ceenews2_04

Upload: antonio-mezzoprete

Post on 04-Jun-2018

223 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/14/2019 CEeNews2_04

    1/6

    D

    uring the initial startup of anew extractive distillationtower* at a Sasol plant in Se-cunda, South Africa, the

    chimney tray feeding liquid to thetowers once-through kettle reboilerunexpectedly overflowed. The plantquickly implemented a successful fix,but the cause of the overflow remainedobscure. Further joint troubleshootingby a Sasol/Fluor team cleared up themystery, and provided lessons for howto avoid similar problems in other ex-tractive distillation units.

    The investigation utilized pressuredrop measurements, neutron back-scatter, surface temperature surveys,

    and hydraulic calculations to estab-lish the force balance that led to theoverflow. Among other things, themeasurements offered strong evi-dence of boiling maldistribution in thereboiler. The team came up with atheory that is consistent with all themeasurements and force balance; itsubdivides the reboiler into a stag-nant region, an intense boiling region,and a kettle region.

    This regional maldistribution pat-tern is believed to be unique to mix-tures such as those encountered in ex-

    tractive distillation, with the boilingliquid consisting of a large concentra-tion of very high boilers and a smallconcentration of volatile components.The mulitifeed and multidraw kettlearrangement, and possible foam for-mation, may also play a role.

    Process descriptionThe process feed to the tower (Figure1) contains: 93% Component A, an or-ganic with an atmospheric boiling

    point of 80C, which is recovered as ahigh-purity overhead product from thetop of the tower; and 7% Component B,an organic that boils at 84C under at-mospheric pressure. B has a higheraffinity to the solvent and leaves at thetower bottom. The solvent boils atabove 250C at atmospheric pressure.

    In the unusual boilup supply sys-tem for the tower (Figure 1), liquid at108C from the bottom trays is col-lected on an upper chimney tray, fromwhere it gravity-flows to a once-through thermosiphon side-reboiler.The liquid from the outlet of the ther-mosiphon is collected on a lower chim-

    ney tray at 109C, and gravity-flows tothe kettle reboiler. Vapor generated inthe kettle reboiler at 140C is re-turned to the tower above the lowerchimney tray. Unboiled liquid fromthe kettle reboiler, also at 140C, over-flows the reboiler weir into the kettledraw compartment, from where itgravity-flows into the tower bottom-product sump.

    Table 1 shows the rapid change incomponent concentration that takes

    place near the bottom of the tower.The tower bottoms contain 2% byweight of Component A, 1% of Compo-nent B, and 97% solvent (ComponentC). One stage up, at the bottom chim-ney tray, the liquid compositionchanges to 30 wt.% A, 14 wt.% B, andonly 56 wt.% solvent. This differenceis the reason for the steep tempera-ture change near the tower bottom.Such steep composition and tempera-ture changes are typical of extractivedistillations, using a high-boiling sol-

    vent to separate relatively low-boilingorganics.

    (Continues on p. 56)

    CHEMICAL ENGINEERING WWW.CHE.COM FEBRUARY 2004 55

    Operations & Maintenance

    *The tower was not designed by the authors em-ployers.

    109C

    #3140C

    #1 109C

    #2 140C

    Thermosiphonreboiler

    Kettlereboiler

    Cleansolvent

    CW

    A

    Contaminated solvent2% A 1% B 97% solvent

    Steam

    PinH

    108C

    Feed

    93% A7% B

    Pout hs, REB

    1 5

    4

    2

    3

    Distillation

    Reboiler StartupCan Pose ChallengesFuture designs can benefit from the lessons

    that were learned via a well-planned

    investigation of this startup problem

    Etienne Rubbers, Kirsten GreenandTerry Fowler,Sasol Technology (Pty.) Ltd.

    Henry Z. Kister and Walter J. Stupin,Fluor

    FIGURE 1.

    The extractivedistillation tower

    discussed in this articleemploys two reboilers

    TABLE 1. DESIGN STREAM COMPOSITIONS

    Stream #1 Stream #2 Stream #3

    Liquid toFlashed Flashed

    Liquid Vaporkettle ex kettle ex kettle

    Temperature,C 109 124 137 140 140

    A, wt, % 30 3 2 2 62

    B, wt, % 14 2 1 1 29

    C, % wt 56 95 97 97 9

    Vaporization, mol% 69.0 71.7 72.3

    Mass flow, kg/h 134,000 72,000 62,000

  • 8/14/2019 CEeNews2_04

    2/6

    The startup problemUpon startup, the lower chimney trayoperated at the design temperature of109C. The kettle reboiler vapor was at

    145C, somewhat hotter than the de-sign 140C. However, the bottom sumptemperature was 109C, much colderthan the design temperature of 140C.This low temperature meant a muchlarger concentration of A (around 30wt.%) in the sump, compared to the de-sign (around 2 wt.%). Because of thediscrepancy, the recovery of A in theproduct stream was much poorer thanthe design recovery. Whats more,flooding was apparent in the solventstripper downstream of the extractive

    distillation tower, due to the significantvariation in its feed composition.

    Operators first tried to improve perfor-mance by manipulating some of thetower operating variables. Raising thesteam rate as far as possible only had asmall effect, raising the sump tempera-ture from 109 to 112C. Reducing the feedflowrate had a much greater effect, rais-ing the sump temperature as high as130C, but at the penalty of lower pro-duction rates. The situation was not tol-erable, and changing operating variables

    did not produce a satisfactory solution.The higher reboiler temperature

    suggests that the flowrate through thekettle reboiler was lower than as de-signed. The only plausible explana-tion for the low sump temperature isthat the liquid from the lower chim-ney tray was bypassing the kettle re-boiler and reaching the bottom sump.Such bypassing can be due either toliquid leaking from the lower chimneytray, or to liquid overflowing into thechimneys. The observation that the

    sump temperature rose at lower feedrates argued against tray leakagebeing the root cause, thus supportingliquid overflow into the chimneys.

    There was another possible routefor liquid to bypass the lower chimneytray. Vapor-liquid effluent from thethermosiphon side reboiler enters the

    vapor space above the lower chimneytray (Figure 1). Though the bottom ofthis nozzle is 300 mm above the hat ofthe chimneys, it was possible that liq-uid trajectories from the side reboilereffluent were finding their way into achimney that had its opening orientedtowards the effluent nozzle.

    Based on the overflow theory, thetower was shut down after one day inservice, and the chimney heights onthe lower chimney tray were extendedfrom 200 mm to 900 mm. Furthermore,

    the chimney into which the side re-boiler could have been blowing liquidwas blanked. Upon return of the sys-tem to service, the sump temperaturereached the design 140C and stayedthere during all periods of normal oper-ation. The reboiler temperature de-clined from 145C to the design 140C.The chimney tray modifications weresuccessful in preventing any furtherliquid bypassing from the lower chim-ney tray into the bottom sump.

    Taking the pulseAlthough the problem thus wentaway, understanding its root causeposed an interesting challenge. Fur-thermore, such understanding wouldbe valuable, both for dealing with sim-ilar systems and for application dur-ing any future debottlenecking of thesame Sasol distillation unit. So, thefollowing checks and measurementswere performed:Pressures: The team measured pres-sures at the vapor space above the ket-

    tle reboiler and in the vapor above thelower chimney tray in the tower. Themeasured pressure drop at startup was48 kPa, higher than the design 2.5-kPa maximum, even though theflowrates were lower. The outlet-linepressure drop was recalculated, usingthree alternative procedures: a Fluormethod, a commercial simulation, anda simple hand method based on PerrysHandbook [2]. All three methods gavenumbers in the 46-kPa range, whichwere slightly lower than, yet well inline with, the field measurementsLiquid level: Neutron backscatterscans were performed to determine

    the liquid level on the lower chimneytray. The level was found at 300 mm.This finding confirmed that, with theinitial 200-mm chimneys, the liquidlevel would have been above the top of

    the chimneys and overflow wouldhave occurred. It also confirmed thatincreasing the chimney height to 900mm would have indeed eliminated theoverflow. Thus, this test fully con-firmed the startup teams analysisand solution.

    The neutron backscatter scans alsoargue against the liquid trajectoriesfrom the side reboiler causing the by-passing. With the trajectories fullyeliminated during the startup fix, thechimney tray liquid level still exceeded

    the original chimney height of 200 mm.Entrainment: Entrainment of liquidin the vapor leaving kettle reboilershas been known to occur when the

    vapor space above the tube bundle issmall [1]. To explore this possibility,the team assessed the design of theexisting kettle in light of good-designcriteria listed in the literature [1]. Theratio of kettle diameter to overflowbaffle height in the existing kettle is1.4, which is near the recommendedminimum (1.3 to 1.6). The ratio of tube

    bundle to kettle diameter is 0.63,which slightly exceeds the recom-mended maximum 0.6.Foaming: The team conducted somelimited tests to check for foaming. Thetray-pressure drop across the columndid not provide any evidence of foam-ing on the trays. So, if foaming tookplace, it would be confined to the kettle,where the solvent concentration washigh and where vigorous boiling takesplace. A sample of reboiler liquid wasshaken, and was observed to produceabout 25 mm of froth above the liquid,which took 12 min. to disappear.

    Because of the high fluid tempera-

    56 CHEMICAL ENGINEERING WWW.CHE.COM FEBRUARY 2004

    Operations & Maintenance

    FIGURE 2. Temperature measurements made during troubleshooting of the kettlereboiler constituted a key input for the analysis of its performance problem

  • 8/14/2019 CEeNews2_04

    3/6 CHEMICAL ENGINEERING WWW.CHE.COM FEBRUARY 2004 57

    tures, it was not possible to simulatethe operating conditions. So foamingcould have played a role, even thoughno stable foam was observed. Older-

    shaw tests [1] during the piloting of thecolumn indicated some wall-supportedfoaming, but the foam was unstable.Temperatures:Laser-guided infraredpyrometers measured the surface tem-peratures around the kettle reboilershell. Several sets of these measure-ments were made. While the tempera-ture varied from one set of readings toanother due to changes in steam pres-sure, the trends observed did not varyand were completely repeatable. Figure2 shows the results of the most compre-

    hensive temperature survey.The kettle reboiler has four liquid

    inlets and four vapor outlets, a con-figuration that is often used to providegood fluid distribution in horizontalshell-side reboilers. The average sur-face temperatures at the two vaporoutlet nozzles furthest away fromthe overflow baffle were low (122and 128C). The surface temperatureat the next vapor-outlet nozzle was141C, and at the vapor nozzle closestto the overflow baffle, the temperature

    was 144C.The surface temperatures at the

    reboiler liquid pool displayed the op-posite trend. Temperatures were high(176 and 186C) near the flange, andwere progressively lower along thefinal half of the reboiler length. Thesurface temperatures at the liquidpool were 160168C in the thirdquadrant away from the overflow baf-fle, 157160C in the second quadrantaway from it, and 154C at the quad-rant just before the overflow baffle.

    Heat transfer: The heat transfercoefficient was calculated to be 839kW/(m2)(C), which is lower than thedesign 1,026 kW/(m2)(C).

    Assessing reboiler operationThe measurements and determina-tions made as outlined above enabledthe team to assess the operationwithin the reboiler.Pressure balance: The pressure bal-ance between the elevation at Point1 at the vapor space above the lowerchimney tray and that at Point 2 atthe reboiler floor (see Figure 1) is asfollows (with all pressures expressed

    in millimeters of liquid head):

    H =Pin+ hs,REB+ Pout (1)

    where His the liquid driving

    head (top of chimney tray liquid levelto reboiler floor), Pinis the pressuredrop in the inlet piping, hs,REBis thestatic head in the reboiler and Pout isthe pressure drop in outlet piping.

    The neutron scan measurementsgaveHas 15.5 kPa at normal operat-ing conditions. The Pinwas calculatedat about 1 kPa. The Pout at normaloperating conditions was calculated at57 kPa; this range was slightly higherthan the startup Pout, which was 46kPa and agreed with the plant data at

    its original startup. With the assump-tion of no aeration in the reboiler liq-uid, and allowing for the head over theweir as recommended [3], hs,REB wascalculated at about 8 kPa. The rightside of Equation (1) thus gives 14-16kPa, which is slightly on the low side(but well within calculation accuracy)of the 15.5 kPa head on the left side ofEquation (1).

    Basic to this force balance is the as-sumption that the static head insidethe reboiler,hs,REB, is equal to the head

    of actual liquid from the top of theoverflow baffle (plus the small headover the weir) to the reboiler floor. Thisassumption makes hs,REB the head ofnon-aerated liquid, which may be ques-tionable for a pool of boiling liquid.

    This assumption is made in somekey literature references on the sub-

    ject [3, 4], and is often the conserva-tive assumption made for design.More-comprehensive models [5, 6]replace the non-aeration assumptionby a model that takes into account the

    generated vapor volume, as well as thetube-bundle two-phase pressure drop.

    Application of these models to kettlereboilers is further complicated by therecirculation of liquid inside the kettleand the existence of fairly clear liq-uid zones between the bundle and thesides of a kettle reboiler [5].

    In the Sasol reboiler, there is an ad-ditional justification for the non-aera-tion assumption. As is discussed below,there is evidence that the nozzle clos-est to the tubesheet enters a stagnantliquid region. In that case, this inletwill see the full hydrostatic head be-tween the overflow (Point 3) and the

    reboiler floor (Point 2) in Figure 1.Based on the Equation (1) force bal-

    ance, the chimney tray overflowed be-cause the pressure drop in the reboiler

    outlet line was high and because thefull static head of the non-aerated re-boiler liquid acted to raise the drivinghead on the chimney tray. Entrain-ment from the kettle could have alsocontributed. This entrainment wouldfurther increase the pressure drop inthe reboiler vapor outlet line, and ac-count for the outlet line pressure-dropmeasurement being slightly higherthan calculated. Calculations showthat about 10-30% entrainment inthe vapor would account for the slight

    pressure drop difference. However,this difference can also be explainedby simple inaccuracies in measure-ment and calculation.

    Although the pressure balance ex-plains the chimney tray overflow, itis unable to explain the observationsfrom our temperature measurements.These are discussed below.Vapor disengagement betweentubesheet and first vapor outlet noz-

    zle: The tubes are 6,100 mm long (thismeasurement does not appear in Figure

    2). Of this length, about 1,500 mm (or25%) lies between the tubesheet and thebeginning of the first vapor outlet noz-zle. So if the heat exchanger were to va-porize the liquid uniformly, practicallyall the vapor exiting at the first vaporoutlet nozzle should initiate in the ex-changer section between the tubesheetand the beginning of that nozzle.

    The kettle overflow baffle is 1,150mm high. A Francis weir formula cal-culation [2] shows that a 72,000-kg/hliquid overflow over the outlet weir will

    incur an additional liquid head of 50mm. So, the liquid level is at least 1,200mm above the kettle floor (without al-lowance for hydraulic gradients or foraeration of the liquid, which will tend tomake the liquid height even taller).

    The first 236 mm of tubes are in the1,000-mm-dia portion of the kettle,with no vapor disengagement space.Over the next 1,090 mm of the tubes,the kettle diameter expands from1,000 to 1,628 mm, at a 30-deg angle.Geometry dictates that the first 350mm of that expansion will be fullysubmerged by the 1,200 mm of liquid.So the total tube length at which there

  • 8/14/2019 CEeNews2_04

    4/6

    is no vapor disengagement whatso-ever is about 590 mm. In this length,40-50% of the vapor reaching the first

    vapor nozzle should be generated.

    Further, in light of the expected vaporgeneration, vapor velocities should belocally high. And, there will not bemuch disengagement of liquid in thenext 500 mm or so of tube length, asthis section has a restricted vapor-passage area due to the smaller diam-eter. (However, as we have concludedbelow, the actual vapor generation inthis section is very low.)Stagnant liquid region (Figure 3):

    For the liquid in the tubesheet region,the surface temperatures measured

    were high. This is also the region thathas little disengagement space above.Liquid temperatures in the range of168 to 186C imply concentration ofComponents A and B of less than 0.5%by weight; in other words, very low.

    Under the assumption that thesetemperatures apply to the mixtureclose to the tubesheet throughout thetube bundle, the material in this re-gion thus consists almost entirely ofthe non-volatile solvent, ComponentC. It then follows that in this region,

    very little boiling and heat transfertake place. This absence will occur ifthere is only a small flux of materialflowing around the tubes in this area.

    The small flow generates only asmall quantity of vapor, and the liquidaround the tubes will get hotter andhotter as its low-boiling componentsflash off. All that will be left is theheavy liquid, which tends to sit there.This behavior turns most of the firstquadrant into a stagnant zone or deadpocket of hot, heavy liquid, mostly

    Component C.It is worth noting that temperatures

    of 168 to 186C were surface measure-ments at the shell, some distance awayfrom the hot tubes. The stagnant layerbetween the tubes and shell is a goodinsulator, and it is possible that nearthe tubes the temperatures will behigher, probably approaching that ofthe condensing steam.

    Since not much liquid vaporizes inthe stagnant zone, the froth density inthis region is high, much closer to thatof the liquid than to that of a vapor/liq-uid mix, and much higher than that ofthe froth or aerated liquid in the other

    quadrants of the kettle. This meansthat there is less flow resistance, dueto the lower static head of the froth inthe central and baffle regions. Conse-quently more incoming liquid willchannel toward the central region.

    Our calculations indicate only a fewhundred millimeters of head differ-ence are enough to account for the es-sential absence of flow in one feed noz-zle while the adjacent nozzle feeds thebulk of the liquid to the reboiler. Asless feed flows towards the tubesheetend, the liquid in the tubesheet regionwill stagnate more and vaporize less.The end result is a self-acceleratingmechanism that promotes the deadzone in the tubesheet region and ahigh-liquid-flux central region.

    The high-L/V, intense boiling, cen-tral region (Figure 3): This regionhas a higher-than-design liquid fee-drate (and a high liquid/vapor, or L/V,ratio), to compensate for the reducedflow to the stagnant region. The highrate results in lower liquid tempera-tures, but with intense boiling andhigher than design vapor generation.

    A pressure balance across the re-boiler (Figure 1) equates the pressuredifference between the common liquidinlet line (Point 4) and the common re-

    boiler vapor space (Point 3), no matterwhich region one travels through. Thepressure difference is given by Equa-tion (2), again with all pressures ex-pressed as millimeters of liquid head:

    P4 - P3 = hs + Pn + PREB (2)

    where hs is the static head in a givenreboiler region, Pn is the pressuredrop in the liquid line leading to theregion, and PREB is the reboiler pres-sure drop in that region.

    In the stagnant region discussedabove, hs approaches the full head of thenon-aerated liquid, and thePterms arequite small due to the low flow. Con-

    versely, in the intense boiling region, theboiling aerates the liquid, which lowersthe liquid head. This induces highflowrates of liquid into the region untilthe pressure-drop terms balance the dif-ference between the aerated-liquid head

    in this region and the non-aerated-liquidhead in the stagnant region.

    The high liquid flowrate inducedinto the vigorous-boiling central re-gion means that the fraction of the liq-uid vaporized in this region is lower,so less of the low-boiling componentsare removed from the liquid. Thislights-rich liquid could possibly per-sist all the way to the two vapor noz-zles closest to the tube sheet.

    The lower temperatures observednear the vapor nozzles closer to the tube

    sheet are the result of the high liquidflux in the tube field and the resultinglower temperatures through the re-gions providing the vapor to these noz-zles. The vapor nozzle closest to thetubesheet is the coldest, because that iswhere the intense boiling starts and ismore vigorous, and there is a smallerfraction vaporized. Also, some of the liq-uid is likely to be projected into the

    vapor outlet nozzles in the form of en-trainment. The action in this region hassome similarities to the action in hori-

    zontal thermosiphon reboilers, in thatthe quantity of vapor generated resultsin a low-density froth, which promotesliquid flow to this region, in contrast tothe behavior with the more dense liquidin the more stagnant zones.

    One might wonder what causedthe measured wall temperatures ofthe liquid in this region to remainhot. The key is that our survey mea-sures the wall temperatures, not thebulk temperatures. The intense boil-ing takes place at the bulk of thebundle, and may not reach the wall.Equation (2) shows that along thewall in this region, there is horizon-

    58 CHEMICAL ENGINEERING WWW.CHE.COM FEBRUARY 2004

    Operations & Maintenance

    FIGURE 3. Boiling behavior varied significantly across regions of the kettle reboiler

  • 8/14/2019 CEeNews2_04

    5/6 CHEMICAL ENGINEERING WWW.CHE.COM FEBRUARY 2004 59

    tal flow of hot liquid from the higher-static-head stagnant region towardsthe overflow baffle. This flow is some-what tempered by the boiling. On the

    other hand, the vapor temperatureswere measured right above the bulkof the bundle, where the boiling ismost intense.The kettle boiling region (Figure

    3):There is a net flow of liquid fromthe central region to the overflowbaffle, because this is the only mannerin which the solvent can permanentlyexit the kettle reboiler system. As onemoves closer to the overflow baffle,there are two effects: this net flowincreases, and the composition of the

    liquid can be expected to contain lessof low-boiling material. As a result,less vigorous boiling and a higher aer-ated-liquid density can be expected inthis region. This higher aerated-liquiddensity will provide more resistanceto liquid feed from the inlet nozzle inthis region and, in turn, reduce theinlet feed to the kettle boiling region.

    This is the one region of the reboilerthat actually operates as it should,like a kettle. Not much entrainmenttakes place here, and there is much

    disengagement space in this region,both of which promote the true kettleaction. As the reboiler-feed liquid flowthrough this region is lower than thedesign and the percent vaporization isclose to or higher than design, the tem-perature of vapor flowing to the lastnozzles is close to or above design.

    If liquid entrainment from the in-tense boiling region is significant, itwill generate a higher pressure lossin the reboiler outlet lines above theintense-boiling region. Since the pres-

    sure difference between Points 5 and3 in Figure 1 is constant no matterwhich path is taken, this higher pres-sure loss in the nozzles above the in-tense boiling zone will induce a higher

    vapor flow into the outlet nozzlescloser to the outlet baffle.

    Mysteries explainedThis overall assessment of the reboileroperation led us to the development ofa model. Described in greater detailin Reference [8], the model quantifiesthe theory presented above, and it ex-plains all of the observations.

    The model confirms a maldistribu-

    tion situation: a single zone can han-dle substantially more than the de-sign liquid flowrate and produce muchmore than the design amount of vapor,

    based on a much improved tempera-ture driving force. Further, the modelconfirms that this high vapor produc-tion comes about because of relativelysmall gradients in head within thetube field, brought about by regionsof low froth density due to high vaporproduction. The relatively inactivezones explain the lower-than-designoverall heat transfer.

    The regional maldistribution patternreported here may be universal amongkettle reboilers. But in most cases, it

    will not result in significant perfor-mance issues. Instead, this maldistri-bution is more significant in reboilersfor mixtures such as those used in ex-tractive distillation, where the boilingliquid consists of a large concentrationof very high boilers and a small concen-tration of volatile components.

    The mulitifeed, multidraw arrange-ment plays a major role. A single feed,single draw kettle would have experi-enced a different maldistribution pat-tern, possibly even more severe.

    A stagnant region or dead zone isformed in the tubesheet quadrant by aself-accelerating mechanism. With littledisengagement space, not much boilingtakes place. The low-boiling compo-nents tend to preferentially vaporizeout of the liquid, so the liquid becomesrich in the non-volatile component. Thenonvolatile liquid, with little aeration,develops a high hydrostatic head, whichresists movement of fresh liquid intothis region and promotes stagnation.This region could constitute a signifi-

    cant performance issue, with lower thanexpected heat transfer rates, or lead tofouling in heat sensitive materials

    The high hydrostatic head in thestagnant zone, and the enhancedamount of fresh feed diverted into thecentral region, initiate a region of highL/V ratio, with vigorous boiling in thebulk of the bundle. The boiling is prob-ably accompanied by entrainment intothe overhead vapor line, low fractional

    vaporization, and a relatively highconcentration of lights.

    Finally, the region near the overflowbaffle receives close to the design feedflow and has ample disengagement

    space, and is the only region behavinglike a kettle.

    This pattern explains all the ob-servations made: the high pressure

    drop, the measured temperatures, thelower heat transfer coefficient (causedby the stagnant region), and even thelikely entrainment.

    The regional maldistribution pat-tern reported here gives completesupport to the assumption that thereboiler liquid head in Equation (1),hs,REB, is the static head of non-aer-ated liquid in the reboiler. This headis set by the static head of the liquidin the stagnant region. This explana-tion closes the loop on the reason for

    the chimney tray overflow. The highreboiler static head, plus the high re-boiler outlet line pressure drop led tothe high liquid heads in the chimneytray. Entrainment from the intenseboiling central zone could also havebeen a contributor.

    The low temperatures at the vaporoutlet nozzles right above the intense-boiling region were caused by the highL/V ratios in this region, possibly as-sisted by entrainment. Above the kettleregion, the L/V ratios were much lower,

    and so was entrainment, leading to thehigher than expected temperatures.

    The high temperatures measurednear the walls of the exchanger werecaused by horizontal flow of hot liquidfrom the stagnant zone along the walltowards the boiling zones.

    The wall-supported foaming ob-served on the Oldershaw tests, andthe slight foaming observed in thefoaming tests, could also be a contrib-uting factor that could aggravate theintense boiling effect.

    The malfunction reported here,namely, excess pressure drop in akettle reboiler circuit when comparedto the available head, was identified[7] as the most common malfunctionexperienced in reboilers. The employ-ment of good design practices, coupledwith compiling a valid reboiler pres-sure balance, can readily circumventsuch malfunctions.

    In contrast, the maldistributionpattern reported here has not previ-ously been reported. It is unique tomixtures such as those encountered inextractive distillation, where the boil-ing liquid consists of a large concen-

  • 8/14/2019 CEeNews2_04

    6/6

    tration of high boilers and a small con-centration of volatile components.This maldistribution affects both ket-tle heat transfer and pressure drop,

    and it, too, needs to be accounted forin the kettle design.

    Lessons for designKettle reboiler designs should:1. Strictly adhere to the reboiler pres-

    sure-balance equation. Four-fifthsof kettle reboiler malfunctions arecaused by excess pressure drops inreboiler circuits [7].

    2. Be based upon the clear liquid headin the reboiler, not on aerated-liquidhead.

    3. Take account of froth density gradi-ents within the reboiler, and the re-sulting maldistribution.

    4. Trade off pressure loss through inletnozzles (which provide liquid distri-bution to the reboiler) against thehead and column height required.

    5. Consider a horizontal thermosiphon

    reboiler as an alternative for mix-tures containing a large concentra-tion of components boiling at hightemperatures together with compo-

    nents that are volatile. Edited by Nicholas P. Chopey

    60 CHEMICAL ENGINEERING WWW.CHE.COM FEBRUARY 2004

    Operations & Maintenance

    AuthorsEtienne Rubbers is a corpo-rate finance consultant forSasol (Rosebank, Johannes-burg, South Africa; Phone:+27 11 441 3472; e-mail: [email protected]), in-

    volved in merger and acquisi-tion activities. Previously, hewas a process engineer in theconcept development group,developing a number of thefirms projects from concept to

    basic engineering, construction and commission-ing. Despite his career change, he still has keeninterest in process engineering. He holds a B.Sc.(chemical engineering) from Wits University(Johannesburg) and a masters degree in indus-trial management from Belgiums University ofLeuven, and is a CFA charterholder.

    Kirsten Green is a seniorprocess engineer for SasolTechnology (Phone: +27 11344 0082; e-mail: [email protected]). She startedher career at Sasols coalpreparation plant, gaining ex-perience in materials han-dling. She then joined the con-cept development group,where she has been involvedin feasibility studies for the

    companys chemical projects. She spent the pastyear on rotational training with MWKL in Lon-

    don, doing basic engineering in petroleum refin-ing. She holds a B.Ing. from the University ofStellenbosch and an M.Sc. from University ofCape Town, both in chemical engineering.

    Terry-Ann Fowler is a seniorprocess engineer for Sasol(Phone: +27 11 344 0092; e-mail: [email protected]).She forms part of a team in-

    volved in conceptual develop-ment and design of a numberof the firms chemical and re-finery projects. She hasgained experience on projectsfrom pilot plant studies,through feasibility and con-

    ceptual development, basic engineering, and onto construction and commissioning. Her B.Sc.and M.Sc., both being in chemical engineering,are from the University of Cape Town, and her

    Ph.D., also in chemical engineering, is from WitsUniversity.

    Henry Z. Kister, a FluorCorp. senior fellow and direc-tor of fractionation technology(Phone: 1-949-349-4679; e-mail: [email protected]),has over 25 years experiencein design, troubleshooting, re-

    vamping, field consulting, con-trol and startup of fractiona-tion processes and equipment.Previously, he was Brown &Roots staff consultant on frac-

    tionation, and worked for ICI Australia and Frac-tionation Research Inc. (FRI). The author of text-books Distillation Design and DistillationOperation, as well as 70 published technical arti-cles, he has taught the IChemE-sponsored Prac-tical Distillation Technology course 250 times. A

    recipient of Chemical Engineerings 2002 Awardfor Personal Achievement in Chemical Engineer-ing, Kister holds B.E. and M.E. degrees from theUniversity of NSW in Australia. He is a Fellow ofIChemE and a member of AIChE, and serves onthe FRI Technical Advisory and Design PracticesCommittees.

    Walter J. Stupin is an exec-utive director for Fluor(Phone: 1-949-349-5209; email:walter.stupin @fluor.com). Hemanages and executes processengineering for petroleum re-fining and chemical plants.His over 40 years of processengineering experienceranges from vice president oftechnology at C F Braun Inc.to research at Fractionation

    Research Inc. (FRI). The distillation field has

    been an area of key interest, in which he haspublished over 20 technical papers. Dr. Stupinholds B.S., M.S. and Ph.D. degrees, all in chemi-cal engineering, from the University of SouthernCalifornia.

    References1. Kister, H. Z., Distillation Operation, Mc-

    Graw-Hill, New York, 1990.2. Perry, J. H., Chemical Engineers' Hand-

    book, 7th Ed., McGraw-Hill, New York, 1998.3. Lieberman, N. P., Process Design for Reli-

    able Operation, 2nd Ed., Gulf Publishing,Houston, 1988.

    4. Palen, J. W. and Small, W. M., Kettle and InternalReboilers,Hydrocarb. Proc.,43 No. 11, p. 199, 1964.

    5. Fair, J. R. and Klip, A., Thermal Design ofHorizontal Reboilers, Chem. Eng. Prog., p.86, March 1983.

    6. Ishihara, K., others, Critical Review of Correla-

    tions for Predicting Two-Phase Flow PressureDrop Across Tube Banks,Heat Transfer Engi-neering, 1, No. 3, p. 23, JanuaryMarch 1980.

    7. Kister, H. Z., What Caused Tower Malfunc-tions in the Last 50 Years?, Trans. IChemE,

    Vol. 81, Part A, p. 5, January 2003.8. Rubbers, E., Green, K., Fowler, T., Kister, H. Z.,

    and Stupin, W. J. Once-Thru Reboiler StartupCan be Exciting, in Distillation 2003: on thePath to High Capacity, Efficient Splits, Proceed-ings of Topical Conference, AIChE Spring Meet-ing, New Orleans, La., March 31April 3, 2003.