the tray is more a va continuum than a liquid continuum, a weir actually se%* little purpose, It could be eliminated entirely, althous! ‘most engineers prefer to play safe by providing at a Lin, weir. Liquid gradients on sieve trays are considerably 1% pronounced than on bubble-cap trays. They ca ignored entirely in pressurized columns smaller tha? 8 fe dia, and in low-pressure columns less than 10 ft d® ‘The most critical variable in sieve tray design is ‘open hole area. Too small an area leads t0 @ pressure drop and, in extreme cases, jetting. An S@ 1 sively large open hole area encourages weeping even dumping, where no liquid passes over the *! tray weir. ee ‘There is considerable uncertainty regarding MOY much weepage is detrimental to column performane® has often been recommended that liquid flow throug) the perforations be kept at less than 25% of the 10% tray flow. However, many sieve tray columns opeF#ty apparently without undue harm to efficiency», Wi more than half the liquid weeping through the ee ‘A modified version of the correlation developed Fair [//] is given in Fig. 11. An operating point abo the line representing the desired ratio of open hole is the his to active tray area (Ay,/A,) is considered a safe design (<25% weeping). A point below the line may be in Some doubt but does not necessarily represent a dump- ing situation. In fact, experience has shown that if the ‘operating point falls anywhere above the curve repre- senting 6-8% open area, the column will most likely operate within acceptable efficiency limits. For a suitable sieve-tray design, hole velocity (caleu- lated for the vapor) can be estimated from the basic F factor relationship given in Eq. (6). Representative values of F for vacuum columns, atmospheric and ‘moderate-pressure columns, and pressurized columns are 11, 13, and 15, respectively Sieve-tray pressure drop combines flow resistance through the perforations with the hydrostatic head of aerated liquid on the tray: hy = hy + hy 9) Where fy = tray pressure drop, in. liquid; A, = dry Pressure drop across holes, in. liquid; and /, = aerated liquid head, in. liquid. Dry pressure drop across the holes can be obtained from the following well-known relationship: 1, = o0s2e(24) 49 Cy Where h, = dry pressure loss through holes, in. liquid; U, = hole velocity, fi/s (used in place of v in Eq. 6); and Cy = dry orifice coefficient (Fig. 12). ‘The pressure drop across the aerated liquid on the tay is obtained from the following relationship: hy = Bly + hoe) any Where h, = aerated liquid head, in. liquid; &yy = head at weir height, in. liquid; hyy = height, in, of liquid Over weir = 0.5 (Q/Ly)™; Q = liquid flow, gpm; 1, = weir length, in; and B = aeration factor (Fig. 13). Tn vacuum columns, where performance can be sig- hificantly compromised by entrainment, two-phase flow through the holes can raise the dry pressure drop be- Yond that for pure vapor flow. It is not unusual in low-pressure systems to observe a total pressure drop 15 to 25% higher than calculated. ‘The normal pressure-drop span for sieve trays is from 15 to 5 in. Hs. Outside these limits, trays may weep ‘Xcessively in vacuum columns, or jet in pressure serv- ice. In cither case, there could be an appreciable loss in “ficiency, Valve trays Arranged on a valve tray are liftable caps that act as Yariable orifices by adjusting themselves to changes in Yapor flow. This design (whose installed cost is 15-20% igher than for the equivalent sieve tray) is said to Provide turndown over a greater range than is possible for sieve trays. sult valves are actually small metal disks oF sips lat are lifted above openings in the deck as vapor Passes across the trays. The valve caps are restrained by legs or spiders, which limit vertical movement. Fabricators cite the following advantages for valve 1. A fairly constant pressure drop across a large por- tion of their operating range Dry orifice coefficient used in estimating dry pressure drop across sieve tray Froth density or aration factor (8) Pere ke ue nc eee rear ‘CEMair. Actual CEM co CaP Dry pressure drop across valve trays is Patan et ke au cee Crary Crm 121 Pie wr 1m bele weir height 5% of flooding) DISTILLATION DESIGN, rd 2. A high turndown ratio. 3. Operation at about the same capacity and effi ciency as sieve trays However, it should be noted that good turndows ratio, the most frequently advertised advantage of valve trays, may also be attained with sieve trays by imposing fa reasonable pressure drop. Even in vacuum columns where pressure drop can be critical, the available turm™ down ratio for sieve trays is adequate for most opera tions. ‘A mechanical problem often encountered with valve wear or corrosion of the retaining lugs or spiders. The constant movement of the valve caps imposes fatigue stresses that are aggravated by operation in a corrosive environment, It is not unusual to find valves missing 0” the bottom trays where high-boiling and corrosive com stituents tend to concentrate. Valve trays, because of their proprietary nature, are usually designed by the fabricator, although it is poss ble to estimate some of the design parameters from vendor literature. Tray performance can be predi from pressure drop charts (Fig. 14 and 15) adapt from the Koch Design Manual [6], whereas column and downcomer areas can be calculated by methods out lined above. ‘The number of caps that can be fitted on a tray is at best an estimate unless a detailed tray layout is pre pared. A 3-in x 2%-in cap pattern is the tightest at Fangement available and has become the standard fo" Jow- and moderate-pressure operations, For such a pat” tern, about 14 caps/ft? fit into the “net” capped are Active area as defined by Fig. 9b does not take int account stilling sections at the inlet and outlet, edge losses due to column support rings, and unavailabl space on top of support beams. It must be assumed that between 6 and 15% of the so-called active areas is not available for valves in large and small columns, resp tively. Tray efficiency (left) vs. column efficiency for the same separation 122 Downcomerless trays ‘The simplest possible tray design is the sieve tray without downcomers. Its successful operation demands tly high vapor velocity through the holes to maintain enough liquid on the tray for proper vapor- liquid contact. At low vapor-rates, downflowing liquid is not retained long enough on the tray, and mass transfer becomes quite inefficient since, as is the case in Spray columns, the only contact is between the rising Vapor and falling liquid droplets. ‘The major disadvantage of straight, downcomerless trays is that their operating range (turndown ratio) is considerably smaller than that of trays having down- ‘omers. Sufficient vapor flow is required to maintain a liquid level on the tray, but it cannot be so high that it will restrict liquid downflow and therefore flood the Column, At their optimum operating point, downcom- ‘less trays have about the same efficiency as sieve trays. However, efficiency will drop drastically as the tray dumps or the floodpoint is approached. Downcomerless trays may be considered when oper- ating conditions and capacity are not expected to vary; Solumn liquids have a high solids content; and ease of leaning is important. Few design procedures for downcomerless trays are Published in the open literature, and even the best of these is unlikely to achieve a degree of reliability greater 440%, The correlation developed by Sum-Shik et al, [/2] is cited here as a suitable procedure for estimat- ing an acceptable hole velocity: Uy = 1.15 fou te (Jar ety by hole velocity, ft/s liquid and vapor densities, 1b/ft° ‘open hole area, ft? = column free cross-sectional area, ft® Column efficiency, % 9} 605 01 Feedstock molar average viscosity at ‘average tower temperature, cP 0203 05 "10 7a BP, = ra Fae G, = orifice coefficient (0.62 for holes and slots) k= pressure loss factor through hole (1.8 for holes % to 7 in dia.) = pressure-drop ratio for vapor and liq- uid flowing through the holes “The column free cross-sectional area can be fixed by applying the same design criteria as were recommended for obtaining the free area in other tray columns. The hole area for a downcomerless tray will be 15-30% of the total area, with a 20-25% open area being consid- ered most effective. Variations of the basic perforated, downcomerless design include trays that have flat grating instead of perforations (Turbogrid); trays that use sinusoidally shaped, perforated plates (Ripple trays); and trays whose surfaces impart directional flow to the vapor (Kittle trays). In addition to the trays with and without down- comers described above, a number of other proprietary tray configurations are available for commercial col- umns (Table 11). ‘Tray and column efficiencies “Tray efficiency compares the separation actually at- tained to that which is possible in a true equilibrium stage. The Murphree efficiency is defined as: Efficien« ‘Vapor cone. from tray ~ Vapor cone. o ay Vara rion oi Bquld = Vopor cme ocay (19) ‘his efficiency differs from the overall column effi- 02 Relative volatility of key component X viscosity of feed ‘st average column conditions, cp Crd 8 - En Soo z Oe is 20-30 40 50 60 70 ‘% of flooding Tray efficiency is generally highest a between 40 and 85% of flooding ciency, which is the niffnber of theoretical equilibrium stages divided by the actual number of column trays. Overall column efficiencies are usually less than indi- vidual tay efficiencies, This is illustrated by the ‘McCabe-Thiele diagram in Fig. 16. The difference be- ‘comes most pronounced when a large number of trays are required to achieve a high product purity. Plate efficiencies are never constant throughout the column, since mass-transfer characteristics change with ‘composition, flowrate and temperature. Especially on the end-trays, where changes in concentration per tray are small, efficiencies drop off drastically as the concen- tration of the contaminating component falls below 100 ppm, Early investigators focused on the effects of physical properties and vapor-liquid equilibria on overall col- tum efficiencies. The best-known correlations are those of Drickamer and Bradford [2], who, on the basis of refinery column data, related efficiency to average liq- uid viscosity; and O'Connell [9], who also incorporated the effects of relative volatility ‘There presently is no truly satisfactory method for accurately calculating tray efficiencies. The only rational procedure available in the open literature is outlined the classic AIChE monograph on bubble-cap trays [J] Here, investigators combined the effects of physical properties, vapor-liquid equilibria, and tray hydraulics, in a diffusion and mass-transfer model that they hoped would represent the conditions of equimolar counter- diffusion through a film. For efficiency estimates, the simplest approach is often the best. The designer can pick off column effi- ‘ciencies from the charts given in Fig. 17, or can estimate them from the performance of similar columns already in operation. “Normal” column efficiencies fall between 60 and 85%. They tend toward the lower value.when: @ High product purity is specified, and especially when the contaminant concentration must be held 194 — aa SING ARGH TT substantially below 100 ppm. = Column performance is controlled by material balance, which is the case if one of the product components is only a small fraction (<10%) of the feed. = Low contaminant concentrations are specified fof both ends of the column. ‘Tray efficiencies can be drastically reduced by liquid entrainment at one end of the operating range ant excessive weeping at the other. Generalized curves it Fig. 16 illustrate that columns usually perform most efficiently between 40 and 85% of flooding. Packed columns Dumped or random packings are suited for the fol: lowing operating conditions: 1, Low-pressure-drop operations (vacuum columns) 2. High liquid/gas ratios (found more frequently it absorbers and scrubbers than in distillation cob ‘umns). 3, Corrosive environments (plastic or ceramic inter nals are required). 4, Small-diameter columns. Liquid distribution in packed towers Liquid distribution is one of the most critical items it the design of a packed tower. Unless the surface of the packing is adequately wetted, tower efficiency will dr0P drastically. For good irrigation, there must be at least four liquid-distribution points/ft®, and the liquid 1 should be kept above 2 gpm/ft? It is sometimes difficult to achieve the latter rate if vacuum distillation service; however, many packed ¢ tumns operate with liquid rates as low as 0.8 gpm, wi only a nominal loss in efficiency. Gravity distribution is the preferred method for i! troducing liquid onto the packing. Spray nozzles, whi! in theory assure a more-even distribution, can plug. ‘wear out over prolonged operation, and can be mé troublesome than gravity-feed devices. The liquid between the distributor and the top of the pact should be no greater than 12 in. A higher fall will t! to break the stream into droplets, and thereby inc liquid entrainment. Packing types 4 ‘Tower packings come in many shapes, materials 3% sizes; in commercial installations, commonly used include slotted rings and Intalox saddles. Depending the manufacturer, slotted rings are called Pall 8 (Norton), Ballast’ rings (Glitsch), Flexirings (Ko) Hy-Pak (Norton), or Cascade Mini-Rings (Mass Tra fer, Ltd,). Intalox saddles are only made by Norto™ Both rings and saddles can be manufactured fro™ wide variety of metals, plastics or ceramics. ng ‘The older types of packing such as Raschig rings Berl saddles have been almost completely superseded PY the more efficient slotted rings or Intalox saddles. OU high-capacity packings are more often found in se bers and in absorbers than in distillation columns. ‘To ensure proper liquid distribution and to minim wall effects, it is necessary to place a lower limit 0M fry 1g size. This ratio ratio of column size to pa ‘equal or exceed 30 for Raschig rings, 15 for saddles, and 10 for slotted rings. The 2-in slotted metal ring is gener- Ally found to be best from the standpoint of capacity. In fact, experience shows it to be the most economic pack- ing size and shape for most distillation applications. ‘Temperature and the corrosive nature of the column environment also influence the choice of packing. Some Beneral guidelines: 1. Metal packing should not be specified if the meas- ured corrosion rate is greater than 10 mils/yr 2, Plastic packing, while usually resistant to most chemicals, may embrittle with prolonged exposure. 3. If subjected to continuous heat, plastic packing should be fiberglass reinforced. 4. Ceramic packing is usually selected for hot, corro- sive service, although it is more fragile than metal or plastic packing. Shifting of the bed can cause the packing to fracture, restricting vapor flow and. boosting pressure drop. Design of packed beds ‘The generally accepted design procedure for sizing Packed columns [8] is the modified version of the Sher- Wood correlation (Fig. 19). ‘This correlation relates bed diameter to fluid densi- ties, owrates, and a characterization factor that has n measured for each type and size of packing: oer Where d = bed dia., ft; W = actual vapor flow, Ib/s; And G = vapor flow, Ib/(s)(f), as in Fig. 19 ‘The pressure drop across a bed increases with the Yapor and liquid rates. If these are raised simultane~ Susly (as is usually the case in distillation), at some Point a significant amount of fluid will be held up in the packing voids. The pressure drop across the bed focs up rapidly until downflow of liquid stops entirely flooding). Unfortunately, this floodpoint has not been defined well enough to be incorporated into Fig. 19. spending on the packing type, the floodpoint falls Rnwhere within a band of scattered points. For eer ngs, flooding takes place at a lower pressure trop (1.5 in, HjO/ft) than it does for some of the more ent packings such as slotted rings (2 10 2.5 i 10/11) pan most packed beds, the pressure drop will range ween 0.1 and 0.8 in/ft. Actually, for pressure drops °F0.5 100.6 in. H,O/R, all types of packings seem to be out equally efficient. Below these limits, the efficiency A. Raschig rings drops off drastically. Performance of the more efficient packings (eg, slotted rings and plas- tic Intalox saddles) does not deteriorate appreciably Mil the drop falls below 0.1 to 0.2 in. H,O/ft Selecting the right pressure drop for a packed bed {kPends mostly on its service. In general, it is desirable {yaPPly the highest pressure drop-consistent with relia- le and economic operation. For moderate- and high~ opeure distillation, the drop should be 0.4 to ot in. H,O/ft; for vacuum distillation, it should be 410 0.2 jn, H,O/t; and for absorbers and strippers 10 04 in, H,O/ft hhe accuracy of the curves in Fi a «ay 19 becomes some- what questionable at the extreme ends of the scale. “Thus in vacuum stills (equivalent to very low values of X; Fig. 19), actual measured pressure drops can easily be 25 to 30% higher than calculated. This discrepancy is attributed to high entrainment, which (as on trays) appreciably affects the flow behavior of the vapor. At high liquid rates, or at high pressures (when X > 3.5), the pressure drop in a functioning column may actually fall above the normally expected floodpoint limit ‘Although high-efficiency packings are designed to maintain reasonably good liquid distribution over the cross-section of a bed, there will always be some fraction of the liquid that reaches the column wall. Once there, the liquid will not readily redistribute itself back into the packing, but instead will bypass a large fraction of the vapor stream. Therefore, one should limit the verti cal height of a bed to ensure that the major fraction of, the liquid will remain on the packing during its passage through the bed. For Raschig rings, maximum bed height should be 2.5 to 3.0 times the bed diameter; for saddles, 5 to 8 times; and for slotted rings, 5 to 10 times. ‘Total bed height should not exceed 20 ft Liquid redistributors (wall wipers) are sometimes installed within a bed 10 bring liquid back into the packing. For slotted rings or saddles, installation of these devices produces little improvement in perform- ance if bed-height criteria” have been met. A notable exception occurs when a very high purity must be achieved at the bottom of the column, and even a trace ‘quantity of bypassed liquid can appreciably affect bot- toms concentration, Packing efficiency ‘Although numerous texts outline theoretical proce- dures for calculating packing height, there is no reliable method that can be universally applied to distillation systems. Instead, industry today uses the HETP concept to convert empirically the number of theoretical stages to packing height. ‘What makes this concept useful is that HETP values are remarkably constant for both organic and inorganic systems. Even with high-surface-tension liquids, good performance is possible so long as the packing wets properly (ie, li rates are kept above 1,000 Ib/(h)(ft2), and difficult-to-wet plastics such as fluorocarbons are avoided. In commercial columns, values of HETP for high- efficiency packings (slotted rings, Intalox saddles) run about 18 in for a t-in nominal packing size; 26 in for a 14-in size; and 35 in for a 2-in size. These values are about 6 to 12 in greater than published values derived from the operation of closely controlled pilot-plant columns (6. Because of reduced irrigation efficiency in vacuum columns, it is usually wise to add another 6 in to the listed HETP, Absorption systems generally exhibit HETPs in the range of 5 to 6 ft. For small columns (dia. <2 fy), an old rule-of-thumb proves surprisingly accurate if the appropriate packing size is used and the packing is properly loaded. HETP (15) For a particular type of packing, the effectiveness (HETP/AP) is fairly constant for all sizes. Little will be Column diameter 125 Generalized pressure drop correlation used for designing packed columns gained by replacing 2-in slotted rings with 1-in slotted rings to improve the HET? in a vacuum column, since the higher pressure drop will nearly cancel any reduc- tion in height. Systematically packed columns "These columns contain preformed sections having a large surface area for a given volume. In order to fune- tion as designated, the sections must be assembled within the column according to a prescribed arrange- ment. Since systematic packing configurations have not been standardized, no design procedures can be univer- sally applied to them. Engineering correlations are spe- tific for each proprietary packing and are usually sup- plied by their respective manufacturers. ‘Koch-Sulzer Packing—This packing consists of parallel layers of corrugated wire gauze arranged in a sloping pattern. Because liquid flow is controlled by capillary action, superficial liquid rates as low as 250 Ib/(h)(ft*) are attained without undue penalty to efficiency. ‘The packing effectiveness (HETP/AP) can be quite high at low pressures, making Koch-Sulzer packing a ‘good candidate for high-vacuum distillation, Another application for which the Koch-Sulzer arrangement should be considered is the replacement of existing trays, ‘or packing in vacuum columns in which its lower pres- sure drop permits raising the column pressure to en- hance system capacity. It must be recognized, however, that Koch-Sulzer packing is usually more expensive than any type of dumped packing. For estimation purposes, it can be assumed that the HETP for all sizes of commercial equipment lies be- tween 10 and 12 in, and that the column can be de- signed for an F factor (op,®) between 1.7 and 2.0. ‘Knit-mesh packing—These packings are offered by a number of fabricators, and tend to have similar con- struction and substantially the same performance char- acteristics. Pressure drop is very low, but efficiency is, not quite as high as that of Koch-Sulzer packing. HETP 126 is somewhat affected by column dia., ranging from 6 it in small columns to 24 in. in larger columns. Commercially available mesh packings includ Goodloe, Multifil and Hyperfl. Like Koch-Sulzet packing, mesh packing is woven from 6 to 7 mil wie land requires the use of materials of construction that will not corrode under column operating conditions Koch Flexipac—Flexipac is similar in construction © Koch-Sulzer packing, but instead of wire mesh use corrugated metal sheets. Its performance falls some where between that of slotted rings and Koch-Sula packing. ‘Kloss and Neo-Kloss packing —These types are suitable for very low pressure operations where separating ef ciency is not critical. Since there is no obstruction in vertical direction (the packings are constructed, tively, with an assembly of coiled springs, or with OF centric cylinders), the pressure drop is extremely 1% ‘Their HETPs range from 2 10 4 ft Glitsch grid—This is an assemblage of stamped pane (having a high open-area) that are stacked on {OP 9% each other. The primary advantages of the Glitsch gt are its low pressure drop and high capacity. It is bet cd for use in strippers and absorbers where ligt rates are high than in distillation columns where HETP tends toward 6 ft Batch Distillation Batch still, once quite numerous, are seldom const ered today because of the considerable labor and atte tion that must be lavished on their operation. ‘There 1 still, however, situations where the choice of bateh 4 tillation is justified and favorable economics ©" 1. demonstrated. These include: small production 124 (usually less than 1 million tb/yr); widely varying IU conditions and product requirements; irregular © equipment; multiproduct separations; successive duction runs with different processes. Batch stills are mainly used today in pharmaceutical, fine-chemical, dye, cosmetic and liquor processes where the frequent practice is to process a variety of products in relatively small batches, Single-stage distillation ‘The simplest example of batch distillation is single- Stage, differential distillation, beginning with an ini- tially full stillpot heated at a constant rate, The compo- sition of the vapor leaving the kettle changes continu- ously but is always in equilibrium with the remaining liquid, Liquid not vaporized is removed as bottoms Product at the end of the run In 1902, Rayleigh developed an expression that re- lates distillate and bottoms composition of the single- age batch still to the fraction of the initial charge that has been vaporized. If the relative volatility is constant, or can be averaged, the following relationship app! (2) aiy(n + an) (16) initial iquid in reboiler, moles; Ly = final tesidual liquid, moles; x;,% = initial, final mole frac tions of the more-volatile component in reboiler; y Vapor mole fraction; and a = relative volatility Since a simple batch still only provides a single theo- retical stage, it is impossible to obtain a complete sepa- tation of a pure component unless the relative volatility is infinite. Therefore its application is usually restricted to preliminary recoveries that are later followed by a ‘nore rigorous distillation; to production where high Purities are not required; and to processing of casy- to-scparate mixtures, such as the removal of low boilers r residue from a product. Distillation with rectification ‘To achieve a reasonable separation and to obtain Product cuts of a desired purity in a batch still, it is Necessary to rectify the vapors from the stllpot in a tray °F a packed column, Usually such a column does not ©ontain many trays, since a batch still consists solely of ‘rectifying section. As shown in Fig. 20, the system will Pinch, regardless of the number stages, once the low boiler concentration in the bottom approaches the in- lersection of the operating and equilibrium lines. As long as the concentration of the more-volatile ‘omponent in the reboiler stays reasonably high, distill- ‘te purity will remain reasonably constant. However, as the component becomes depleted, overhead concentra- on drops rapidly. _ When reflux ratio is held constant during the separa- tion of two key components, the overhead composition Will change continuously, and final distillate concentra- ion will be the average of the entire cut collected in the educt receiver. ‘Too high a takeoff rate (low reflux atio) will speed collection of product, but at the ex- Penseof product quality. Too low a product withdrawal Tate (high reflux ratio) will result in maximum recovery Hon spec” product, but at an unacceptably long th cycle. fo maintain a constant overhead composition, it is sary that the reflux ratio be reduced continuously is delays onset of a col- umn pinch, but does so at the expense of a longer distillation time. The product run is considered com- plete when a further increase in reflux ratio will reduce forward flow to such a low level that batch cycle time will be extended beyond acceptable economic limits. "The customary arrangement for operating batch col- ‘umns is to combine constant and variable reflux ratios. Usually two or three different reflux ratios are applied to each overhead cut. It has been found that for a given number of stages, variations of reflux ratios and collect- ing times have litde influence on the total capacity of a still if they are within reasonable limits established for operation of the still at either constant-reflux or con- stant-overhead composition, ‘Considerable effort has been expended by investiga- tors to fix the effect of liquid holdup on separation efficiency in batch stills. For many years, any holdup in fa column was considered detrimental. More-recent ‘studies indicate that a limited amount of liquid in the column may actually optimize column capacity. “The existence of a finite holdup creates a “flywheel effect,” which causes the composition to change more slowly than “instantaneous” equilibrium conditions tend to indicate. Thus, overhead purity will stay high longer than theoretical calculations might show, especially when the reflux ratio is relatively low (perhaps <8) and the content of the still is turned over rapidly. The flywheel effect is less pronounced when the reflux ratio is high, a condition frequently encountered when high-purity cuts are required. Literature references seem to indicate that a column holdup of up to 10% may actually improve batch-still performance [3,7]. Above 15%, holdup appears to have ‘a negative effect and column capacity is reduced. Ca- pacity can be further optimized by reducing the exter- nal holdup of distillate. The elimination of reflux drums by inline flow-splitters will minimize equilibration time and thereby permit rapid concentration changes be- tween cuts Reboiler ‘concentration ‘with 7 stages coneentration Pee acme crossing of operating and equilibrium lines aeRO aT 127 3 A 3 6 10 15 20 25 90 35 Number of theoretical stages or reflux ratio Estimate of theoretical stages and reflux ratio for a batch distillation ‘These rules-of-thumb apply to batch-still operation: 1. Once a column hi been installed, capacity for a given product specification is only minimally af- fected by changes to reflux ratio and to the length of a product cut. Asthemore-volatilecomponentisremoved from the eboiler, separation becomes progressively more difficult . Too low a reflux ratio cannot meet the product speci fication no matter how many trays are installed. vis impossible to recover in a single operation at high purity a low-boiling component that repre- sents only a small fraction of the initial charge For optimum separation capacity, minimize or climinate reflux holdup. Design for a liquid holdup in the column equivalent to 10 to 15% of the initial batch charge. Precise design of a batch distillation system can be extremely complicated because of the transient behav- ior of the column. Not only do compositions change continuously during rectification of an individual charge, but successive batches may start from varying compositions as “slop” cuts and “heels” are recycled. ‘The literature dealing with batch stills abounds in jures for hand and computer calculations. These have been well summarized in most standard distilla- tion texts, and their application together with liberal doses of experience will generally produce an acceptable column design. ‘The number of equilibrium stages required for a given batch distillation can be estimated directly from Fig. 21. Whereas this estimate may not result in an optimum configuration in every case, it will usually provide an acceptable design for most systems. The graph empirically represents, for a number of commer- ial stills, performances that have been correlated from the basic Fenske equation. "This curve is considered suitable for separations re- quiring a product purity in the range of 300 10 | 128 8,000 ppm (0.03 to 0.8 wt%). If more than a single pure overhead product is to be recovered, the number of theoretical stages are determined by the smallest rela~ tive volatility between any two of the major system components, ‘The graph also contains an empirically developed curve for reflux ratios, which permits selection of the highest reflux ratio at which the column will have © ‘operate any time during the batch cycle. Such a value useful for preliminary selection of a column diamete™ and estimation of cut time. Efficiency of batch columns may vary widely during ‘a single cut as the concentration profile in the columm shifts over a wide range. For long intervals, operation may take place under pinched conditions where effi ciencies as low as 30% have been measured. Without experimental data, a reasonable value to expect for al overall batch column-efficiency is from 50 to 60% For packed columns, height will depend on types of internals used and column diameter. The following guidelines are recommended for converting chart value to packing height. Size and type of packing >2in dumped (slotted rings or saddles) 1.5-in dumped (slotted rings or saddles) <10-in dumped (slotted rings or saddles) Mesh packing (<1 Mesh packing (1 References 1. AICHE Bubble Tray Design Manual, 1958 2. Drickamer, HLH, and Bradford, J Ry Tia AFCA, Vol 38,1935 Eich, FG, Eficency Calculation fr Binary Batch Recifcations Holdup, Chom. Eng, July 8, 1900, p13 3 4. Glitch, nc. Bulletin 34900, Se, 1974 5, King, C.J, "Separation Process,” McGraw-Hill, New York, 197! 6 Koch Flextry Design Manual, Bulletin 960, 1960, 7 Layben, W.L, Some Practical Aspects of Optimal Batch Disilat Dasigy Unt Eig Chen. Pc Des. Dp, WoL 10, No, I, 197 8, Norton Co, "Packed Towers 9, OContel, H.E. Time ATONE Vol 42, 1946, p. 74 ie 10, Perry and Chilton, C. H., “Chemical Engineers’ Handbook,” Set" ‘McGraw-Hill, New York, Sth ed 1973. a 11, Smith, B.D, "Design of Equilibrium Stage Process,” McCraw Hh York, i963 12, Sum Shik, others, Chin. Pom. 13, 968, p66 15, Tebal Es eMassTraafer Operations 2nd ea, MeGiraw Hil, NS York 156 Van Winkle, M., *Distiation,” McGraw: Hil, New York, 1967: Van Winkle. M. and Tod, Gham Eng, Sept 20, 1971, p. 136 The author ‘Otto Frank is Supervisor of Proce, Design and Ntethods a Allied Cems Corp bere he Seven for he soos Foie ial i a sista Oger he

You might also like