On the design and simulation of an airlift loop bioreactor with

13 Pages·2009·1.42 MB·English

foodandbioproductsprocessing 87 (2009) 215–227
ContentslistsavailableatScienceDirect
Food and Bioproducts Processing
journal homepage: www.elsevier.com/locate/fbp
On the design and simulation of an airlift loop bioreactor
with microbubble generation by fluidic oscillation
William B. Zimmermana,∗, Buddhika N. Hewakandambyb, Václav Tesarˇc,
H.C. Hemaka Bandulasenaa, Olumuyiwa A. Omotowaa
aDepartmentofChemicalandProcessEngineering,UniversityofSheffield,SheffieldS102TN,UK
bDepartmentofChemicalandEnvironmentalEngineering,UniversityPark,NottinghamNG72RD,UK
cInstituteofThermomechanicsoftheAcademyofSciencesoftheCzechRepublicv.v.i.,18200Prague,CzechRepublic
a b s t r a c t
Microbubblegenerationbyanovelfluidicoscillatordrivenapproachisanalyzed,withaviewtoidentifyingthekey
designelementsandtheirdifferencesfromstandardapproachestoairliftloopbioreactordesign.Themicrobubble
generationmechanismhasbeenshowntoachievehighmasstransferratesbythedecreaseofthebubblediameter,
byhydrodynamicstabilizationthatavoidscoalescenceincreasingthebubblediameter,andbylongerresidencetimes
offsettingslowerconvection.Thefluidicoscillatorapproachalsodecreasesthefrictionlossesinpipenetworksand
innozzles/diffusersduetoboundarylayerdisruption,sothereisactuallyanenergeticconsumptionsavingsinusing
this approach over steady flow. These dual advantages make the microbubble generation approach a promising
componentofanovelairliftloopbioreactorwhosedesignispresentedhere.Theequipment,controlsystemforflow
andtemperature,andtheoptimizationofthenozzlebankforthegasdistributionsystemarepresented.
©2009TheInstitutionofChemicalEngineers.PublishedbyElsevierB.V.Allrightsreserved.
Keywords: Airliftloopbioreactor;Microbubblegeneration;Fluidicoscillators;Transportphenomena;Biorefineries
1. Introduction aresusceptibletoflocformation.Typically,microbesorcells
thatsedimentoutofthesuspensionaccumulateinstagnation
Airliftreactorsareperceivedtohaveperformanceadvantages zonesatthebottomexpire.
over bubble columns and stirred tank bioreactors for many Given the importance of energy usage in the operation
applications, biorenewables production in particular. Where of ALBs, it is surprising that the sparging system, which is
the product is a commodity biochemical or biofuel, energy the central power consumption feature of the ALB, has not
efficiency is the primary concern. There are multiple objec- received more attention. Jones (2007) gives a good review of
tivesfortheoptimizationofenergyefficiency,however.The the major features of ALB, including the conventional types
hydrodynamics of stirring is an important consideration, as of sparger design. Chisti and Moo-Young (1987) classify the
are the phase transfer of nutrient influx and the efflux of spargers used in the ALB as dynamic and static. Dynamic
inhibitorproductsandbyproducts.Finally,themetabolismof spargers use injection through nozzles to disperse the gas
cellsormicrobesengagedinthebiochemicalproductionarea introduced. Static spargers are typically less reliant on the
majorconstrainingfactor–masstransferfromthebulkliquid momentum of the jet, and the gas is introduced typically
tothebioculturemustbemaintained.Therearetwoimpor- throughaperforatedplate(seeDeshpandeandZimmerman,
tant reasons to use airlift loop bioreactors (ALB) that arise 2005a,b)orlesscommonly,throughaporousbaffle(Heijnen
fromtheairlifteffects:flotationandflocculation.Smallbub- andVan’tTiet,1984).Thisstudywasmotivatedbythedevel-
bles attached to particles or droplets significantly lower the opmentofanovelmicrobubblegenerationtechniquebasedon
densityoftheaggregate.GrammatikaandZimmerman(2001) fluidicoscillationdivertingjetsusedinsparging(Zimmerman
describethesegeneralizedflotationeffects.Suchaggregates etal.,2008).
∗ Correspondingauthor.Tel.:[phone];fax:[phone].
E-mailaddress:w.zimmerman@sheffield.ac.uk(W.B.Zimmerman).
Received7November2008;Receivedinrevisedform20March2009;Accepted23March2009
0960-3085/$–seefrontmatter©2009TheInstitutionofChemicalEngineers.PublishedbyElsevierB.V.Allrightsreserved.
doi:10.1016/j.fbp.2009.03.006
216 foodandbioproductsprocessing 87 (2009) 215–227
The paper is organized as follows. In Section 2, the Thisrangeoffrequenciesincludesthetargetrangeofthenatu-
microbubble generation mechanism by fluidic oscillation is ralbiochemicalpathwaysoscillationsofglycolysisinyeast,so
discussed,leadingtodesigncriteriaforspargingsystemsand arelikelytobeexcitedbybacterialcultureswhichexhibitgly-
nozzlebanksthatachievehighenergyefficiency.InSection3, colyticcyclesinaerobicmetabolism.However,italsooccurred
aspectsofthedesignofALBsthatareinfluencedbytheincor- to us that there was the potential to generate microbubbles
poration of the microbubble generator by fluidic oscillation, byanovelmechanismusingfluidicoscillation.Typically,bub-
thedesignitself,andsimulationofansuchanALBarepre- bles generated from a rigid aperture are approximately an
sented.InSection4,asummaryisgivenandconclusionsare order of magnitude larger than the aperture. Unfortunately,
drawn. evenusingextremelysmallholesdoesnotnecessarilyensure
generatingsmallairbubbles.Bubblesgrowduringtheprocess
2. Microbubblegeneration
oftheirformationandwhenfinallyseparatedfromtheaper-
tureexit,theirdiameterisoftenmanytimeslargerthanthe
Theconceptforthefluidicoscillatordrivenmicrobubblegen- hole diameter. The separation is controlled by surface ten-
erationmechanism(Zimmermanetal.,2008)stemsfromthe sion of the water; and an important negative factor is the
systems biology objective of using oscillatory nutrient feed fact that many contaminants – especially organic – present
streamsinaconventionalfermentororchemostattoinves- inwastewaterareknowntoincreasethesurfacetension.The
tigatethekineticsofmetabolicpathways.Zimmerman(2005) finalbubblesizeiscommonlydeterminedbytheovercoming
demonstratedfromthesimulationoftheglycolyticpathway of the wetting force of the pre-bubble on the solid walls by
in yeast that oscillating the glucose feed stream could cre- buoyantforcesorbycurrentsintheliquid.Thefluidicoscil-
ate information-rich time series responses in extracellular lation,ifappropriatelyconfiguredandtuned,couldinterfere
metabolic production, such as excreted ethanol. Specifi- withthisbalanceofforcesasthepre-bubblereachessizesnot
cally, it was shown that three kinetics coefficients of the muchlargerthantheaperturediameter,sincetheaccelera-
Michaelis–Mentenkineticsofthebranchpointofsuccinicacid tion force of the oscillation can be made arbitrarily large by
production could be inferred by data assimilation with high usinghighfrequenciesandlargeamplitudeoscillationfrom
fidelityduetothepresenceofstrongnonlinearityexcitedat jet diversion. Such a microbubble generator would have the
oneoftworesonantfrequenciesofthepathway.Theeigen- benefitofverylittleadditionalenergy(pumpingheadloss)to
systemanalysisofthepathwayshowednatural(butdecaying) break off the bubble, so should be highly energy efficient in
frequenciesinthepathwayat15Hzand52Hz.Thelatterwas formation.
identifiedbyeigenvectoranalysisasAMP-ATP,NAD-NADHco- Inthissection,wediscussthepotentialdesignbenefitsof
metaboliteoscillations.Theformerinvolvedfourreactionsto microbubblessoproducedandthemethodologiesoftheirpro-
whichpyruvatewasessential. ductionthatwehaveengineered,primarilywiththetargetof
As such oscillations are not observable in unforced increasing the mass transfer for aeration in bioreactors, but
chemostats, in order to test the hypothesis, the first author theprincipleappliestojustaboutanynutrientintroduciblein
sought an approach for designing a bioreactor which would thegaseousphase.
have an easily controllable and oscillating nutrient stream.
As liquid phase oscillation is difficult to achieve at low 2.1. Thebenefitsofmicrobubbles
flowrates, the obvious stream to control would be pneu-
matic. Thus the introduction a gas stream with a nutrient Whyisthatwhenblowingacontinuousstreamofairthrough
in the gaseous phase would be ideal. The easiest nutrient asmallopeningthatwedonottypicallygetsmallbubbles?
to introduce for most biocultures is oxygen, so the target For instance, when a bubble is formed from a single open-
switchedfromyeastfermentationtoprocessescontrolledby ing,theliquidattachedtoitsperimeterprovidesananchoring
aerationrate,animportantexampleofwhichiswastewater effectasthewettingforceattachesthegrowingbubbletothe
treatment. solidsurface.Unlessthisanchoringforceisdisturbed,thebub-
Purification,treatment,orremovalofcontaminantsfrom ble will grow until the buoyant force on the bubble exceeds
waterisoftenassociatedwiththeirdecomposition.Animpor- theanchoringrestraintonthebubble,causingittopinchoff.
tantmethod,thebiologicaltreatmentofwater,isessentially Typically,thebuoyantforcedoesnotexceedthatofthewet-
anintensificationofthenaturallyoccurringdecompositionof tinganchoruntilthebubbleisaboutanorderofmagnitude
contaminantbyactionofmicro-organisms,mainlybacteria. larger than the diameter of the hole. The process is sensi-
Though this method alone cannot remove all possible con- tive to the wetting properties of the solid surface as well.
taminants(andespeciallyfortreatmentofvariousindustrial If the bubble contacts the surface over a larger region than
wastewatershastobeoftencombinedwithothermethods)it the aperture perimeter, or if the solid surface is hydropho-
iseffectiveandcommonlyappliedmethodforbreakingdown bic,thegasphaseofthegrowingbubblewillformasecond
themajorpollutants:organicmatter,nitratesandphosphates anchorforcewiththesolidsurfaceoverawiderarea,increas-
(Stevenson, 1997). The efficiency of the process – or, more ing the buoyant force and thus bubble volume required to
specificallyoftheaerobicstageofthebiologicaldecomposi- overcome it. If the surface is hydrophilic, then this attrac-
tion–islimitedbyavailabilityofoxygenneededforgrowth tive force is absent. In the next subsection, the generation
andactivityofaerobicmicro-organisms. mechanism for microbubbles by fluidic oscillation is dis-
Weproposedtocombineafluidicoscillatorwithanozzle cussed. In this section, the desirability of small bubbles is
bankwiththeintentionofproducinganoscillatorystreamof discussed.
bubbles.Tesarˇetal.(2006)hadpreviouslydevelopedasimple Themajoradvantageofsmallbubblesisthesurfaceareato
andinexpensiveno-movingpartfluidicdevicewithadjustable volumeratio.Nearlyallinterfacialtransportprocesses–heat,
frequency in the range of 1–100Hz, controllable by chang- mass,momentum–aredependentonthesurfaceareaofthe
ing the length of the feedback loop, but for essentially high interfacebetweenthephases.Itisgeometricallyclearthatthe
Reynoldsnumberflowsachievableevenalowairflowrates. surface area to volume ratio of a spherical bubble increases
foodandbioproductsprocessing 87 (2009) 215–227 217
inverselyproportionatetoitsradiusordiameter: the control of continuous flow through a pipework network
anddistributeddiffuseraerationsysteminpilottrials.
S = 4(cid:2)r2 = 3 (1) Asanaside,theargumentgivenaboveformassandheat
V (4/3)(cid:2)r3 r transferenhancementbysmallerbubbleswithequivalentvol-
umeofdispersedphaseholdsformomentumtransporttoo,
Thismayappearalgebraicallyequivalent,butsupposewe
with some modification. The classical Stokes law serves as
maintain that the total volume V0 of the bubble phase is to aguidefortheresidencetimeofamicrobubbleinaviscous
remainconstant,then
liquid:
3
S= rV0 (2) U = 2g(cid:3)(cid:4)d2 (4)
stokes 9 (cid:5)
l
isthesurfaceareathatthetotalphaseexhibits.Forinstance,if
1lofairisdistributedin100(cid:2)msizebubbles,thereare10m2of Due to the square, it is clear that the residence times of
interfacialarea,easilycomparableorgreaterthanareasonable smallbubblesismarkedlylongerforthesameheightofliq-
sizedair–liquidinterfaceofthecontinuousphasesinatank uidthanforlargerbubbles.Thussmallerbubbleshavemuch
opentotheatmosphere.Furthermore,thebubblesaremov- longer to transfer their momentum from the bubble to the
ing.Themasstransfercoefficientofasinglebubbleordroplet liquiddraggedalongwiththem,eventhoughtheyhaveless
maybedifficulttoassess,asitdependsonthehydrodynamics momentumtotransfer.Thesetwoeffectswouldbalance,but
ofbubblerise,itsenvironment,andtheconstitutiveproperties forthesurfaceareatovolumeratio–momentumisalsotrans-
ofthemediumandmaterialtransferred(seeDeshpandeand ferred,byshearstress,acrossthesurfaceareaofthebubble.
Zimmerman, 2005a,b), but in most flows it is dominated by Thereforethefluxofmomentumismarkedlyincreasedbythe
convectionforcesandcanbegloballyfittedtoamasstrans- decrease in bubble size, by the same ratio of Eq. (2). It fol-
fercoefficientphenomenologicalequation,wheretheoverall lowsthatmicrobubbleshaveahigher“draggingability”when
fluxisproportionaltotheinterfacialareaSofthedispersed rising or flotation capability with the same volume of fluid
phase.Thustransferdynamics,massor,byanalogy,heatflux, holdup.Thiseffectispotentiallyveryimportantforimproved
arerapidlyenhancedbydecreasingthebubblesize.Theabove mixinginariserregionofanairliftloopbioreactor,provided
argumentforthebenefitintransferefficiencyistypifiedbythe thebubblescanbeproducedenergeticallyefficiently,i.e.the
commonchemicalengineeringphenomenologicaldescription costofthemicrobubbleproductionperunitvolumedoesnot
ofinterphasemasstransferfluxJ(mol/s): rise due to rising friction factor. For design purposes, if the
goalistoachievethesamemixinglevelorriserperformance
J=KlS(cg−cl) (3) withmicrobubbles,thenpotentiallythiscanbeachievedbya
lowervolumetricflowrate,sincethelongerresidencetimein
whereK isthemasstransfercoefficient(unitsofvelocity),S theheightofliquidpermitshigherholdupatlowervolumetric
l
istheinterfacialarea,andcgandclaremolarconcentrations. flowrates.
Mass flux J, all things being equal, increases proportionate These benefits have been tested in laboratory and pilot
to S, and therefore inversely proportionate to the diameter scale experiments. Shi (2006) demonstrated 8-fold smaller
d of the microbubble. Bredwell and Worden (1998) inferred bubbleswithoscillatoryflowthanwiththesamevolumetric
K in an oxygen microbubble column from a plug flow con- flowratesteadyflowthroughthesamenozzlebank,measur-
l
centrationmodelforthedissolvedoxygen.Alaserdiffraction ingan8-foldincreaseindissolvedoxygentransferefficiency
techniquewasusedtocomputetheinterfacialareaS.Worden accordingtothestandardASCEtest.Inarecentlycompleted
and Bredwell (1998) demonstrate that the very high mass pilotscaletrial,Zimmermanandco-workers(internalreport)
transferratesofmicrobubblesrequiremodelingofanintrin- havefound3-foldincreaseinaerationratesusingthefastest
sically transient nature. They found that the presence of frequencyoscillationpossibleintheirfluidicoscillatorsystem
non-transferred gas in the microbubble limited the mass over steady flow through the same flexible membrane dif-
transferrates. fuserarraywiththesamevolumetricflowrate(∼2m3/hper
But one might argue that this flux enhancement effect diffuser).Theyalsorecordedadecreaseinpowerdrawonthe
is balanced by the cost of producing microbubbles. As we blowertoachievethisflowrateof13%,whichwillbediscussed
pointedoutinthebeginningofthissection,onewouldthink inSection2.3.
that to produce smaller bubbles requires smaller holes or
pores.Therefore,withcontinuousflowthroughthesesmaller 2.2. Microbubblegenerationbyfluidicoscillation
openings,thefrictionforcewouldbeexpectedtobepropor-
tionatelylarger.Asfrictionincreaseswithsurfaceareaofpores 2.2.1. Instabilityofparallelpercolation
or channels, one would expect the head loss on the pump Thedesirableaeratorwouldproducesimultaneouslyalarge
duetohydraulicresistancetoriseinverselyproportionateto numberofverysmallairbubbles.Thiswouldresultinalarge
theopeningdiameter.Sothetransferperformanceincrease totalair/waterinterfaceareaandthereforehighrateofoxy-
is offset by the energetic decrease, and no expected overall gentransportintowateracrossthisinterface.Inthevarious
efficiency is likely. This argument, however, argues against attemptsatreachingthissoughtaftersituation,theaerators
seekingtoproducesmallerbubblesbyminiaturizingthehole. havebeenmadewithalargenumberofparalleltinyapertures
The“win”canonlyoccurifthefrictionlossremainsaboutthe exitingairintowater.Unfortunately,thedesirablebubblefor-
same,butthebubblesizeisreduced.Adifferentmechanism mationhasneverbeenobtained–becauseofthefundamental
forbubbleproductionisrequired.Thenextsectiondiscusses instabilitypropertyofthebubblegrowthmechanism.Atthe
the fluidic oscillator driven microbubble formation mecha- beginningofthebubbleformation,thedistributionofairflow
nismreportedbyZimmermanetal.(2008),andthefollowing intotheaperturesisstable.This,however,ceasestobeonce
sectionreportsitssurprisingdecreaseinfrictionlossagainst oneofthegrowingbubblessurpassesthehemisphericallimit
218 foodandbioproductsprocessing 87 (2009) 215–227
shape.Immediately,itsfurthergrowththenbecomeseasier. powerfuloutputflowthroughtheoutputterminalsY1orY2is
Theairenteringintoitmeetsalowerpressuredifference(cid:3)p controlledbymuchweakercontrolflowsinputintothecon-
toovercomethanintheotherapertures.Asaresult,thispar- trolterminalsX1 andX2.Thisparticularamplifierisbistable
ticularbubblestartsgrowingfasterattheexpenseoftheother –itusestheCoandaeffecttoremaininoneofitstwostable
bubbles,whichmayevencompletelyceasetogrow. stateswhenthecontrolactionisabsent.
ThisisdemonstratedinFig.1,whichshowsphotographs The oscillator is made from the amplifier by providing it
takenwhenwatchingasteadyairflowintoanaeratorposi- with a feedback loop. There are several possible feedback
tionedatthebottomofashallowwater-filledtank.Theaerator alternatives even for this jet-diverter type of amplifiers (as
–actuallythesameasintheotherexperimentsdescribedhere discussed in Tesarˇ, 2007). The feedback loop shown at the
–hastheairexitaperturesintheformofarowof0.6mmdiam- right-hand part of Fig. 2 is the particularly simple one, con-
eterparallelholes.Despitetheairflowpathbeingequallyeasy sisting of nothing more than just a tube of suitable (or, in
inalltheholes,thebubblesareformedinonlyoneofthem. Fig. 2, adjustable) length, connecting the two control termi-
Theexceptionalcaseischosenbymerechance–inthetwo nals X1 and X2. As shown in the right-hand part of Fig. 2,
photographs in Fig. 2 under nominally identical conditions, the jet issuing from the main nozzle attaches to either one
different orifice are seen to be active. In this aperture, the of the two attachment walls due to the Coanda effect of jet
bubbles grow to sizes substantially larger – by more than a attachment to a nearby wall and is thereby led into one of
decimalorderofmagnitude–thantheholediameter.Allthe thetwooutputterminals.Becauseofthechangeofdirection
otherorificesareinactive,nobubblesareformedinthem. duetothedeflection,airflowtrajectoriesinsidethejetinthe
This failure to obtain the desirable parallel formation vicinityofthecontrolnozzlesarecurved.Thiscurvaturecre-
of tiny bubbles is very fundamental – it is the very basic atesaradialpressuregradientacrossthejet.Inthesituation
Young–Laplacelawofsurfacetension(cid:6)thatgovernsthebub- showninFig.2,thiscausesadecreaseinpressureatthecon-
ble formation. According to it, the pressure difference (cid:3)p trolportX1,whichthendrawsairthroughthefeedbackloop
across the air/water surface is inversely proportional to the fromtheoppositecontrolterminalX2 wherethepressureis
curvatureradiusRofthesurface: higher. It takes some time for the flow in the feedback loop
tube to gain momentum, but when this happens, the con-
2(cid:6)
(cid:3)p= R (5) trol flow in X1 suffices, because of the amplification effect,
forswitchingthemainjetfromtheterminalY1anddiverting
Whenabubbleisformedinaroundorificeexit,itscurva- itintototheotherterminalY2.Asthedeviceissymmetric,
tureradiusisinitiallyextremelylarge,butdecreasesfast.The this jet switching is – after a delay needed for the feedback
pressure needed for the bubble growth increases. The criti- flow to gain momentum into the opposite direction – then
calsituationisreachedoncethegrowingbubbleattainsthe reproduced in the opposite way, thus leading to a periodic
hemispherical shape. Thereafter, its radius R increases with switchingprocess.Tesarˇetal.(2006)demonstratethatthefre-
increasingvolumeand,consequently,thepressuredifference quencyoftheoscillationiscontrolledprimarilybythelength
(cid:3)pdecreases.Alltheairthenentersintotheparticularbub- of the feedback loop and the supply flow rate. The acoustic
blethatfirstformsabubblewhereitmeetslessoppositionto regimeoffrequenciesbetween1and100Hzcouldbereadily
flow.Thismakesitsfurthergrowtheasier.Thisinstabilityis achievedinamodelconsistingoftheplatestacksshownin
essentially the same mechanism as Saffman–Taylor viscous Fig.2.
fingering(ZimmermanandHomsy,1991). Tesarˇ (2007) reviews, apart from this jet deflection type,
manyotheralternativetypesoffluidicoscillatorsthatmaybe
2.2.2. Jetdiversionfluidicoscillation usetodriveALBs.Allofthemwhenusedinthisapplication,
The key idea of the microbubble generation method as aresuppliedwithsteadyairflowandproduceself-excitedair
described by Zimmerman et al. (2008), is to limit the bub- flow oscillation. Essentially, the oscillation due to an intrin-
blegrowthtimebythedurationoftheperiodofanoscillator sic hydrodynamic instability caused by the presence of the
thatsuppliesairintothebubble-formationapertures–which feedbackaction–thechangeinthelargeoutputflowdueto
maybenozzles,diffusers,porousbaffles,perforatedplatesor thereturnflowofasmallfractionofthefluidintoalocation
microporousmaterials.Thegrowthisterminatedattheendof whereitcanactagainstthecausewhichgeneratedtheoutput
eachoscillationhalf-period.Thebubbleisthenremovedfrom effect.Theprinciplesunderlyingthefluidicoscillatorscanbe
theaperturesothatthegrowthofthenextbubblehastostart classifiedintothreegroups:
anewinthenextperiod.Nobubblecanreachthelargesize
typicalforsteadyblowing. (1) Twinvalveoscillatorswithmutual–phaseshifted–block-
Theessentialpartoftheaerationsystemisthereforeaflu- agebyfluidicamplifierscapableofproducingtheflowturn
idicoscillator–preferably,duetotheadvantagesofreliability, downeffect.Thisisaratherrarelyusedprinciple.
robustness,andlowprice,anoscillatoroftheno-moving-part (2) Anexternalfeedbackloopaddedtoasingleamplifiervalve
type–oneofthevarietydescribed,e.g.,byTesarˇ(2007).Shown (orseveralvalvesforminganamplifyingcascade).Thisis
inFig.2isanexampleofaparticularlysuitablefluidicflow- theveryobviousandalsoverycommonoperatingprinci-
divertingoscillatordesign.Itsmaincomponentisthefluidic ple,mostlyusedtogenerateanoscillatoryorpulsatilefluid
amplifier(originallydesignedforanotherapplication)shown flowinaconnectedloadwiththejet-deflectionamplifiers.
intheleftpartofFig.2.Detailsofitsgeometryandproperties Becauseofthesymmetryoftheamplifier,theoscillators
aredescribedbyTesarˇetal.(2006).Steadyairflowissupplied usuallypossestwofeedbackchannels,oneconnectingY1
intotheterminalSanditsflowintoonethetwooutputter- with X1 and the other connecting Y2 with X2. It is the
minalsY1 andY2 iscontrolledbythecontrolactionapplied principle used here, having the less obvious single loop
tothecontrolterminalsX1andX2.Thecontrolactiondeflects layout.
thejetoftheairissuingfromthemainnozzle,whichiscon- (3) Internal feedback oscillator, sometimes using a geome-
nectedtoS.Thedeviceisdescribedasanamplifierbecausethe tryreminiscentofafluidicvalve,sometimeswithrather
foodandbioproductsprocessing 87 (2009) 215–227 219
Fig.1– Examplesoflargebubblegenerationbysteadyblowingofairthroughsmallsubmergedorifices.Theaeratoris
providedwithanarrayof600(cid:2)mparallelholeshavingtheiraxesorientedvertically.Insteadoftheexpectedstreamof
manyparallelsmallbubbles,theparallelpercolationinstabilityleadstogenerationoflargebubbles–ofsizemorethan
10-foldlarger(mostbubblesonthetopleftare∼10mm).Thetoprightisanimagefromthesamesystembutwiththe
fluidicoscillatoractive.Theflowratesarecomparablebutnotequal,2.5:1inratio.Thesteadyflowdoesnothavesteady
bubbleproductionatthelowerflowrateoftherighthandimage.Atthehigherflowrateofthesteadyflow,theoscillator
inducedbubblesaretoocloselyspacedandcoalesce.Eightimagestakenfromtheoscillatordrivenexperimentina8.5mm
by7.4mmwindowwereusedtocollect237bubblesizes(diameterofequivalentareacircularsection).Itshouldbenoted
thatthenozzleblockis12cmlong.Themeanbubblesizeofthedistributionoffrequencyis700±25(cid:2)m.
remote from it and perhaps retaining only topologi- diverter configuration adopted here is particularly useful
cal similarity. Also used to generate an output fluid when employed for the flow diversion, as it can feed alter-
flow. natelyonebankortheotherofaeratornozzles,astheseare
shownintheschematicFig.3.Theswitchingprovidesshort
Any of these principles can be used to design an oscil- pulsesofmomentuminthenozzles,whicharrivewithregular
lator for generation of the oscillatory air flow for ALB. The frequencyattheexitapertures.
Fig.2– Left:themodelofthefluidicjet-deflectionamplifierusedinthetests.ItisastackofPMMAplateswithlaser-cut
cavities–containingnomovingmechanicalparts.Thescrewsare1/4inchheads.Right:thefluidicoscillatorismadefrom
theamplifierbyprovidingitwiththefeedbackloop(shownhereofadjustablelengthfortuningtheoscillationfrequency)
connectingitstwocontrolterminals.
220 foodandbioproductsprocessing 87 (2009) 215–227
Fig.3– Schematicrepresentationofthefluidicdiverteroscillatordrivingthemicrobubblegenerationsystemwithtwo
nozzlebanksfedbyalternateoscillation“strokes”.
Applyingthefluidicoscillatortothetaskofgenerationof Fig. 4. The water and air orifices are arranged in a grooved
themicrobubblesrisesis,however,notnecessarilystraightfor- channel so that they form two rows, with their axes mutu-
ward.Theproblemencounteredistheremovalofthenascent allyatrightangles.Thebubblesthathavegrownduringthe
microbubblefromitsairinletaperturewhileitisstillsmaller firsthalfoftheoscillationperiodontheairsideareblownoff
thanthesizeofthecritical,hemisphericalcap. fromtheiraperturesinthesecondhalfoftheperiod,during
Naturallygrownbubblesseparatefromtheiraperturesdue which the water flow pulse is admitted to the water side of
to the hydrostatic lift force. This, however, reaches the nec- the groove. Fig. 5 shows the success of this method, where
essarylevelonlywhenthebubbleisrelativelyverylarge.The theoscillatoryflowgeneratesmicrobubblesfromthe600(cid:2)m
smallbubbles,notyethavingreachedthecriticalhemisphere holesofsubmillimetresize,whereasthosegeneratedbythe
shape, need not and often do not separate. When supplied steadyflowaredominatedby>6mmbubbles.Oneobservation
withtheoscillatoryairflow,theymaystayattachedincreasing fromFig.5(seebubblesizedistributioninFig.1)isthatthe
anddecreasinginsizeintherhythmoftheoscillation. smallbubblesareapproximatelymonodisperseandregularly
One possible solution is applying the oscillatory input spaced,andthereforedonotsufferfromcoalescence.Crabtree
air flow pulses with considerable momentum, sufficient to andBridgwater(1969)provideamechanisticargumentforthe
dislodge the nascent microbubbles. Adjusting the proper non-coalescenceofchainsofbubblesbeinghydrodynamically
momentumis,however,ratherdifficultandexperienceshows stabilizedastheyrise.
it may need re-adjustments with changing operating condi- Additionally, another method of microbubble formation
tions. wasdemonstratedinalayoutinwhichshortsmall-diameter
A rather less sensitive solution is the original concept of holesconnecttheexitorificewithalarger-diametermanifold.
theauthorstoblow-offthenascentmicrobubblesbyanadja- Theaeratorissuppliedwiththeoscillatoryflowdeliveredby
centwaterflowpulseintheotherhalfoftheoscillationperiod. thefluidicoscillatorandoperatesinaperiodicmanner,dur-
Inparticular,thiswasdemonstratedinamodelpresentedin ingwhichalternatingairandwatercolumnsmovebackand
Fig.4– Left:configurationfortheblow-offmechanismforbubbledetachment.WhilethesamediverteroscillatorasinFig.3
isused,onlyoneofitsoutputs,A,isusedtodelivertheairtotheaerationaperturesduringthehalf-period.Theother,B,
deliversawaterflowpulseintoasystemofadjacentwater-flownozzles.Centre:thebubbleformationhalfoftheperiod.
Right:waterflowpulseremovesthenascentbubblesfromtheirapertures.
foodandbioproductsprocessing 87 (2009) 215–227 221
configuration but without the water flow – air flow in both
adjacent banks – was conducted, with the observation that
smallbubblesformfrombothbanksofholes.Therefore,the
essentialfeatureistheorientationeffect–ahorizontalcom-
ponentisnecessary.Anartist’sconceptionofhowhorizontal
orientationisnecessaryforonlyairflowisshowninFig.6.
The key element in this mechanism is the recoil of the
nascent microbubble against the solid wall, and particu-
larlythewettingpropertiesofthemicrobubble.Forinstance,
we fabricated nozzlebanks from microchannels wet etched
in both glass and PDMS microchips. The glass microchips
formed small bubbles of the size of the aperture with the
microchannel apertures (60(cid:2)m characteristic size) oriented
horizontally, but ∼500(cid:2)m bubbles when oriented vertically
(Varma,2007).ThePDMSanaloguesonlyformedlargebubbles.
ThebubbleattachestothePDMSsurfaceandgrowsalongthe
microchipexteriorsurface.AsbubbleswetthePDMSsurface
Fig.5– Tinybubblesgeneratedinanexperimentwiththe
butnottheglasssurface,itisclearthatwettingpropertiesof
aeratorsystemmodelcorrespondingtoFig.4.Actually,the
the orifice and adjacent solid wall are extremely important
samenozzlebankisusedastheonegeneratingthelarge
for the formation of microbubbles by the fluidic oscillation
bubbleswiththesteady(non-oscillated)airflowthrough
mechanism.
the0.6mmholesasshowninFig.1.
2.3. Energyefficiency
forththroughthemanifold.Duringthefirsthalfoftheperiod,
the small holes are filled with air from an air column mov- One of the unexpected outcomes of an ongoing set of pilot
ingpastthem.Inthefollowinghalf-period,theliquidcolumn scale trials in wastewater treatment with a pneumatic dis-
comesandmovesby.Itentersthesmallholesanddislodges tributorsystemfortwobanksofconventionalaerators,called
theairfromthem.Thismayseemtobeasomewhatcomplex membranediffusers,wasthedecreaseinpowerconsumption
mechanism,buttheformationofthealternatingwaterandair by the fluidic oscillator inserted as the splitter between the
columnsinthemanifoldwasobservedtobeautomatic,new twobanks(seeFig.3).Typically,oneexpectsthattheinsertion
watercolumnsformingandgraduallyprogressingdownthe of a fitting into a flow distribution system, such as a bend,
manifoldastheaircolumnsbetweenthemdiminished,hav- valve, or splitter will add an additional hydraulic resistance
inglosttheirairbyfillingthesmallholes.Ancertainadvantage to the system. So the design trade-off for microbubble gen-
ofthismethodistheratherlowrequiredoscillationfrequency, erationdrivenbyafluidicoscillatorwouldbeexpectedtobe
correspondingtothereciprocatingmotionsoftheliquidand increasedmasstransferperformancescalingwiththeratioof
aircolumns. thediametertypicalbubblegeneratedbyfreestreamsteady
Althoughthe“blow-off”configurationsucceedsincreating flowandthemicrobubblediametergeneratedbyfluidicoscil-
smallbubblesonthescaleoftheaperture,thepricepaidis lation, and the expected additional head loss at a constant
that it uses half the momentum for air flow and bleeds off volumetricflowrate.Clearly,theunexpectedhydraulicresis-
halfthevolumetricflowrate.Itisareasonablequestiontoask tancedecreaseofoscillatoryflowrequiresanexplanation.We
astowhetherthispriceisessential.Sothetestofthesame believetherearetwocomponentstothedecrease.
Fig.6– Anschematicofthemicrobubblegenerationfromahorizontallyorientednozzle/gasflowpath.Therearefour
distinctphases:(1)pushingoutofahemisphericalcapatthebeginningoftheupstrokeofthefluidicoscillator;(2)buoyant
riseofthegrowingbubbleduringtheremainderoftheupstroke;(3)suctionofthebubbletocollideagainstthesolidsurface;
(4)recoilofthebubblefromthesolidsurfaceandbreakoffofthebubble.
222 foodandbioproductsprocessing 87 (2009) 215–227
2.3.1. Coandaeffectfrictionreduction works as well for a single impingent jet. Tesarˇ et al. (2007)
The fluidic oscillator in Fig. 3, on time average, serves as a showasimilarconclusionexperimentallyforturbulentheat
splitter.Theaggregateflowratethrougheachexitchannelis transfer.
equal, yet the flow never flows down both channels simul- Thespeculationhereisthatoscillatoryflowreducesskin
taneously.Ifthefeedbackloopwereomitted,i.e.thecontrol frictionsincetheviscousboundarylayerisdisruptedinform-
ports closed off (Fig. 2), then it would be expected that the inginthedirectionperpendiculartotheflow.Themomentum
fluid would fill the ducts on both sides due to the splitter pulsesfindmuchlessresistanceinpushingdownthecenterof
action. This is exactly what happens in our control experi- thechannelthanfromtheslowermovingfluidnearthewall,
ment, where the fluidic oscillator is replaced by tee-splitter. astheviscousfrictionhasnothadtimeto“diffuse”outward
Thetee-splitterandtheclosed-offcontrolportsontheoscil- fromthesinkofmomentumatthewall.
latorachievedalmostexactlythesameenergyconsumption Thisargumentforskinfrictionreductionworksaswellfor
at constant volumetric flow rate. In both cases, the splitter turbulentflow,butthetimescalesforturbulentwallboundary
action results in the mean flow having a stagnation point layerestablishmentareshorterinscalingfactor,butgiventhe
at the geometric point of the split, regardless of the design much higher Reynolds number achievable in turbulent flow,
ofthesplitter.Conversely,whenthejetflowsthrougheither thisfeaturecanbeovercomewithhigherflowrates(orfaster
channel in the fluidic oscillator driven flow, the jet attaches oscillation). The classical estimates for turbulent boundary
tothecurvedsidewallinFig.1accordingtotheCoandaeffect. thicknessandskinfrictionare:
Althoughthisflowsmoothlycurvestowardtheoutletport,the
diverted jet has no stagnation point. The friction loss along ı 0.385 0.0594
= C = (7)
thewallatandnearastagnationpointisappreciable,andis x Re1/2 f Re1/2
x x
completelyavoidedbyeitherdivertedjetinafluidicoscilla-
tor.
Withoutanydetailedexperimentalstudy,theresultsfrom
our pilot trials suggest that these two resistance reduction
2.3.2. Boundarylayereffects
effects–Coandaeffectremovingthestagnationpointofthe
Turbulent flow in ducts experiences a viscous sublayer near
splitter and skin friction reduction by slow boundary layer
thewallinwhichdissipationislargest,exceedingtheinterior
formation – are estimated to be about equal in importance,
dissipation from eddy motions in the bulk. It is well known
about6–7%reductioneachwithonevolumetricflowrate,with
that solid bounding surfaces induce most of the dissipation
the inference based on a roughly linear decrease in energy
lossinstatisticallystationaryturbulentflows.Butwhatabout
consumption with increasing oscillation frequency at high
oscillatoryflowswiththesameaveragevolumetricflowrate?
frequencies, but a plateau in reduction at low frequencies,
Sinceouroscillationisa“positivedisplacement”syntheticjet,
but too little data for a more accurate assessment. Higher
itisconceptuallyusefultoviewitasaseriesofmomentum
flow rates led to greater energy consumption savings, con-
pulsesseparatedbymomentum“gaps”.Thefluidissuddenly
sistent with the implication of Eq. (2) and our assumption
acceleratedbythemomentumpulse,andthenitsinertiatrails
aboutthescalingofthetimetosetupaturbulentboundary
offuntilthenextmomentumpulseisexcited.Aconceptual
layer.
modelforthisistheclassicalboundarylayerproblemofthe
suddenlyacceleratedplate,forwhichtheframeofreferenceis
changedtothestationaryplatewiththefluidsuddenlyaccel- 3. Designaspectsofanairliftloop
erated.Thelaminarresultispresentedintheclassicalworkby bioreactor
Rosenhead(1963).Thethicknessofthelaminarboundarylayer
ıandtheskinfrictioncoefficientC aregiven,indimensionless In the previous section, the design aspects of a microbub-
f
form,by ble generator component of an airlift loop bioreactor were
discussed, demonstrating that the usual design trade-off
ı 5 0.664 betweenfrictionlosseswithsmallaperturesanddistributor
= √ C = √ (6)
x Rex f Rex channels and performance gains in transfer efficiency with
smallbubblescanbe“triangulated”withthefluidicoscillator
wherexisthedownstreamcoordinatefromthestartofthe principle,withtheoscillatoryflowresultinginlessfrictionloss
pulse.WithalaminarboundarylayerathighReynoldsnum- whilestillgeneratingsmallbubbles.Thisargumentposesthe
ber, one could argue that the time to set up the boundary advantageofusingsuchafluidicoscillatordrivenmicrobub-
layershouldbeinverselyrelatedtothedimensionlessbound- ble generator in many chemical engineering processes, but
arylayerthickness,andthusscalingwiththesquare-rootof still leaves many design questions. Before a design can be
theReynoldsnumber.Sothetimetosetupaboundarylayer confidently implemented, information must be collected on
is large. What if the period of the fluidic oscillator switches performanceaspectsthatareaffectedbyinclusionofthenovel
before the boundary layer is set up? That problem for dual elementinthedesign.Inthissection,wewilladdressthelikely
laminarimpingentjetshasbeenstudiedbyHewakandamby influenceofthemicrobubblegeneratoronthetypicalperfor-
(2008).Theheattransfercoefficientfortheoscillatingimpin- manceofanALB.However,thisisrecognizedasnosubstitute
gent jet was found to be much higher than under steady foractualperformanceobservationandresultsofoperational
dual impingent jets, as the oscillation disrupts the forma- studies.Thereisa“chicken-and-the-egg”problemherethat
tion of the boundary layer that limits the transfer to the in order to design an ALB properly with this novel compo-
surfacetoconductionthroughtheboundarylayer.Thisprin- nent, it is necessary to construct, commission and operate
ciple works as well for mass and momentum transfer. The aprototype.Buttheprototypemustbedesigned.Thescope
transferratetotheimpingentsurfaceismuchhigherdueto of this article is the design of the prototype, whose opera-
thedisruptioninsettinguptheboundarylayerinthedirec- tionalperformancevariationcanthenbereportedonindue
tion opposite of the impingent jets. Clearly, the argument course.
foodandbioproductsprocessing 87 (2009) 215–227 223
totheriserregion.Sufficientclearspaceforthisdrawingin
should be available so that friction losses are not apprecia-
ble.AsthefreeandforcedconvectionflowinFig.7hasonly
afiniteamountofkineticenergyavailable,assuppliedbythe
injectionflowandcontributedbythedraggingoftheliquidby
therisingbubbles,frictionlossesshouldbeminimisedevery-
where,astheflowisholisticthroughouttheALB–changesin
one section are propagated throughout. In this respect, the
oscillatory flow which was responsible for lower friction in
thepipeworkofthewastewatertreatmentexperimentmen-
tionedinSection2.3maycontributewithintheALBaswell
to lower friction losses. In the underwater visualizations of
microbubblecloudformation,aconcerted,periodicmotionof
toroidal shaped clouds of bubbles were observed, indicating
thattheoscillatoryeffectcanextendbeyondthemicrobubble
generationinthebase.Thesetoroidalcloudswereobserved
atrelativelylowfrequencies,1–5Hz,i.e.longfeedbackloops
∼25m.
3.1.2. ALBriser
Fig.7– SchematicdiagramofaninternalALBwithdraught The riser is the phase transfer work-horse of the ALB. The
tubeconfiguredwithatailormadegroovednozzlebankfed gas–liquid mass transfer and liquid–bioculture mass trans-
fromthetwooutletsofthefluidicoscillator.The ferarethedominantfeaturesofthisregion.Thegassparger
microbubblegeneratorisexpectedtoachievenearly is usually at the bottom of the riser, and the bubbly flow is
monodisperse,uniformlyspaced,non-coalescentsmall responsible for the lowest density in the fluid mixture, and
bubblesofthescaleofthedrilledapertures. fortheco-currentflotationeffectintheALB.Thissectionis
themajortargetforperformanceenhancementfortheintro-
duction of microbubbles. If nutrients are introduced in the
3.1. KeydesignfeaturesofanALB gasphase,suchasoxygenforaerobicmetabolism,thehigher
mass transfer flux should lead to greater bioculture activ-
Inthissubsection,wewillreviewkeydesignfeaturesofthe ity, or conversely, low gas flow rates could be introduced to
ALBwithaviewtotheinfluenceofthemicrobubblegeneration save energy consumption while achieving the same oxygen
mechanism.AfullerreviewisgivenbyJones(2007). transferrate,duetothehigheroxygentransferefficiencyas
Fig.7showstheschematicdiagramoftheinternaldraught discussedinSection2.1.Accordingtothehydrodynamicbub-
tubeALBconfiguredwithatailor-madegroovednozzlebank blechainstabilizationhypothesisbyCrabtreeandBridgwater
suchasdemonstratedinFig.4.Thisisexpectedtotoachieve (1969),weexpect,asinFig.5andillustratedinFig.7,thatthe
nearlymonodisperse,uniformlyspaced,non-coalescentsmall microbubblesgeneratewillbenearlymonodisperseanduni-
bubbles of the scale of the drilled apertures. The cloud of formlyspacedastheyrise.Thisshouldsignificantlychange
dispersed microbubbles should resemble Fig. 5. This is the themultiphaseflowdynamicswithmoreuniformprofilesand
majormodificationoftheALBproposedhere.Theremaining predictablemasstransfercoefficients.Itisthedeformability
features of the internal draught tube, riser, and downcomer andpolydispersityofbubblecloudsthatmakesmasstrans-
regions are conceptually the same as the standard design, ferandmomentumtransfereffectsfromgastoliquidphase
which drives the recirculating flow from buoyant effects – a reliant on empirical correlations. With uniformly spaced,
combinedforcedandfreeconvectionflow,asthereisaninjec- monodisperse bubbles, the mass transfer coefficient can be
tionofmomentum,butalsoofdensitydifference.Thebubbly predicted from level set method modelling such as those of
flow region has lower density, and rises due to a combina- DeshpandeandZimmerman(2005a,b),andthehydrodynamic
tion of buoyant and hydrodynamic forces (Grammatika and effectsbytheconcertedmultibodymicrohydrodynamicsanal-
Zimmerman,1999,2001).Thedowncomerflowisassuredby ysisofGrammatikaandZimmerman(2001).Thegeneralized
thekinematics–iffluidrisesintheriser,thenthebottomof flotation analysis of this paper is particularly important as
theriserisamasssinkandthetopisamasssourcedueto the introduction of smaller bubbles in the riser region can
continuity. Consequently, this drives a flow from the top of leadtoamuchgreaterflotationeffectfromthecollection,by
therisertothebottom,shownbythearrowsinFig.7. microbes,ofsufficientmicrobubblesontheirsurfacestohave
an appreciably larger flotation efficiency, such as the major
3.1.1. ALBbase featureofdissolvedairflotationseparations.Thetailoringof
MerchukandGluz(1999)havethemostpromisingworkonthe themicrobubblescaletoachievethedesiredleveloffluidiza-
ALBbasearea,butuntiltheircontribution,itwouldbefairto tionofthebioculturewithoutcollectingthewholeofthephase
statethatmostresearchershaveviewedthebaseasoflittle atthegas–liquidinterfaceattheheadspaceisamajordesign
consequencetoALBperformance.Asthegasspargerandthe problemforthisnovelALB.Performancedataatthemoment
bubbledistributionarelocatedinthebase,thisisthemajor isrequired,aspreviousstudiesontheeffectofchangingthe
focus of our design, as can be seen in Fig. 7, where our flu- geometricparametersoftheperforatedplatedistributorsys-
idic oscillator and tailored grooved nozzle bank distribution temhavelargelybeenabouttheuncoveredareaoftheplate,
systemhasbeensubstitutedforthetraditionalbase.Theonly notthesizedistributionofthebubblecloudsgenerated.
hydrodynamic issue otherwise about the base how the liq- Typicallygas–liquidmultiphaseflowmayhaveawiderange
uidfromthedowncomerregionisdrawninthroughthebase of regimes, but dispersed or bubbly flow is common in all
224 foodandbioproductsprocessing 87 (2009) 215–227
applications. These possible flow regimes are dependent on
thegeometricalconfigurationandthegasandliquidsuperfi-
cialvolumetricflowrates.Theinfluenceonflowregimeofthe
microbubblegenerationsystemisnotknownapriori.Never-
theless,itwouldseemhighlylikelythatitwillhaveastrong
influence,forinstance,thetoroidalmodeofbubblecloudfor-
mationobservedinairdiffusertrialsdiscussedbeforehaveno
analogueinthestandardpipelineflowregimesasthebubbles
arepracticallymonodisperseandnon-coalescent,yetother-
wiseexpectedtobeintheheterogeneousflowregime.Clearly,
the observed flow regimes will depend on the liquid viscos-
ityforlowloadingofmicrobes,andontheeffectiverheology
of the suspension when there is a large volume fraction of
microbes.TherobustnessoftheflowdiagraminFig.7mustbe
testedexperimentally,particularlywithregardtoentrainment
inthedowncomerregionandchannellingintheriser.
TheperformanceofanALBislikelytobeaffectedbythe Fig.8– Processflowdiagramoftheairliftbioreactorgas
modificationoftheflowregimes,asthebiologicalgrowthhas separator.
beenshowntobesensitivetoflowregime(Vialetal.,2000).Our
ownbiologicalgrowthstudiesfocussedontheyieldofyeast perature control is to be achieved through a cooling fluid,
biomassfrommicrobubblegenerationwithfluidicoscillation, preferablywater,andutilizationofacoppercoil.Thecoilis
whichwasshowntobe15%higherthangrowthundersteady placedinsidethebioreactortosupporttheinternalbafflethat
flow(Zhang,2007).However,welimitedourgrowthstudyto actsastheflowguide.Thetemperatureandthepressure/flow
thebubblyhomogeneousflowunderlowflowrateconditions, rateofthecoolantaretobemonitoredcontinuouslytocontrol
wherethehighergrowthratecouldbeattributedtohigherdis- theflow.Athermocoupleisplacedinsidethebioreactorasthe
solvedoxygenlevelsfromhighermasstransfercoefficients, controlprobethatregulatesthecoolantcontrolvalveusingan
nottheflowregimeitself. externalcontroller(LabViewPCIcontrolcard).Theflowinthe
temperatureregulationcircuitisdesignedtoswitchbetween
3.1.3. ALBdowncomer hotandcoldcoolantstreams,addingtheflexibilitytousethe
This is probably the least dynamically important region of samecircuittoheatthesystemifnecessary(Fig.8).
theALB,asthedowncomerflowisdeterminedbykinematic The compressed air stream comes from a constant head
considerations, and its composition and bioculture occupa- compressorisconnectedtothediffuserthroughafluidicoscil-
tionaredependentontheparticularsoftheriserdesign.This lator that can be tuned to adjust the oscillation frequency
region has the highest multiphase density, largely occupied in the range of 1–100Hz. The novelty in the process is the
withliquid,perhapswithhighdissolvedoxygen(gas)concen- inclusion of the oscillator to generate smaller bubbles. As
tration,butthegasphaseholdupisexpectedtobelargeonly a comparison of aeration levels between oscillated and the
ifmicrobubblesaresmallenoughtobeentrainedintheliq- steadyairflowhastobemade,thecontrolandthemeasure-
uidflowaspassivescalars.Suchentrainmentisunlikely,for ments of the air flow circuit is crucial to the experiments.
instance,withpurelyfreeconvectiondrivenflow,asnoneof Both the pressure and the temperature are to be measured
thebubblesshouldbesufficientlylargetodisrupttheorderly upstream and downstream from the fluidic oscillator. The
flowasdepictedinFig.7.However,ifthereisastrongforced oscillatorhasthenozzleeffectthatexpandstheflowandthe
convectioncomponent,itispossiblethatthebubblerisewill temperaturechangeandthepressurelossacrossithastobe
sufficiently transfer momentum to the liquid that the liq- considered carefully in the comparison. This means that a
uidphaseflowwillbestrongenoughtoentrainsomeofthe slightlyhigherpressurehastobeemployedwhentheoscil-
microbubbles in the downcomer. Clearly, this aspect of the lator is in place as the time averaged pressure head to the
novelALBdesignposedhereismuchdifferentthanincon- diffusershouldbesameforbothcases.Tothisend,apressure
ventionalALBs. regulatorisusedbetweenthemainsandtheoscillator.Tem-
peratureofthecompressedair,whichisdifficulttocontrol,is
3.1.4. Gasseparator tobemeasuredpriortothediffuserandthesurfacetension
Theregionatthetopoftherisertothetopofthedowncomeris andthevolumeexpansion/contractionaretobecompensated
termedthegasseparator.Itisnotourintentiontochangeany- inthecalculation.
thinginthedesignofthissensitivearea,butitiscommonly Thebioreactordischargesitsloadtoaholduptankatthe
acceptedthatthisisthemostsensitivepartoftheALBdesign. base level for further processing. A centrifugal or a positive
Residencetimeoffluidinthegasseparatordependsglobally displacement pump is to be used to pump the slurry to the
onthedesignandontheconditions,particulargasholdup,in tankdependingonthethickness.Forthinslurry,thegravity
theriseranddowncomer.MerchukandSiegel(1988)discusses flowwouldbesufficient.
manyofthekeyaspectsofgasseparatordesign.
3.3. Bioreactordesign
3.2. Flowcircuitdesignandinstrumentation
Thecapacityofthetankis250landanexcessof10%ispro-
Intheequipmentdesign,theprocessflowintothebioreactor vided as a safety measure to prevent spills. It consists of
hasbeenconsideredcarefullyfortightcontrolofthetempera- a conical bottom with an angle of 75◦ a cylindrical section
tureandairflow.Thisisdeemednecessaryforthecomparison and a bolted end plate with a vent. If the anaerobic condi-
oftheperformancewithandwithoutoscillations.Thetem- tionsaretobemaintainedtheflowthroughtheventcouldbe