5.

FLOW IN THE CYLINDER

by John L. Lumley, Cornell University (excerpt from his book:
Engines an introduction. Cambridge University Press, 1999)

INTRODUCTION
Thegasflowinthecylinderhasaprofoundinfluenceontheperformanceoftheengine,
whether intended or not. In the early days of the automobile, it was seldom intended,
because relatively little was understood about turbulence, the primary player in this
drama. Engines were designed largely by empiricism; from time to time a particularly
successful design emerged, and its characteristics (to the extent that their relevance was
recognized)werepreservedinsubsequentengines.
Thereareacoupleofnotableexceptionstothis:Inthefirst,between1903and1907H.
R.Ricardo[45](thefatheroftheoctanenumber),workingatCambridgeUniversitywith
Professor Bertram Hopkinson, did pioneering work on the effect of turbulence on
combustionandheattransferintheICengine,particularlyontheeffectoftheincreased
effective flame speed on knock, and on the possibility of stratified charge, among other
things.Thisled,duringtheFirstWorldWar,togreatimprovementsinthedesignoftank
engines (giving short flame travel and high turbulence levels, permitting higher
compression ratios without knock), and after the war led to design modifications of the
flathead, orsidevalve engine, which resulted in the same performanceas the overhead
valveengine,andwhichweregenerallyadopted,andresultedinpatents.
The other notable exception involved measurements made of the swirl and tumble
produced by various inlet configurations, and the effect of the swirl and tumble on
combustion[63],[86].Theflowmeasurementsweremadeusinghighspeedphotography
of goose down (!) in a glass cylinder. The combustion measurements used schlieren
techniques.Theauthorsdeterminedthemajormeanflowandturbulentcharacteristicsof
swirlandtumble,andtheireffectsoncombustion(wewilldetailthislater).
TheworkofRicardo(above)andthisprewarworkwerelargelyforgotten,andhadto
berediscoveredduringthe1980sand1990s.
An example of an early engine designed largely on an empirical basis is the DOHC
pentaheadengine,whichwasdevelopedforracingduringtheteensofthecentury,and
was afterward extensively used in aircraft. It was designed principally to maximize the
valvearea,whichkeepstheMachindexaslowaspossible,althoughtheMachindexwas
not understood at the time. It was known to be particularly successful, but it was not
understooduntilrecently(exceptfortheworkofRicardo[45]andtheNACAwork[63],
[86])thattheorientationoftheinletvalvesinducestumble,whichisthenbrokenupasthe
pistonapproachesTC,resultinginhighturbulencelevels,andhigheffectiveflamespeeds.
The central location of the spark plug also gives relatively short travel distances for the
flamefront,whichRicardo[45]understood.Thecombinationofshortdistancesandhigh
effectivespeedsresultsinshortburntimes,whichmeansthatthecompressionratiocanbe
increasedatthesameoctanenumberwithoutknock,oneofRicardosbasicfindings[45].
In modern engines, short burn times can be taken advantage of in other ways. For
example, higher exhaust gas recirculation rates (EGR) could be tolerated, resulting in
higher efficiency and lower NOx production. This would be preferable to increasing the
compression ratio in a modern engine, since an increased compression ratio could raise
theunburnedhydrocarbonemissions,becausethecrevicevolumewouldbecomealarger
fractionofthetotalvolume.Wewilldiscussallthesepossibilitiesinduecourse.
Now, engines are designed in an attempt to consciously bring about some of these
effects. Unfortunately, our ability to apply existing limited understanding of fluid

mechanics and turbulence to the complicated situation in the cylinder is still fairly
rudimentary,andwhiletheprocessismorerationalandlessempiricalthanitusedtobe,
itstillhassomewaytogo.
Also,enginedesignersaretryingtodoratherambitiousthings.Thesecomeunderthe
general heading of flow management, with the goal of reducing brakespecific fuel
consumptionandemissions.Thereisacomplexinteractionamongtheflowmanagement,
thecharacteristicsofthecatalyst,theemissions,andthefuelconsumption.
For example, the lean burn engine burns a homogeneous mixture of perhaps 24:1.
Suchamixtureisrelativelydifficulttoburn,andrequiresahighturbulencelevel.Exhaust
gas recirculation (EGR) is another technique to meet the same goal. EGR keeps
approximatelythesameproportions,butreplacessomeoftheexcessairwithrecirculated
exhaust gas, which has the advantage that a 3way catalyst remains effective (current
catalysts are not effective when the gases contain an excess of oxygen). Again, high
turbulencelevelsarerequiredtoobtainreliablecombustion.
The stratified charge engine attempts to segregate the incoming fuel vapor so that it
does not mix with all the air in the cylinder, so that the engine will run at an effective
air/fuelratio of perhaps 50:1,while the fuelis confinedto a small region of the cylinder
volumewheretheair/fuelratioisonly15:1,whichwilligniteandburn.
Stratifiedchargeenginesnowaredirectinjectionengines;thatis,theyinjectthefuelspray
athighpressuredirectlyintothecylinder,usingvariousstrategiestokeepthefuelspray
from mixing with the entire contents of the cylinder. In the early development (say,
twenty years ago) of the stratified charge engine, various techniques (other than direct
injection) were used to segregate the charge, and they were not very effective. We will
discusssomeofthesetechniqueslater.Suchenginesdidnotachieveair/fuelratiosmuch
abovethoseoftheleanburnengines,andarenowthoughttohavebeenprobablyaboutas
homogeneous as a poorly managed homogeneous charge engine. Terms like nearly
homogeneous,orweaklystratifiedmightbetterbeusedtodescribetheseengines.The
suggestionhereis,thathomogeneouschargeengines,unlessconsiderablecarebetaken
toinducehomogeneity,arenotparticularlyhomogeneous.
Leanburnresultsinalowercombustiontemperatureoverall,producingfeweroxidesof
nitrogen. Because the mixture is approximately stoichiometric or slightly rich in the fuel
cloudinthestratifiedchargeengine,theoxidesofnitrogenarenolowerthanthoseofthe
leanburnengine,althoughtheair/fuelratiomaybemuchhigher.However,combustion
quality is poorer than in stoichiometric homogeneous engines (due to slower burn and
lowertemperature),andasaresulttheemittedhydrocarbonsarehigher.COislowerfor
the lean burn engine, but rises again for the stratified charge engine, due to the poor
combustion.Moreover,lowerlevelsofNOxoutoftheenginedonotnecessarilyresultin
lowerlevelsofNOx outofthetailpipe,becausetheconversionefficiencyofthecatalystis
lowintheexcessairofaleanburnorstratifiedchargesystem.Bycomparison,EGRoffers
fuel economy and engineout NOx benefits comparable to a mildly lean combustion
system,butallowstheuseofathreewaycatalystforoxidationofhydrocarbonsandCO,
and reduction of NOx. This is also a costeffective approach. Note that, for $150 Ryobi
offersa4strokecyclegasolineweedwhackerwithEGRLosAngelesCountytakenote!
(LosAngelesCountyrecentlypassedanordinanceoutlawingweedwhackersbecauseof
theirhighemissionslevel.)
Foratleast the past fifteen years, the motivation for lean stratified charge engines in
the automotive industry has been improved fuel economy. Not only are HC and CO
emissionshigher,butonehastostartworryingaboutparticulates,asinadiesel,because
ofthepoorcombustion.ThefewleanburnenginesthatcurrentlymeetFederalemissions
standards operate lean over only a very small portion of theirspeedload map. None of
themmeetsU.S.50statestandards.

 Managementofflowintheenginecylindertobringaboutanyoftheseeffectsisvery
trickytodo, aswe will see,and the fluid mechanical detailsofhow it works are poorly
understood.

PHASES OF THE FLOW

Theflowinthecylindercanbedividedintoseveraldistinctphases.Theflowintothe
cylinder through the inlet valve or valves (forming a jet) does two things: first, the
geometrical configuration of the inlet ports and the valves, and their opening schedule
creates organized motions in the cylinder, known as swirl (about the cylinder axis) and
tumble(orthogonaltothecylinderaxis);second,thejetitselfisturbulent,andinaddition
muchofthedirected(nonturbulent)energyinthejetisconvertedtoturbulence,resulting
inaveryhighturbulencelevelduringtheinletstroke.
Duringthesecondhalfoftheinletstrokemuchofthisturbulencedecays,whichisto
saythattheintensitydecreasesmarkedly,bothbecausethesource(thejet)iscomingtoan
end,andduetotheeffectsofviscosity.Theorganizedmotionsarecarryingfuelvaporand
droplets,residualgases,andallthecontentsofthecylinder,downthesidesandacrossthe
bottomanduptheotherside,androundandround(dependingonhowmuchswirland
tumble there is). In addition, the turbulence is spreading itself, organized momentum,
fuel, residual gases, in fact anything that can be transported, throughout the cylinder,
trying to make everything as uniform as possible. This is what turbulence does best
(transporting things), and it does it many orders of magnitude better than molecular
transport.
Duringthecompressionstroke,theincreaseofdensityandthechangesinlengthscales
(due to the change in geometry of the charge as it is compressed) have the effect of
amplifyingtheturbulencewhichremainedfromtheinletjet,althoughtheviscousdecay
and turbulent transport continue. In addition, the swirl and tumble are affected by the
samephenomenon.
Ifthepistonandcombustionchamberhavebeendesignedtoproducesquish(thatis:if
the piston crown approaches very close to some part of the combustion chamber roof),
thenthiswillhavetwoeffects:first,thefluidsqueezedoutofthesquishclearancevolume
willproduceorganizedmotions,mostofwhichwillbreakupintoturbulence;andsecond,
the change in geometry due to the squish will have dynamical effects on the organized
tumbleandswirl.
Near TC, some of the organized motions may find they have insufficient room to
maintain their form, and they will break up into turbulence, increasing the turbulence
level.Bythistimeconditionsinthecylinderhavebecomecrudelyhomogeneous,dueto
the transporting effect of the turbulence and the organized motions, unless a concerted
efforthasbeenmadebythedesignertoavoidthis.Bearinmindthatthetimeavailablefor
transport (measured by the time scale of the turbulence itself) is not large, so the
uniformityachievedcanonlybecrude.
During combustionthe turbulence levelrises somewhat. Then, during the power stroke,
the geometrical changes result in a strong attenuation of the turbulence, and any
organizedmotionthathassurvived.This,combinedwiththeviscousdecay,resultsinthe
turbulencebeingsharplysuppressed,sothatbythetimetheexhaustvalveopens,thereis
virtuallynothingleft.Verylittleturbulenceisgeneratedduringtheexhauststroke.
Wewillexaminethesevariousphasesonebyone.

AVERAGING

Weneedtotalkfirstaboutaveraging,sinceturbulenceisdefinedbyfluctuationsaboutan
averagevelocity.

If we were dealing with a process that was not changing statistically from instant to
instant (even though the instantaneous values were chaotic) such as the wind speed at
midday on a day without severe changes in the weather, then we could use a time
average, integrating over a time long compared to a typical time over which the values
fluctuate. Such a situation is called stationary. Unfortunately, the situation in an engine
cylinderisnotstationary.Successivecyclesaresimilar,butwithineachcyclethesituation
isstatisticallyquitedifferentfrominstanttoinstant.
In such a situation, it is useful to use a phase average. Imagine measuring, say, the
circumferentialvelocityUataparticularlocationinthecombustionchamber.Thiswillbe
afunctionofcrankangleandoftheparticularcycleinwhichwemeasuredit,say

where i indicates the number of the cycle in which it was measured. Then the phase
averagedvalueisdefinedas

where is the number of measurements available. There are a number of interesting
questions which wecould address, such as, how many individual valuesdo we need to
include in Equation 5.2 to arrive at a stable value? This and many other questions are
coveredin[96],andneednotconcernushere.
Now that we have a convenient definition of an average value, which is a function
only of crank angle, we can define the turbulent fluctuating velocity as the difference
betweentheinstantaneousvalueandtheaveragevalue:

StatisticsofU(,i)cannowbedefinedas(forexample)

themeansquarevalue.
Notethatthemeanvaluesdefinedinthiswayarefunctionsonlyofthecrankangle.There
is another way to define an average in this situation. Consider the time scale of the
turbulence.Forexample,wewillfindbelowthattheturbulenceduringthemiddlethirdof
the intake stroke has an intensity (the rootmeansquare value, from Equation 5.4) of
roughly10 ,where istheaveragepistonspeed,andalengthscaleofroughlyb/6[98],

wherebisthebore.Thismeansithasatimescaleofroughly .Thistimescaleis1/60

the time it takes the piston to complete the intake stroke (presuming that the engine is
nearlysquare,bS).Now,whatcanwedowiththis?Ittellsusthatthereisatime,say
S/ 60 60 8 times larger than the times typical of the turbulent
, that is about
fluctuations,andyetshorterthan(about1/ 60=1/8)thetimeoftheintakestroke.

where 608.Wecanusewhatiscalledamovingaverage:

where=S/8Vp.Wecancallthistheindividualcyclemean.Figure5.1showsacartoonof
suchaninstantaneousvelocity,thephaseaverage,andtheindividualcyclemean.

Thisquestionofwhetherthereisanintermediatetimescale,largerthanthesmallscale
of the turbulence, but shorter than the longer scale of the intake stroke, is the same
question that arises in meteorology, whether there is an intermediate scale between
turbulence and weather, that would permit a distinction between them. It also arises in
viewingpicturesinanewspaper.ThedistancebetweentheBendaydotsthatmakeupthe
pictureisthefinescale,andthescaleofthedetailinthepictureisthelargescale.Forthe
picture to be interpretable by our eye, a scale must exist that is large compared to the
Benday dot scale and small compared to the scale of the detail. The situation is only
unambiguous when there is a wide separation between the scales. There are many
interestingstatisticalquestionsthatcanbeaskedaboutthissituationsomeofthemwere
discussedin[66].
Heywood[47]saysthereisconsiderabledebateintheliteratureoverwhetherthecycleto
cyclevariationintheindividualcyclemeanisaphysicallydistinctphenomenonfromthe
turbulence,whichwehavedefinedasthedepartureoftheinstantaneousvelocityfromthe
phase average, but which could also have been defined as the departure of the
instantaneous velocity from the individual cycle mean. The way we have defined it, it
includesboththisotherdefinitionofturbulence,andthecycletocyclevariation.
Cycletocycle variability can be related to the turbulence in the inlet manifold. To
begin with, we have evidence that the flow in the cylinder at TC (and after) is very
sensitivetothevelocityfieldinthecylinderatthemomenttheinletvalvecloses,whichis
determined by the flow in the inlet manifold while the cylinder was filling. Exponential
dependenceonsmalldifferencesininitialconditionsisacharacteristicofturbulence[49].
Inthecaseofflowinthecylinder,thedependenceontheinitialconditionsisknownboth
from computational evidence and from experimental evidence. In [80] the authors
examinetheflowinasinglecylinderenginewithapancakeshapedcombustionchamber,
with an inlet and exhaust manifold. The inlet valve is shrouded, to produce high swirl;
such a configuration also has relatively little cyclic variability of the large scales. The
authors calculated the mean flow structure, modeling the small scales of the turbulence.
Theyfoundthattheswirlratio,andthephasingoftheswirlstructure90CADATCwas
verysensitivetotheinitialvelocityfieldatinletvalveclosing.
In [79], the author made measurements of the flow in the same engine, this time
without the shroudon the inlet valve, so thatthe flow isundirected. The measurements
were made by introducing very small particles that follow the flow, and photographing
them at closelyspaced times, obtaining velocity vectors from the differencesin position.
ThisisknownasParticleImageVelocimetry,orPIV.InFigure5.2Ishowthisflow.From
Figure 5.2 it is clear that there is substantial variation in the large scale structure at TC.
This variation in large scale structure will affect the flame propagation and heat release,
andmustconsequentlyberesponsibleforthecycletocyclevariability.

 In [43] the author carried out preliminary calculations of the unsteady flow in this
sameengine,includingtheinletandexhaustmanifolds,resolvingonlythelargestscales
(modeling the smaller scales). Again, he found large variability of the large scale
structures, completely consistent with the PIV measurements and the mean flow
calculations.
The flow is deterministic, and the flow at TC is entirely determined from (and very
sensitive to) the flow at the inlet valve closing. Since this is entirely determined by the
flow in the inlet manifold during filling, it must be this flow that is responsible for the
cycletocyclevariability.Theflowintheinletmanifoldisverycompressibleaswehave
seen, and large amplitude sound waves are present in the manifold. It might at first be
thoughtthatthesewereresponsibleforthefluctuationsinmanifoldflow.Measurements
ofthepressurefluctuationsinthemanifoldofa2.2LRenaultJ7T(fourcylinders,withtwo
valves per cylinder) at several engine speeds [14], however, show that the pressure
fluctuations (and hence the sound waves) are quite periodic; the very small differences
fromcycletocyclearebelowtheresolutionofthepressuretransducer.
Wehavetoconcludethatthedifferencesintheinitialvelocityfieldinthecylinderwhen
the inlet valve closes are attributable to the turbulent velocity fluctuations in the inlet
manifold.Thesemayhaveanindirectinfluencealso,sincethevelocityfluctuationsaffect
thepressuredropacrosstheinletvalveduringfilling,andmayhaveasmalleffectonthe
sound waves in the manifold. The relatively small differences at inlet valve closing
attributabletothemanifoldturbulenceareenoughtocauseconsiderablevariabilityatTC,
duetotheextremesensitivityoftheturbulencetoinitialconditions.

Note that the various kinds of average (the phase average, and the moving average,
which could also be applied in space within the engine cylinder) are not statistically
equivalent. It is necessary in any discussion to be clear about which average is being
discussed.Sincethecycletocyclevariationandthesmallscaleturbulenceinthecylinder
appeartobephysicallythesame,differingonlyinscale,itprobablymakessensetolump

them together, because the cycletocycle variations are also responsible for transport in
thecylinder.Intherestofthisbook,Iwillusethedefinitionofturbulenceasthedeparturefrom
thephaseaverage.
Thisdoesnotmean,however,thatthecycletocyclevariationshouldbeignored;itis
responsibleforunevennessintherunningoftheengine,andisgenerallyundesirablefor
that reason. In addition, if there is cycletocycle variation, some of the cycles will be of
lower efficiency, and this will affect the average efficiency. I would rather describe the
situationthisway,however:thatitisdesirabletosuppressasmuchaspossiblethelargest
scale variability of the flow entering and in the cylinder (which will be responsible for
perceptible unevenness and fluctuation in efficiency, the longer, the more perceptible).
Thissaysthesamething,buttheemphasisisalittledifferent.

A WORD ABOUT TURBULENCE

The reader will need to understand just a little about turbulence for this discussion to
makesense.Itwillnotbenecessarytoconsultatextonthesubject,suchas[96].
Allfluidflowiseitherlaminarorturbulent.Theflowofpancakesyrupisanexample
oflaminarflowsmooth,steadyanduncomplicated.Whenthekitchenfaucetisturnedon
fullandtheflowleavesthefaucetinanunsteady,rough,chaoticmanner,filledwithsmall
eddies,thatisturbulentflow.Mostoftheflowintheuniverseisturbulent.
Whether a flow is laminar or turbulent is usually (and certainly in our case)
determinedbyaReynoldsnumber,whichcomparesinertialtoviscousforces.IfU(m/s)is
atypicalvelocityintheflow,L(m)atypicallength,and(m2/s)thekinematicviscosity,a
Reynolds number is defined as UL/. It is dimensionless. In any given flow, there are
usually several Reynolds numbers that can be defined, depending on what velocity and
lengthscalesareused,buttheyareallrelated.Theflowofpancakesyrupisdominatedby
viscousforces,whilethewaterflowisnot.TheReynoldsnumberofthepancakesyrupis
low, while the Reynolds number of the water flow is high. If the Reynolds number is
large,theflowwillprobablybeturbulent.
AstheReynoldsnumberincreases,alaminarflowbecomesunstable.Thatmeans,that
small disturbances, which would have been suppressedby viscosity when the Reynolds
number was smaller, will instead be amplified, and the flow will become chaotic and
unsteady.Inthisway,energyinthemeanflowistransferredtotheturbulence.
Turbulent flows contain a wide range of scales that is, the velocity field contains
motions of all sizes, from eddies that are essentially large enough to fill the space
available, in our case the engine cylinder, down to eddies often substantially below a
millimeterinsize.Thesizeofthelargesteddycanbeguessedbyaskingforthediameter
of the largest sphere that will fit in the available space since turbulent eddies are
approximately the same size in all directions. Hence, at TC the largest eddy will be
roughlytheclearanceheight,whileatBCitwillberoughlythecylinderbore.Thekinetic
energyoftheseeddiesvaries:thelargesteddiesarerelativelyweak;asthesizedrops,the
energyrisesrapidlytoapeak,andthenfallscontinuallydowntothesmallesteddies.The
mostenergeticeddy,atthepeak,whichisresponsibleformostofthetransport,isabout1/6
thesizeofthelargesteddythus,1/6ofthebore,or1/6oftheclearanceheight[98].
Turbulenceconsumesenergy.Meanflowenergyisconvertedtolargescaleturbulence
and those large scale motions are themselves unstable, and break down into smaller
scales,andthoseintheirturnbreakupintostillsmallerscales,untilthesmallestscaleis
reached. It is a curious experimental fact that the amount of energy consumed in a
turbulentflowhasnothingtodowiththeviscosity,butisdeterminedentirelybythelarge

turbulentscales.Thatis,theamountofenergytakenfromthemeanflowisdeterminedby
theprocessofcreatingthelargestturbulenteddies.Thisamountofenergyispassedonto
thenextsmallereddies,andontostillsmallereddies,andisfinallydissipatedbyviscosity
at the smallest scales. The amount of energy passed from the mean flow to the largest
eddiesandontoeddiesofprogressivelysmallersizeiscalledthedissipation,theenergy
consumed per unit mass of the flow, and is designated by . If u is a turbulent velocity
scale,therootmeansquarefluctuatingvelocity(typicalofthemostenergeticeddies),and
lisalengthscaleofthemostenergeticeddies,thenitisanexperimentalfactthat

You can understand this by writing u3/l = u2[u/l). u2 is proportional to the energy of the
turbulence,andl/uisproportionaltothenaturaltimescaleofthemostenergeticturbulent
eddies,whichdetermineshowfastthemostenergeticeddiesloseenergytothenextsize
eddies.Fromthisyoucanseethatu3/lislikedu2/dt.
The dynamical behavior of the smallest eddies is determined entirely by the rate at
whichtheyarefedenergy, , andthesizeofthekinematicviscosity.Thesmalleddies
are where the mechanical energy is converted to heat, although the rate of conversion is
determinedbythelargescales.Iftheviscosityisraised,doesnotchange,butthescaleat
which the dissipation takes place increases. The smallest scale is called the Kolmogorov
microscale,designatedby,anddefinedas

Smallerscalescannotexistbecausetheviscoustransportsmearsthemout.Inmostflows
theKolmogorovmicroscaleisafewtenthsofamillimeter.Wewillfindthatinanengine
cylinder,itisoftheorderof0.01mm,or10m.
Asnotedabove,inanenginecylinder,weexpectthemostenergeticscaletobeofthe
orderofh/6ifthepistonisnearTC,oroftheorderoftheb/6ifthepistonishalfwaydown
or more. Consider a cylinder halfway between TC and BC on the intake stroke, and
supposetheengineisrunningatapistonspeedof5m/s.Assumeaninletdensityof=
0.5 kg/m3,and suppose S=b. We will find in Section 5.5that the turbulent velocitywhen
thepistonisinthemiddleoftheintakestrokeisabout10timesthemeanpistonspeed,or
perhaps50m/shere.Weexpectthatl= b/6,andletustakeb = 82.6mm.Thisthengivesus
an 9 MW/kg,giving7.4106m=7.4m.Asturbulentflowsgo,intheenvironment
andthelaboratory,thisisextraordinarilysmall;intheatmosphereortheoceanisabout
1mm,andinlaboratoryexperimentsnotmuchsmallerthanafractionofamillimeter.
Thevalueof 9MW/kgsoundsenormous,butbearinmindthatthemassinacylinder
islessthanagram,thatthislevelofdissipationdoesnotlastlong:roughly 5.5ms,
anditresultsinatemperatureriseoflessthan50K.Forhigherpistonspeeds risesvery
3
fast,since .willfallslowlyas .
We can look at the cylinder at TC. Now l is 1/6 of the clearance height, h. If r = 8, the
clearanceheightisapproximatelyh = 11.8mm.Hence,l = h/6 2mm.Wehavealready
seeninSection3.3.4thatatTC,intheabsenceofswirl,tumbleandsquish,u .

The kinematic viscosity in a gas is proportional approximately to . With r = 8, the
isentropic temperature atthe end of compression hasincreased by afactor of 2.3, while
thedensityhasincreasedbyafactorof8,givingavalueforthekinematicviscosityof=

2.8410 6 m2/s.With =5m/s,thisgives 7.9kW/kg,andavalueof7.34m,still

aboutthesamesize,andquitesmallbynonautomotivestandards.
Notice,ifweusebothourEquations5.7and5.8,wecanwrite
where Re = u l / is the Reynolds number of the turbulence, based on the turbulent
fluctuatingvelocityandthelengthscaleoftheenergycontainingturbulenteddies.Notice
thatthisReynoldsnumberdeterminestheratiooftheenergycontainingandsmallscales

in the turbulent flow; when Re 1, the energycontaining and small scales are
approximatelythesame,andatthispointtheturbulencecannolongermaintainitself.To
exist, turbulence needs to have a range of scales; when the small scale and the energy
containing scale are the same, this means that viscosity acts directly on the energy
containingscales, and that makes the turbulence too dissipative, andit dies. Atour two
states,halfwaydownontheintakestrokeandatTC,wehavescaleratiosof l /1.910 3
and 1.6 10 2 , for turbulent Reynolds numbers respectively of Re 2.3 10 4 and Re 8.7
10 2 .
Inagas,measuresthemoleculartransportofmomentum,c,wherecisthe
rootmeansquare molecular thermal velocity, and is the mean free path. That is, is
proportional to the product of the velocity scale and the length scale of the process
responsible for the property transport, the molecular motion. In a turbulent fluid, the
turbulence transports properties, by simply carrying the fluid containing them to a new
location. This transport process is nothing like (physically or mathematically) the
molecular transport; nevertheless, it is often convenient to parameterize it by what is
calledaneddyviscosity, T .Justlikethemoleculartransport, T ,theturbulenttransport
coefficient, is also proportional to the product of the length and velocity scales of the
physicalprocessresponsibleforthetransport,theturbulence.Hence, T u l .Wecanwrite

sothat,inourtworepresentativesituations(halfwaydownontheintakestroke,andatTC
onthecompressionstroke)theturbulenceisapproximately10 4 and10 3 timesaseffective
attransportingmomentum(andeverythingelse)asthemolecularmotion.Valuesof T in
the two situations are T 6.9 10 1 m2/s and T 1.4 10 2 m2/s, far larger than the
molecularvalues.
During a time t, the turbulence will carry a property (momentum, for example) a
distanceLofroughly[96]

Considerthesituationhalfwaydownontheinletstroke.Ifwemeasuretintermsofthe
timetocompletetheinletstroke,say1/3ofthattime,whentheturbulenceismostintense,
t=S/3 =,then

Thus, this intense turbulence is capable of transporting momentum across a substantial

fractionoftheenginecylinderinthetimeavailable.Notethatittransportsfuelvapor(or
anything else) just as well as momentum, since the transport is physical the fluid
containing the momentum or fuel vapor is simply moved to another place, so the
effectiveness of the transport does not depend on the nature of the thing being
transported,anditworksaswellforallproperties.
Ontheotherhand,atTC,theturbulenceismuchweaker,andthetimeavailable
ismuchshorter.Theburntimecorrespondstoroughly0.28S/ =,sothat

soonlyafractionoftheclearanceheightcanbecoveredinthetimeavailable.
There are, of course, other factors involved, and these are extremely approximate
values,intendedonlytogivetheflavorofwhatisgoingon.Itisclear,however,thatthe
turbulence available during the burn is marginal, considering that it is primarily
responsible for the flame propagation. It is equally clear that it will be very difficult to
keep the fuel vapor segregated during the inlet stroke, in the face of the aggressive
transportpropertiesoftheturbulenceavailablethen.
ESP (see Chapter 8) assumes that the charge is homogeneous that the aggressive
transportduringtheintakestrokehasresultedinuniformity.Tocalculatetheturbulence
level,itusesturbulencemodelsthatare(insomerespects)alittlemoresophisticatedthan
ourcrudeapproximations,andinotherslessso.

TURBULENCE INDUCED BY THE INLET JET

I will assume here that we are discussing a homogeneous charge engine, without
significantswirlortumble,whichIwilladdresslater.
The flow through the inlet valve forms a hollow conical jet, roughly at the angle of the
valve seat. We may expect that the Reynolds numbers will in general be high, since the
viscosityofairatroomtemperatureisrelativelylow.Usingourestimatefortheaverage
velocitythroughtheinletvalveEquation2.15,andthewidthofthegap(assumingavalve
seatangleof45),takingconditionsatidle, =2m/sandp=0.4bar,wegetaReynolds
number of about 6 10 , which will still guarantee that the jet is turbulent (since the
3

Reynoldsnumberbasedontheturbulentvelocityandlengthscalesu andl, whichisthe

onethatcounts,willbeintheneighborhoodof610 2 1,[96]).TheReynoldsnumber
willbemuchlargerathigherpistonspeeds.Thisjetisissuingintoanenclosedspace,soit
strikes the piston crown and the cylinder walls. It forms secondary vortices from the
interaction with these surfaces, and these ultimately are dissipated or break up into
turbulence.Theimpingingjet(asitiscalled)alsoformsturbulentboundarylayersonthe
surfaces.Thejetisitselfunstable,andfluttersbackandforthirregularly,transformingthe
meanflowenergyinthejetintoturbulentenergy.
Wecanmakeacrudeestimateofthetotalenergyintroducedbythejet.Ifthisenergywere
alltobeconvertedwithoutlosstoturbulence,andspreaduniformlythroughthecylinder,
itwouldgiveusaroughestimateofthepeakturbulenceintensitywecouldexpectduring
the inlet stroke. It would be a considerable overestimate, because we are assuming no

losses;inSection5.4wesawthatthelosses(dissipatione)arelarge.Wecanbeginfrom
ourestimateforthemeanvelocitythroughtheinletvalve,Equation(2.15).

10

Now, we want to calculate the total kinetic energy that enters the cylinder in the jet
throughtheinletvalve.Letusmakeourestimatewhenthejethashalffilledthecylinder,
inthemiddleoftheinletstroke,whentheturbulencefromthejetisatitspeak.

Thefirsttermis

The second term is the total volume that has entered the cylinder, which is just half the
displacementvolumeVd/2.Hence,thetotalkineticenergyis

Ifthisenergyisspreaduniformlyoverthecylindervolume,wegetatotalenergyof

Equatingexpressions5.17and5.18,wehave

sothat

In[47]thepeakvalueof u/ isgivenasabout10.Using Equation5.20,estimatingCi =
0.35,andusingthevalveandcylinderdiametersin[47],weobtain12.4.
Obtaining a value that close from my simple calculation must be regarded as
largely an accident. I neglected a great deal, notably the dissipation, which we have
alreadyseeninSection5.4isquitelarge,andshouldresultinaconsiderabledecreasein
theturbulentvelocity.
Wehavebeenconsideringanenginewithoneintakevalvepercylinder,forthe
sakeofsimplicity.Inamodernenginewearelikelytohavetwo.Roughlyspeaking,this
will cut the velocity through the jet by a factor of two, and hence also cut the turbulent
velocitybyafactoroftwo.Aspistonspeeddrops,theturbulentvelocitymaynolongerbe
largeenoughtoprovideashortenoughburntimetoavoidknock.Thesolutionistoopen
onlyoneoftheintakevalvesatlowerspeeds,tobringthejetvelocity,andtheturbulent
velocity up again. This is what Honda does, for example, in the VTEC engine (although
thatisalsodonetoinduceswirl).
Before we move on, we should consider decay of turbulence. The turbulent
energybalancelookslikethis:

whereq2/2istheturbulentmeankineticenergyperunitmass,isproduction(therateat
which energy is being extracted from the mean flow and fed into the turbulence), is
transport (the rate at which energy is being moved to another place by the turbulence
itself)andDisdissipation,D=e,therateatwhichtheturbulentkineticenergyisbeing
converted to heat by viscosity. We do not need to concern ourselves with the
mathematicalformsofthesetermsnow.Duringtheinletstroke,theturbulenceawayfrom
thedirectpathoftheinletjetisreceivingenergyfrombothandT.Aftertheinletvalve
hasclosed,theturbulenceinthecylinderisrelativelyhomogeneous,duetotheeffective

11

transport properties of the intense turbulence during the inlet stroke. In fact, during the
last third of the inlet stroke, the jet through the inlet valve is much weaker, and the
productionandtransportaremuchlessimportant.Wehaveseenthatthedissipationterm
canbeverylarge.Thistermwillresultinarapiddecreaseoftheturbulentkineticenergy,
intheabsenceoftermsfeedingtheenergy.Wecanmakearoughestimateofhowfastthis
canhappen,ifwewrite /2 3 /2sothat

Wearesupposingthattheturbulenceinalldirectionsisequallyintense(apropertycalled
isotropy). This is probably not true, but is close enough for a crude estimate we are
interested only in orders of magnitude here. We may easily solve Equation 5.22 if we
assumethattheturbulentlengthscalelremainsconstant,togive

whereu0istheinitialvalueofu.Ifwetakeu0=10 ,andl=b/6=S/6,thenwehave

If we ask, How long does it take to reduce u/u0 to onehalf its initial value?, we easily
find t/S=0.05.If t/S=1.0,theintensityu/u0isreducedtoapproximately0.05ofits
initial value. This is something like the correct value. Actually, the turbulent intensity is
observedtodroptoabout0.15bytheendoftheintakestroke,andthenonlydropsto0.1
duringtheentirecompressionstroke.Wehaveleftoutsomeimportantmechanisms:the
attenuation due to the expansion during the last third of the intake stroke, and the
amplificationforthesamereasonduringthecompressionstroke.ESPdoesthiscalculation
correctly;allwearetryingtodohereisgetafeelingfortheorderofmagnitudeforsome
oftheseeffects.

INDUCING SWIRL AND TUMBLE
Bythewayinwhichthevalvesandportsarearranged,andthescheduleofvalveopening,
meanflowscanbeinducedinthecylinder.Theflowintothecylinderisturbulentandthe
mean velocity is often smaller than the turbulent velocities. Motions like this are often
calledcoherent,meaningthattheyareorganizationburiedinthedisorganizedturbulence.
When measurements are made, velocitypatterns must bemeasured during a number of
cycles (perhaps 2030) and the results averaged to find this organized part. I reproduce
here two figures from [105], Figures 5.3 and 5.4, which illustrate schematically tumble
and swirl, two types of coherent motion that can be induced in the cylinder. It is also
possibletoinduceacombinationofthetwomotions.Infact,itisessentiallyimpossibleto
generateswirlwithoutinducingsometumble,sothatthetwoarealwaysassociated.

12


Itispossibletogeneratetumblewithoutswirl.However,tumbleisalwaysassociatedwith
othersecondarymotions,sinceitisgeneratedbyflowthroughtwovalves.Wepicturein
Figure 5.5 schematically this secondary motion. This secondary motion has the effect of
isolatingthefluidenteringthroughthey>0valve(inFigure5.5)fromthefluidentering
through the y < 0 valve, something we will return to later. As we have indicated in the
figure, the size of this secondary motion is approximately b/2, and the intensity is also
about onehalf of the intensity of the main tumbling motion. Swirl and tumble, or a
combinationofthetwo,representthemostgeneralmotionthatcanbeinducedatthescale
of the cylinder. It is clear from the existence of the halfscale motion associated with
tumble,thatmanymorecomplicatedmotionsarepossibleatsmallerscale.

13

 Thereareanumberofreasonsforinducingswirland/ortumble.Highturbulence
levelsatignitionproducehighereffectiveflamespeeds,andmorereliablecombustionat
veryleanair/fuelratios,orwithEGR.Atnormalair/fuelratios,thehigherspeedwillallow
theflamefronttoreachtheendgasbeforethechemicalreactionresultinginautoignition
hastimetotakeplace,permittinghighercompressionratioswithoutknock.Thiswasthe
motivationintheearlydaysofenginedevelopment,whenfueloctanenumberswerelow.
Nowthemotivationismorelikelytobereliablecombustionwithveryleanfuel/airratios,
orwithEGR.Inanyevent,onepossiblereasonforswirlortumbleistopromotehigh
turbulencelevelsatignition.
Tumblecanalsobeusedforstratification.Wewilldiscussthislater.
Aswehaveseen,theturbulenceresultingfromtheconversionoftheenergyintheinletjet
decaysratherfast,andnotmuchisleftatignition.Theideaofswirlortumbleisto
encapsulatesomeofthemomentumoftheinletvalvejetintheorganizedmotion(swirlor
tumble),whichislessdissipativethantheturbulence(becauseoflargerspacialscale),and
hencewillretainitsenergylonger.Inaddition,thevorticityinthetumblecanbe
augmentedbythecompressionprocess,orbysquishinthecaseofswirl.Then,justbefore
ignitiontheswirlortumblecanbeinducedtobreakupintoturbulence,producingamuch
higherlevelofturbulenceatignitionthanwouldbepresentjustfromthedecayed
turbulencefromtheinletjet.

























Tumble appears always to break down to turbulence, because as the piston ap
proaches TC, there is not room between the piston crown and the cylinder head for a

14

vortex with a diameter of the order of b; only motions with scales of the order of the
clearanceheightcansurvive,sothevortexbreaksupintoturbulenceofthisscale.Inthe
caseofswirl,ifthecombustionchamberispancakeshaped,theswirlcansurvivethrough
theburn,andwewilltalkaboutthatinamoment.However,ifthecombustionchamberis
apentahead,withsquishprovidingthetransitionfromcylindertohead,and/orwiththe
pistoncrownprotrudingintothehead,theswirlmustaccommodateitselftothechanging
shapeofthespaceavailabletoit(fromcirculartorectangular,andincreasinglynarrow),
andwillalsobreakupintoturbulence.
In fact, in practical openchamber fourvalve pentroof engines, squish is a minor
factor.Thereissimplynoroomwithfourvalvesandasparkplugtohaveanysignificant
squisharea.Thesameistrueinmoderntwovalveengines.Hence,swirlsurvivestoTC,
anddoesnotbreakupintoturbulence.Itisnowclearfromcomputationalfluiddynamics
and flow visualization that even in engines that were designed for high swirl, is
principallytheassociated,unavoidable,tumblethatyieldstheturbulencejustbeforeTC.
There were apparent correlations betweenswirl and burn rate, but theseappear to have
heldonlybecausetherewasalwaystumblewiththeswirl.
Theonlyexceptiontothisisthehighcompressionbowlinpistonengine,whereswirl
plusthesquishdoresultinspinup,andthegenerationofturbulence.
ThebreakupofthetumblejustbeforeTCisnotwellunderstood,andisprobablyquite
interestingfromafluidmechanicalpointofview[29],[38],[101].Experimentsonrotating
flowinanellipsoidshowthatrotationaroundtheintermediateaxisisnotstable,andwill
changeovertorotationaroundtheminoraxisiftheellipsoidissufficientlyflattenedthat
is,iftheratiooftheaxesislargeenough.Whentheflowchangesover,astronglychaotic,
turbulent,flowisgenerated.Inanengine,thissuggeststhattumble,asthepistonrisesand
thespaceavailablebecomesmoreflattened,willchangeovertoswirl,withthegeneration
ofstrongturbulence.Thelosseswillbelarge,probablyenoughsothattheswirlgenerated
willbeverysmall.
Theotherpossiblereasonforinducingswirlistostratifythecharge;thatis,tokeepthe
fuelrich charge segregated, so that it does not mix with the remainder of the air in the
cylinder.Inordertodothis,itisnecessarytodampsomeoftheturbulence,selectively.It
turns out (as we shall see) that one of the effects of swirl is to damp strongly the
turbulenceneartheedgeoftheswirl,andifthecombustionchamberispancakeshaped,
so that the swirl can persist until ignition, the swirl vortex at ignition consists of a high
turbulencecoresurroundedbyaverylowturbulenceannulus.Ifthefuelaircloudcanbe
inducedtoenterthislowturbulenceannulus(anditcan,aswewillsee)thenitwillstay
there, and will not be spread throughout the cylinder. Ignition will not be a problem,
because the fuel/air mixture is locally near stoichiometric in the cloud. At ignition, the
flame front will at first be in the lowturbulence annulus, and will spread slowly, but it
will soon leave for the leaner mixture in the highturbulence core, where it will bum
reliablyandspreadrapidly.
Another type of stratification associated with swirl is important in direct injection
engines. The hot residual gases are considerably lighter than the cold incoming charge
(perhaps0.25ofthedensity),andwillmovetowardtheaxisofrotation.Fuelvaporisof
considerably higher density than air, and with direct injection engines it is possible to
obtain a rich mixture with a density ratio of perhaps 1.52.0, which will gravitate to the
outside of the swirl. Whether these will stay put depends on whether the density
gradients are large enough to damp the turbulence, which would otherwise mix them.
Thesedensitygradientsaredefinitelylargeenough.
Eveniftheswirlisbrokenupbysquishandapentahead,ifthiscanbedelayeduntil
close enough to ignition, a similar scenario will hold. At ignition, the turbulence will be

15

high,soflamepropagationwillberapid.Whatisimportantistonotprovideenoughtime
fortheturbulencetospreadthefuelcloudbeforeignition.Ifbreakupcanbedelayeduntil
roughly 15 CAD before ignition, then the resulting turbulence cannot spread the fuel
cloudmorethan1/3oftheavailableverticaldistance(perhaps0.25S)beforeignition.
Swirl andtumbleare usually present together. The combination of tumble and swirl
resultsinatiltedaxisofrotationofthesecondarymotioninthecylinder,andthistilted
axis precesses. So long as the tilt of the axis is not too great (say, 1:3), the suppression
effectontheturbulenceneartheedgeoftheswirlisstillpresent[80],
Bothswirlandtumblearenormallyspecifiedbyaswirlratio,ortumbleratio.Ineither
case,theangularvelocityofthesolidbodyrotationwiththesameangularmomentumas
theactualvelocitydistributionintheswirlortumbleiscomparedtotheangularvelocity
ofthecrankshaft.

WhereRsandRtarerespectivelytheswirlratioandthetumbleratio,andsandtare
respectivelytheangularvelocitiesofthesolidbodyrotationsthathavethesameangular
momentumastherealswirlortumbleflow.hasunitsofrevolutions/sec.
Productionenginesusuallyhavevaluesoftheswirlortumbleratiosbetween1.0and2.0.
Experimentalenginesachievenumbersuptoperhaps6.0.Ifwecallthetangentialvelocity
attheedgeofthesolidbodyrotation,sothats=2/b,thenwecanwrite

Hence

For a square engine, roughly 3/2 of the swirl ratio Rs gives the ratio of the tangential
velocity to the mean piston speed. Thus, in production engines we expect tangential
velocities below three times the mean piston speed, while experimental engines may
achieveninetimesthemeanpistonspeed.Wecansaythesamethingsfortumble.
The swirl and tumble ratios are normally measured in steady flow on a test rig
consisting of the cylinder head and valves, with a tube replacing the cylinder. A paddle
wheel,orsimilardeviceisplacedattheexitfromthetube,andtherotationofthedeviceis
used as a measure of the swirl or tumble. This approach is sound for swirl, but is less
satisfactoryfortumble.Itisusedbecausemeasurementoftheactualswirlortumbleina
motoringengine,oranoperatingengine,isextremelydifficultandexpensive.Thiscanbe
done, using various optical techniques (particle image velocimetry [80], laser doppler
velocimetry[100])ifanexperimentalenginehasbeenspeciallypreparedforopticalaccess
to the cylinder. The match between the steady flow swirl ratios and the swirl ratios
measuredinamotoringengineisgoodiftheactualvelocitydistributionisintegratedto
obtaintheangularmomentum.Forexample,[80]hadasteadyflowswirlratioof6.0,and
thePIVresultsbuildfromnearzeroatTContheintakestroketoabout6.5atBC,thenfall
slowly to about 5.0 at TC on the compression stroke. On the other hand, estimating the
swirlratiofromthemaximumtangentialvelocitygiveslowvalues;thisgives2.93instead
of 5.0 for the conditions of [80]. In [100], at TC the swirl ratio (estimated from the
maximum tangential velocity, measured by LDV) was 1.27, while the steady flow value
was about 2.6. At 150 CAD before TC, the estimated value was a little closer, at 1.75. It

16

appearsthatestimatingtheswirlratiointhisway(fromthemaximumtangentialvelocity)
gives about 0.6 of the true value. This number would vary for different velocity
distributionsintheswirl;however,thevelocitydistributionsareusuallyquitesimilar,at
leastforopenchamberfourvalvepentroofengines.Presumablythesameremarkscanbe
madeaboutthetumbleratios.Fortwovalveengines,and/orbowlinpistondesigns,itis
moredifficulttogeneralize.




















Swirl and tumble ratios can be obtained quite accurately for both production and
research engines using computational fluid dynamics. This is not a cheap solution,
however, because it requires the construction of a grid for the inlet manifolds, ports,
valvesandcombustionchamber,unlessthishasbeenconstructedforsomeotherpurpose.
Althoughtheswirlvelocityfieldisoftenmorecomplexduringtheintakestrokeandearly
onthecompressionstroke,byTCofthecompressionstroke,the meanvelocityfieldhas
settled down to a single vortex, centered in the cylinder. I include here (Figure 5.6) PIV
measurementsofthevelocityfieldatTCfrom[80].Bearinmindthatthisisaparticularly
simpleengine.

LiftStrategies

As can be seen from Figure 5.3, the fourvalve head lends itself to producing tumble,
whichcanbeoptimizedbymodifyingtheangleoftheinletrunners.Thisisprobablyone
explanation for the popularity of the four valve head, even though other valve
arrangementsmighthaveheattransferorMachindexadvantages.
Ontheotherhand,ifbothinletvalvesareopened,itdoesnotproduceswirl(presuming
thattheheadisreflectionallysymmetricabouttheplanebetweenthetwovalves).Inorder
to produce swirl, it is necessary to keep one of the inlet valves closed, or nearly so. In
[105],theyexperimentedwithvariablevalveliftina79mm76.2mmengineat =5
m/s. With equal lift they attainedRs = 0; with one valveclosed, they obtained Rs = 0.85.
TheRt=0.52withthevalvesequallyopen,butrosetoaboutRt=1.0withoneclosed.In
the Honda VTEC engine, for example, this strategy is used to obtain high swirl and
tumble.Usually,thevalvethatisnotopeningisactuallyopenedjustacracktoallowthe
fuelsprayedbytheinjectorintotheporttoenterthecylinder.Becausetheairflowisvery

17

small,thisissometimesusedtokeepafuelrichcloudnearthecylinderhead,toproducea
stratifiedcharge.

PortandValveConfigurations
Swirl can also be generated by the configuration of the inlet port, or by shrouding the
valve,ormaskingit.

I reproduce here Fig. 5.7 from [47], showing various inlet port configurations, which
producesimilarvaluesofswirlratios,between2.5and2.9.Thedeflectorwallportusesthe
port inner side wall to force the flow preferentially through the outer periphery of the
valve opening, in a tangential direction. The directed port brings the flow toward the
valve opening in the desired tangential direction (see [47]). These various ports pay
various penalties in lower discharge coefficients. The helical ramp ports appear to have
higherdischargecoefficients,becausethewholeperipheryofthevalveopenareacanbe
used [47]. The directed port, with its straight passage, has the most restricted flow area,
andthelowestdischargecoefficient.Sinceonlyonewallisusedinthedeflectorwallport,
theareaislessrestricted,andthedischargecoefficientishigher.
Port design couldprobably be approached rationally. For example, itseemspossible
that there is a design producing a minimum discharge coefficient for given swirl or
tumble, a design that could be found by variation of design parameters using
computationalfluidmechanics.However,inrealenginedesignthefinal(suboptimal)port
choice is made based on nonfluidmechanical considerations of packaging, fuel injector
targeting,andthelike,sothatconcernwithoptimalportsisprobablynotproductive.
Anotherwayofinducingswirlisbyshroudingormaskingthevalve.Figure5.8from
[47]illustratesshroudingandmasking.
Shroudingisoftenusedonexperimentalengines,butneveronproductionengines,since
itrequiresthatthevalvebepreventedfromturning.Normally,avalveisturnedslightly
by the lifter each time it is raised, and this has advantages in uniformizing valve head
temperatureandseatwear.Somespecialprovisionhastobemadetopreventthisrotation
if shrouding is used. Masking avoids this problem, and can easily be included in a
productionengine.However,bothshroudingandmaskinghavethedisadvantagethatthe
effective valve open area is reduced, which reduces the volumetric efficiency and
increasestheMachindex.

18

EFFECTOFCOMPRESSION

To understand what happens to either swirl or tumble during compression, we have to

talk about fluid mechanics for a moment. Conservation of angular momentum is the
principlethatwewillapply.Weareallfamiliarwiththeillustrationoftheiceskaterwho
spins on one toe with her arms outstretched, and then pulls in the arms, reducing the
radiusofgyration(asitiscalled),andincreasingtheangularvelocitysubstantially.Thisis
conservation of angular momentum her angular momentum with arms out and with
arms in is the same, but the moment of inertia is larger with the arms out, and smaller
withthearmsin,sotheangularvelocityissmallerwiththearmsout,andlargerwiththe
armsin.
Intheenginecylinder,theswirlvortexisbeingsquashedlengthwise,whilethetumble
vortexisnothavingitslengthchanged(wewillconsidertheeffectofsquishseparatelyin
a moment for now, assume there is no squish). However, the situation in the engine
cylinderiscomplicatedbythefactthatthedensityischanging,aswellasthelengthofthe
vortex. We need a more general statement of conservation of angular momentum with
density change. Deriving this here would take us far from our subject; the interested
reader can consult a good fluid mechanics text, such as [12]. What we need is called
Cauchysequation,whichstates(inasimplifiedform)that,inaninviscidfluid,ifavortex
ofinitialvorticity0 ,initialdensity0 andinitiallengthl 0 isstretchedtoalengthif,and
changedtoadensityf ,thenthefinalvalueofthevorticityf isgivenby

The same equation can also be obtained by starting with the integral form of angular
momentum conservation, and assuming solidbody rotation. Equation 5.28 can now be
appliedintheenginecylinder.

EffectonSwirlandTumble

First,considerswirl.Ifwestartatthebeginningofthecompressionstroke,isthestroke
plustheclearanceheight,S+h;atTC,l f issimplyh.Theratio

19


whereristhecompressionratio.Thedensityratio,ontheotherhandis

sothattheratioofthefinaltotheinitialvorticityis

Hence,thereisnochangeinthevalueoftheswirlvorticityduringcompression.
Tumble,ontheotherhand,isanotherstory.Here,theaxisofthevortexistransverse
to the cylinder, so that there is no change in the length of the vortex during the
compression(again,assumenosquish).Thus,

sothattheeffectofthedensitychangeistospinupthevortexquitestrongly.Thetumble
vortex will break up before the piston reaches TC. Also, in both cases, the vortices are
losing energy to the cylinder walls during the stroke, and we have not taken this effect
into consideration. Nevertheless, it is clear that the tumble vortex should contribute a
considerablyhigherturbulencelevelatignition,ifthisiswhatisdesired.
Wecanmakeacrudeestimateoftheeffectoffrictiononaswirlvortex.Thewallshear
stressisapproximately

whereuTisthefrictionvelocity(definedbythefirstpartofEquation5.33),andwehave
madethecrudestpossibleapproximation:uT/v=1/30,whichisapproximatelyvalidover
averywiderangeofReynoldsnumbers(see[96]).Weare,afterall,onlytryingtofindout
if this effect is important or not. Using the fact that the rate of change of the angular
momentum is equal to the applied torque, and supposing that the piston is stationary
(another crude assumption which will greatly simplify the problem), we can obtain the
equation

Thismaybeimmediatelyintegratedtogive

where istheinitialvalueofv.IfwetakeS=b,andassumethat 3
(correspondingtoRs=2),areasonablevalue,weobtain

asthetimerequiredfortheinitialtangentialvelocitytodropbyonehalf.Thatis,itwill
takesome38strokesforthistohappen.Despitealltheapproximations,itisclearthatthis
isnotanimportanteffect.Thereweremoreapproximationsthatwedidnotmentionwe
neglected the torque on the ends of the mass of gas. However, it is obvious that adding
thesewillnotmakeenoughdifferencetomakethisimportant.
Tumbleismoredifficulttotreat,becausethelossesarenotsosimpletoparameterize.
We have a cylindrical vortex with its axis transverse to the cylinder, and as the piston

20

comesup,andthevortexiscompressedbetweenthepistoncrownandthecylinderhead,
thegeometryisdifficulttomodelinanysimpleway.Theonlythingthatisclearisthatthe
lossesareconsiderablygreaterthantheyareforswirl.Thisisclear,forexample,in[80],
wheretheinitialswirlratiois6,andthetumbleratiois2.Attheendofthecompression
stroke, the swirl ratio should still be 6, but the tumble ratio should have gone up to 16.
However, from the observed tilt of the vortex axis, the tumble ratio is evidently still of
order2.Thesimplestwaytoviewthis,isthatthelossesaresogreatthatalltheadditional
energy(putinbythecompression)isconvertedintoturbulence.
Ifthereissquishwecantakethatintoaccountusingthesameequation.Considerthe
effect on swirl. We can do the calculation in two steps. First, consider a simple
compression without squish. The combustion chamber is cylindrical and the same
diameterasthecylinder.Thepistonwillstopwitharelativelylargeclearancevolume(a
low compressionratio) toleave roomfor the squish. According to ourequation (above),
Equation 5.31, there will be no change in the vorticity. Now bring in the sides of the
combustionchambertoproducethesquish.Duringthisphasetherewillbenochangein
thelengthofthevortex.If,forexample,theradiusisreducedtoonehalfitsvalue,thearea
is reduced to onequarter of its initial value, so the density increases by a factor of four,
andhencethevorticityisupbyafactoroffour.IhavesketchedthisinFigure5.9.Wecan
applythecompressionandthesquishsequentially,becauseEquation5.28relatesonlythe
initialandfinalstates.Recallthatsquishisnegligibleexceptinbowlinpistonorbowlin
headengines.
When tumble and swirl are present simultaneously, Equations 5.30 and 5.31 are
simultaneouslyvalid.

EffectonTurbulence

A swirling flow has interesting effects on turbulence. The radial distribution of angular
momentum can act dynamically like a stably or unstably stratified temperature
distributioninagravitationalfield(see[96]).
We will have to stop for a moment and talk a little about turbulence dynamics. See
Equation5.21andthediscussionimmediatelyfollowingit.Weneedtoconcernourselves
onlywiththeproduction.Inameanflowoncircularstreamlines,withtangentialvelocity
U,theproductiontermis

where is called the Reynolds stress. It is the mean value of the product of the
fluctuatingturbulentvelocitiesinthestreamwiseandradialdirections.Thisproductisnot

21

zero,becausethevelocitiesarecorrelatedthatis,whenonegoesup,theothertendstogo
up(orperhapsdown)also.Wewillgetbacktothatinamoment.
Theterm(y)correspondstothevelocitygradientinaparallelflow.Ifthemeanflowis
asolidbodyrotation,sothatthespeedUisproportionaltotheradius,thenthisvanishes.
Thereisnoshearinasolidbodyrotation.
In a turbulent flow,lumpsoffluidare continuallychanging position. Although they
areinteractingwiththeirsurroundingsandinterchangingmomentumwiththefluidwith
which they are in contact, theyare trying to conserve angular momentumalso,just as a
lumpoffluidinaparallelflowistryingtoconservelinearmomentum,whileinteracting
with its surroundings. We can make a very crude estimate of if we suppose that a
lump having the local tangential velocity of the swirling flow is displaced radially
outward(apositivefluctuationv)toanew,largerradius,andapproximatelyconservesits
angular momentum during this displacement. Conserving angular momentum would
meanthatits velocitywouldbe r1,soitsswirlspeedwilldropasitmovestoitsnew
radius.AvelocityfieldinwhichU r1 iscalledafreevortex;thishasthesameangular
momentumatallvaluesofr.Whenthedisplacedlumparrivesatitsnew(larger)radius,if
thelumpfindsitsswirlspeedtobeslowerthanthemeanswirlspeedofthesurroundings
atthenewradius,itproducesanegativevelocityfluctuationu;thisgivesavalueofthe
product <0.Ontheotherhand,ifthedisplacedlumpfindsitselfmovingfasterthan
thelocalmeanswirlspeed,itwillhaveapositivefluctuationu>0.Hence,weexpect <
0ifthemeanswirlspeedhasaradialslopelessnegativethanthatofafreevortex,and
>0ifthemeanswirlspeedhasaradialslopemorenegativethanthatofafreevortex.
Considerradial profiles of mean swirlspeed that lie betweena free vortex andsolid
bodyrotation(thatis,U r).Thiswillcoveranyprofilethatwewillfindinaswirlvortex
inanenginecylinder.Thenweexpectthat <0,andthemeanshear <0.Hence,
weexpecttheproductiontobenegativethatis,itwilltakeenergyoutoftheturbulence
and put it back into the mean flow. It will act to suppress the turbulence. It will be a
stabilizinginfluence,likeastabletemperaturedistribution.

InFigure5.10IhaveplottedthedistributionoftangentialspeedinthevortexofFigure5.6.

22

 Itisclearthattheswirlvortexisalmostinsolidbodyrotationouttoaradiusof
approximately 28 mm, and that the tangential velocity has an essentially constant value
afterthat.Thesolidbodyrotationisduetoahighturbulencelevel.Intenseturbulenceis
likeaveryviscousfluiditwillactivelytransportmomentuminsuchawayastotryto
bringthemeanprofiletosolidbodyrotation(astateofzeroshear).Weexpectthatthere
will be very little production out to a radius of 28 mm, and strong negative production
beyondthatpoint.

We can make a crude estimate of the strength of this negative production. We
havedeterminedthat <0;uandvshouldbewellcorrelated,ifthefluctuationsinuare
essentially caused by the fluctuations in combined with the conservation of angular
momentum. Hence, we should expect u2, where u is the r.m.s. value of the
turbulent velocity fluctuation. Let us take U r out to a certain radius, say near 28 mm
andconstant=U0thereafter.Thentheproductionwillbeequalto(intheregionwhereitis
negativeandnonzero,forr>28mm),

Weshouldcomparethiswiththedissipation,whichisgivenbyD= =u3/l ;letustakel=
b/6andevaluatetheratioP/Datr=b/2:

LookingatFigure5.10,weseethatU0=17m/s.Figure5.11,whichwewilldescribeina
moment,givesthepeakvalueofthemeasuredturbulentfluctuatingvelocityinthisflow
at u = 3.46 m/s. As a result, we have U0/3u 17/10.4 1.6. This estimate is extremely
crude,butitsuggeststhatthesuppressiveeffectofthisnegativeproductiontermmaybe
important.Inthisregionwehave,accordingtoourestimate,destructiveforces(negative
production plus dissipation) =2.6 . This means (approximately) that the fluctuating
velocityintheperipheralregionwilldecaytoavalue1/2.6ofthefluctuatingvelocityin
the core in the same time, or that the fluctuating kinetic energy in the peripheral region
will be about 15% of the core kinetic energy after the same decay period. This is
numericallyapproximatelywhatFigure5.11shows.
Let us look at Figure 5.11, reproduced from [80], which presents the measured and
calculatedturbulencedistributionscorrespondingtoFigure5.6.Lookfirstatthemeasured

23

distribution.Inthecenterofthecylinder,theturbulencelevelisroughly3.4m/s,orabout
the mean piston speed. The turbulence level is thus about twice what can be achieved
withoutswirlandtumble.Itisalsoevidentthat,fromaradiusofabout26mmouttothe
cylinder wall (at least to the resolution of the measurements a word about that in a
moment)thereisverylittleturbulence,intheneighborhoodof1m2/s2.Thisisexactlythe
regioninwhichwehaveestimatedthatthereisanegativeproductionroughly1.6times
the dissipation. Note: if there is no turbulence, there is no negative production. The
negativeproductionwehavecalculatedwaspresentwhentherewasturbulence,anditis
whatkilledtheturbulence,butnowthattheturbulenceisdead,thenegativeproduction
has died also. In Figure 5.11 we are looking at the flow just before TC, but the value of
negative production we have calculated corresponds to the conditions during the
compressionstroke.
Thecalculateddistributiondoesnotreproducethis.Theturbulencefallsoffinasortof
Gaussianshapeclosetothecylinderwall,butdoesnotfalltozero.
This is because the turbulence model used cannot reproduce the behavior of the
turbulence in a swirling flow. Developing a turbulence model that is capable of
reproducingsuchbehaviorisonthefrontiersofresearchatthemoment(see,forexample,
[55, 83, 84]; these references refer to rotating flows, but rotating and swirling flows are
closelyrelated).
Veryclosetothewallthereisaturbulentboundarylayer.Thislayerisnotresolvedby
thePIVtechnique.Thecylinderhasaboreof92mm,whichgivesaradiusof46mm.The
PIVdataarelimitedtoacircleofradiusroughly35mm,ascanbeseeninFigure5.11,so
thereis a layer next tothe wall of thicknessroughly 11 mm that is not observed. In this
region,thereisarelativelythinturbulentboundarylayertryingtopropagateagainstthe
stabilizing negative production at its edge. This boundary layer is unstable, and is
dominated by shear, unlike the flow just above it, which is dominated by rotation.
Boundarylayerslikethisintheatmosphere,propagatingupwardagainstbuoyantstably
stratified fluid, are wellunderstood, but our boundary layer in this rotating flow is not,
andwecannotpredictthethicknessofthisboundarylayer,otherthantosaythatitwillbe
thinner than a similar boundary layer in a nonrotating flow. The turbulence in the
boundary layer will be relatively weak, approximately /30, which makes it about one
tenth of the intensity of the turbulence in the center of the cylinder. It will also be of
relatively small scale, roughly 1/3 the thickness of the boundary layer, which is a small
fraction of the bore. The effective diffusivity in the boundary layer, ul, will therefore be
roughly2%(ormuchless)ofthediffusivityinthecenterofthebore.
In this flow of [80] the initial tumble ratio was 2, while the swirl ratio was 6. As we
haveseen,theswirlratioaftercompressionshouldstillbe6,butthetumbleratioshould
be 16 (r = 8). However, as we noted above, the tumble ratio stays roughly constant,
indicatingthatalltheadditionalworkdoneonthetumblevortexbythecompressionhas
beentransformedtoturbulence.However,itisevidentfromFigure5.11thatthenegative
productionintheouter1/3oftheradiushasmanagedtokilltheturbulenceputinbythe
breakinguptumbleaswell.
In several recent works [9], [44], [57], detailed budgets of angular momentum, mean
and turbulent kinetic energy have been calculated in the engine cylinder. These suffer
frombeingmodeledturbulence,andfairlysimplemodels.However,thiswouldnotbea
problem our analytical models are also quite simplistic. Unfortunately, the budgets
presentedareglobalbudgets,integratedoverthecylindervolume,sothatnoinformation
isavailableonlocaldifferencesinproduction.
Itisclearthatthereisalotgoingonherethatwouldrepaycloserattention.

24