ThepiperigconsistsofahydraulicallysmoothPVCpipe,with adiameterof 186
mm and a length of 83 diameters. Ten pressuretaps are mounted on the pipe,
making it easy to measure the pressuregradient. The pipe is tted such that a
traversecan be mounted ontop, making it possible to traverse the ow through
thecenterofthepipe. Velocitiesinthepiperigcouldbevariedfrom5to12.5m/s.
Thecoordinate systemusedin thepipehasits reference
( y = 0)
onthe centre lineofthepipe,yispositiveabovethecenterline,andnegativebelowthecentreline.The velocities in thepipe aredenoted
U x
,U r
andU theta
and arethe axial, radialandcircumferentialvelocitiesrespectively.
An openloop wind tunnel is used forthecylinder wakemeasurements. The test
sectionis45cmx45cmand110cmlong. Acylinderwithadiameterof47.5mm
is tted in the center of the test section, leaving 50 cm of distance downstream
for the ow to develop. Measurementsare takenat
x/D = 10
. Velocities in the windtunnelcouldbevariedfrom4to30m/s.Inthewindtunnelthecentreofthewakeisthereference
( y = 0)
inthe coordi-nate system,yispositiveabovethecenterline,andnegativebelowthecenterline.U,VandWaretheaxial,vertical,andtransversevelocitycomponentsrespectively.
4.1 Calibration and testing
4.1.1 Velocity calibration
Inthepiperigthevelocitycalibrationwasperformedforvelocitiesbetween5and
12.5m/s. Athird order polynomialt to thecalibrationdata includingthezero
velocitypointgavearesidualoftheorderof
10 − 1
,whileasecond orderttothe datawithoutthezerovelocitypointgavearesidualoftheorderof10 − 3
. Asecond order polynomialwasthereforefoundmostsuitableforvelocitycalibrationinthepiperig. Theresidualisdened asthesumoftherelativedeviations betweenthe
polynomialt andthettingdata.
Inthewind tunnelthe velocity rangedfrom 4to 30 m/s. Figure 8showsthe
distributionofthemeasuredvelocityin apointinthecylinderwake.
0 2 4 6 8 10 12 14 16
0 1000 2000 3000 4000 5000 6000
U(E 1 ) hw [m/s]
# []
distribution lower calibration limit centre of distribution
Figure 8: Distribution of streamwise velocity,U, in a measurement point in the
wakebehindacylinder
The velocity scatter falls under the lowest freestream velocity obtainable in
the windtunnel. This isnotideal asthe polynomialt in that regionmost likely
will causeanerrorin theestimated velocitybut itcouldnotbeavoided. A third
order polynomialt tothe calibrationdata includingthe zerovelocity pointwas
chosenandgavearesidualof
10 − 1
. Thematchbetweenthepolynomialtandthe calibratondataforwire1isshowningure9.4.1.2 Eectiveangle calibration
Figure10showsthecosinettingoftheeectiveanglecalibrationdata.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0
2 4 6 8 10 12 14 16 18
e − e 0 [V]
U [m/s]
polynomial calibration data
Figure 9:
−20 −15 −10 −5 0 5 10 15 20
0.7 0.8 0.9 1 1.1 1.2
U(E 1 ) hw /S [−]
Angle [ o ] Calibration data Cosine fit to data
Figure10:
Wire
α e
1 31.69
2 35.88
3 35.64
Table1: Eectiveanglesfoundfrom calibration
Trueangle
U pitot
U V W Calc.angle|V |
Rel.error-20.00 9.9692 9.3490 0.4565 -3.5968 -21.05 10.0275 0.0058
-15.00 9.8884 9.6096 0.4545 -2.7205 -15.81 9.9976 0.0110
-10.00 9.9434 9.7241 0.4218 -1.7365 -10.13 9.8869 0.0057
-5.00 9.9169 9.8504 0.2921 -0.8429 -4.89 9.8907 0.0026
0.00 9.9007 9.8726 0.1849 0.1024 0.59 9.8749 0.0026
5.00 9.9608 9.8039 0.1368 0.9367 5.46 9.8495 0.0112
10.00 9.8936 9.6796 0.0971 1.7844 10.45 9.8431 0.0051
15.00 9.9672 9.5099 0.1003 2.6105 15.36 9.8622 0.0105
20.00 9.9852 9.2656 0.1193 3.3928 20.12 9.8680 0.0117
Table2: Test ofsolutionondatasetforwire1
The residuals of the curvet for the wires was of order
10 − 2
. The eective anglesgivenbythecalibrationis givenin table1.Theangles are in thevicinity ofthe idealvalue of
35 . 26 ◦
and vary within an aceptablerange. Totesttheeectiveanglesandtheφ
anglesthedatasetfromtheanglecalibrationofwire1canbesolved. Table2showsthetrueowangle,theU
velocitymeasuredbythepitot,thecalculatedvelocitycomponents,thecalculated
α
, thelength ofthe calculatedvelocityvector andthe relativeerrorbetweenthevelocitymeasuredbythepitotandthelengthofthecalculatedvelocityvector.
Table2showsthatthecalculated
α
fallswithin± 1 ◦
ofthetrueowangle. The relativeerrorbetweenU pitot
and|V |
is1 . 2%
atmost. TheY p
componentvelocity,V,should bezerobutshowsavariationwithrespectto
α
. ThemaximumvalueofV correspondstoaowangle of
2 . 9 ◦
ora
4 . 7%
relativeerror,whichisarelativly large error. The variation in V corresponds to an angle of2 . 2 ◦
. Aanesland [1]
performedthe samemeasurements, and reportsa
3 − 4%
relativeerrorin V and W forsimilarconditions.The largedeviation seemsto be a combination of misalignment of the probe
relativetotheowandadependencyofthesolutiononthetrueowangle. Aprobe
pitchangledierentfromzerowouldgivethemisalignment,producingapermanent
oset. Thevariationwith
α
is mostlikelycausedbyawrongvaluechosenforφ 1
,since
φ 1
is determined by visual observation and is not likely to be exact. The variationinVcouldbeusedto ndthecorrectvalueofφ 1
,thecalculationscould thenbererunned,and theremainingconstanterrorin Vshould becausedbythepitch. Suchaprocedure hasnotbeenattempted in thisproject. Several authors,
i.e. Cantwell and Coles [3] uses yawingto perform similar corrections, Cantwell
andColesuseditforax-wireprobe.
Twoprolesof theturbulent pipeow hasbeentaken ata Reynoldsnumberof a
approximately
10 5
. Themain dierencebetweenthetwoprolesisthat theyare takenattwodierentvaluesofφ 1
. Thisisdonetoinvestigatetheeectofrotating theprobe.4.2.1 Pressuregradient
The static pressure in the pipe is measured using the pressure taps distributed
along the pipe. Instead of measuring the absolute static pressure, the pressure
dierenceismeasuredbetweenthedierentpointsandachosenreferencepointat
X/D = 70 . 5
. Figure 11showsthedropin staticpressurealongthepipe. Theline drawnintheplotisastraightlinefrom therstmeasurementto thelast.0 10 20 30 40 50 60
Pressure drop, Re = 9.7413e+4 Pressure drop, Re = 1.0329e+5
Figure11:
AsexpectedthehighestReynoldsnumbergivesthebiggestpressuredrop. Both
measurementseriesshowanearlylineardropinpressure,asonewouldexpectfor
afully developed ow. Thepressuregradientisapproximatedby consideringthe
pressuredrop from
X/D = 11 . 3
toX/D = 70 . 5
. Based onthepressuregradient thefrictionvelocitymaybefoundfromequation22. Thefrictionvelocitycanalsobeestimatedfromthefrictionfactor,f.
u ∗ = U avg f
8 1
2
(42)WhentheReynoldsnumberandtheroughnessheightofthepipewallisknown
thefrictionfactorcanbefoundfromtheMoodydiagramorfromanequation. The
PVC pipe is hydraulically smooth, and thefriction factor cantherefor be found
φ 1 Re D dP dx [ P a/m ] u ∗ dP
from Prandtls equation for smooth pipes [11]. Table 3 gives the results for the
dierentvaluesof
φ 1
The two means of calculating the friction velocity gives similar results, the
values given by the friction factor are about
6%
higher than that given by the pressuregradient. Inthefollowingthefrictionvelocityobtainedfromthepressuregradientisassumedtobecorrect.
4.2.2 Mean velocities
Figure12showsthevelocityprolefortheaxialvelocity
U x
,forbothφ 1 = 90
andφ 1 = 180
.Thevelocityonthecenterlinewasmeasuredasecondtimeaftertheprolewas
taken, the resultwaswithin
1%
of therst measurement forbothproles. Both proleshavethesameshape.Ingure13thedatafor
φ 1 = 90
andφ 1 = 180
isplottedagainstthelogarithmic law. Torbergsen [9] didmeasurementsin the samepipe rig forRe = 75000
and obtained agood t with the logarithmic law usingκ = 0 . 41 andB = 5 . 5
. White [12] claimsthatκ = 0 . 41 andB = 5 . 0
giveabettert toexperimental data. Bothdatamatchesthechoiceof
κ = 0 . 41
. B=5.0assuggestedbyWhitegivesthebest tto themeasurementdatain thelog-lawregion.10 2 10 3
Inafullydevelopedturbulentpipeowonewouldexpectthetransversevelocity
components
U r
andU θ
to be zero. Figures 14 and 15 show that neither of thevelocitycomponentsareexactlyzeroacrossthepipe. Theradialpipevelocity
U r
,isfairlyconstantoverthecrosssectionofthepipe,butshowsomevariation,especially
close tothepipe walls. Therangeof variationin velocityis about
± 0 . 125 m/s
orapproximatelyanangle of
± 0 . 8 ◦
relativeto the averageaxial velocity. Both the proles forφ 1 = 90
andφ 1 = 180 ◦
showthe samekindof variationwith respect
to
y/R
but theyare osetrelativeto one another. Theoset equalsabout1 . 6 ◦
ofprobepitch,whichiswithintheerrorrangeonemustexpectwhentheprobeis
alignedwiththeowvisually.
Thecircumferentialvelocity
U θ
showapeculiar variationoverthecrossection of the pipe, varying overa rangeof± 0 . 45 m/s
or± 2 . 86 ◦
relativeto theaverage
velocity. If thedeviation of the circumferentialvelocity is compared to the local
axialvelocitytherangeofanglevariationexceeds
± 4 ◦
,this isshowningure16.Thevariationof
U θ
withrespecttoy/R
followsthesamepatternforbothseriesof measurementsand closely resembles atypicalinverse tangent function. If the
variation were to be explainedphysically it would implythat the owinside the
pipe wasrotating about the centerline. The velocity doeshowevernot decrease
close to the wall, but increases rapidly, this implies an unphysically large shear
stressonthewall.
Sincetheowisassumedtobeunphysicaltheradialvariationmustbecaused
byoneormoreerrorsin thesetup,datareductionorcausedbylimitationsofthe
−0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15
Figure14: Radialmeanvelocity,
U r
−0.5 0 0.5
Figure15: Circumferentialmeanvelocity,
U θ
−5 0 5
Figure 16: Circumferentialmeanvelocity,
U θ
, expressedasangulardeviationrela-tiveto
U x
probe. Thedistinctshapeoftheprolecanmakeonewonderifthisvariationcould
be linked to an error in the arctanexpression in equation 18, but no such error
hasbeenfound. Aninterestingobservationcanhoweverbemadebyobservingthe
dierenceinthetwocenterlinemeasurements,bothfor
U r
andU θ
therelativeerrorbetweenthetwocenterlinepointsislarge. Whiletherepeatedmeasurementsgave
solutionsfortheaxialvelocity
U x
withinonepercentoftherstsolution,theradialand circumferential repeated solutions can vary up to
0 . 5 ◦
and
1 ◦
respectively.
Thisisalargedeviationcomparedtotherangeofthecalculatedvaluesfor
U r
andU θ
. By re-examiningthe measurementdata there was found to be asmall drift in wirevoltage, whichcould notbecorrected forby consideringthe temperaturechange. Ideally themeasurement seriesshould havebeenrepeated, but theerror
wasdiscoveredtolate. Basedonthisobservationpartsofthelargevariationfor
U r
and
U θ
mightbecausedbyvoltagedrift. TheshapeofthevariationofU r
andU θ
doeshoweverseemtobeafunctionofyorasomeotherpropertyrelatedtoy,not
onlyapossiblevoltagedrift. Butwhatpropertycouldthatbe?
U x
isafunctionofy,butissymmtricaboutthecentreline.Thegradientof
U x
alsovariesasafunctionof y but is notsymmtric. Insection 2.2 the possible error ofmeasuring in large
velocitygradientswas dicussed, this couldpossiblybethecause. This discussion
iscontinuedin thenextsectiononshearstresses.
4.2.3 Turbulentshear stresses
As discussed in section 2.3.3, the
u x u r
shear stress is expected to be thedomi-nantshearstressandbehavelinearlyacrossalargeportionofthepipecrossection
accordingtoequation37. Figure17showsthetheoreticalrelationandthe
experi-−1 −0.5 0 0.5 1
Theexperimental resultsfollowthelinear relationwell. The gradientis abit
smallerthan1inthecentreregionwhereviscousshearisassumedtobeneglectable
but thedeviationis small,averysimilar resultwasfoundbyTorbergesen[9]. At
thecentrelinethemeasuredshearstressisclosetozero,asexpected.
Closeto the pipe wallthe turbulent shear stressesare expected to decay and
droptozerointheviscoussublayer. Thespatialresolutionoftheprobeisnotlarge
enoughtomeasure closerto thewallthanabout
y + = 70
whichisfar outsidethe viscoussublayer. Themeasurementdatashowlittletendencytodropofclosetothewall. For
φ 1 = 90
there isalittledropfory/R < − 0 . 9
,whilefory/R > 0 . 9
there is actually anincreasein shear stress measuredfor both datasets. Theincreasedshearstresscouldjust beoutliers,butithappensforbothdatasets.
The two other shear stresses,
u x u θ
andu r u θ
should theoretically be zero as there is no meanvelocity gradientresultingin productionof neitherof them. Ingure18thenormalizedstressesareplotted. Inthecentreregionthemagnitudeof
u x u θ
andu r u θ
arerelativelysmallcomparedtothemaximumvalueofu x u r
,about3
%
, but not zero. Moving closer to the wall both shear stresses increase slowly until|y/R| ≥ 0 . 8
,where theshearstressesincreasemorerapidly. Themagnitude and variation of the shear stresses can not be explained physically, and must berelatedtothemeasurementprocess.
For
|y/R| ≥ 0 . 8
thevelocitygradientexperiencedbytheprobevolumeislarge.The exact eect of an exessively large velocity gradient compared to the probe
volume is unknown. But it will result in calculated velocities dierent from the
truevelocity,as the wallis approached. Theresult canbe agradientofboth
U r
and
U θ
withrespecttoyasobservedinsection4.2.2. Subsequentlythisislikelyto−0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3
resultin unphysicalshear stresses. Insection4.2.2 thedeviation forboth
U r
andU θ
wasfoundtoincreaserapidlyfor|y/R| ≥ 0 . 8
,thismatchestheresultfoundforu x u θ
andu r u θ
and supports the theory that a largevelocity gradientbiasestheresult.
WhencomparingtheresultswiththatfoundbyAanesland[1],thedatascatter
fortheprolesobtainedin thisprojectislargeranddonotcollapseasneatlyinto
a line astheresults ofAanesland. What causes this is uncertain,but a to short
samplingtime couldpotentiallybethereason.
4.2.4 Turbulentnormal stresses
Figure 19 displaysthereduced normalstress forbothmeasurementseries, which
revealsthat
( u x ) +
isthelargestnormalstress,asonewould expect.Torbergsen [9]found
u x +
onthe centrelineto beapproximately0.85 for< = 0 . 75 e 5
. TheresultsshouldbecomparableastheReynoldsnumbersisof thesame order. According to the resultsof Torbergsen,u x +
is fairlyconstant onthecen-trelineforincreasingReynoldsnumbers,butincreasesclosertothewallduetothe
increasedvelocitygradient.
u x + cl. = 0 . 85
matchestheobtainedresultsfairlywell, there is however somescatter in the data as is already mentioned in the end ofsection4.2.3. Thedataseriesforthetwodierentprobesalsogiveslightlydierent
results. Movingcloserto thewall
u x +
isunderestimatedcompared tothe results ofTorbergsen,butmatchestheresultsofAanesland[1]better.u r +
andu θ +
showthe same variation asreported by Torbergsen for
|y/R| ≥ 0 . 6
but the scatter is largeforvariationofφ 1
.Valuesfor
u x +
,u r +
andu θ +
cannot beestimated without givinga relativlylargepotentialerror.
u θ +
fory/D = ± 0 . 6
,canforinstancebeestimatedas1 . 1
but−1 −0.5 0 0.5 1
Figure19: Reducednormalstress
thescatter isof theorder0.2, andcould potentiallybelargerifmeasurementsat
morevaluesof
φ 1
weretaken. Thelargevaritioncanbecausedbypooradjustmentof
φ 1
,itisasmentionedearliersetvisually.Ingure20theturbulenceintensityrelativetothelocalstreamwisevelocityis
plotted.
Thestreamwiseturbulence intensity onthe centreline,
u x + ≈ 3 . 3 − 3 . 6
. Tor-bergsenreporteda valueof approximately3 . 5%
. On thecentrelineu r +
andu θ +
should beequaldueto symmetry,there ishoweversomedierencewhich ismost
likelycausedbymisalignementoftheprobe.
4.3 Cylinder wake
Thecylinderwakecanbeanalyzedbothasameanowandasatimevaryingow.
Both approacheswillbetestedin thissection. Ideallyseveralmeasurementseries
for dierentvaluesof
φ 1
should havebeentakento gainmoreinformationabout theproberesponse. Onlyonemeasurement serieswastakenhowever. A velocityprole wastakenin the test sectionbefore thecylinder wasinserted,to mapthe
referencefreestreamconditions.
Vortexsheddingfromacircularcylinderisamuchstudiedow,itdidhowever
provediculttondnearwakeresultsinthesamerangeofReynoldsnumbers
( 10 4 )
, and downstreamdistancex/D.A surveyusingafourwirehotwireprobebyOngand Wallaceat
< = 3900
wastheclosestmatch found[7]. Theyusedafour wire hot wire probe with a crossection of 1 mm x 1 mm. For the ow investigatedin this project
< = 30717
, placing it in the subcritical range together with the resultsofOngandWallace. Theresultscanthereforebeassumedindependentof−1 −0.5 0 0.5 1
Figure20: Turbulenceintensity
Reynoldsnumberandcomparable.
Emphasisis placedonassesmentof thephysical validityof theresult, aswell
ascomparisonwiththeresultsofOngandWallace.
4.3.1 Mean velocityproles
The prole of the streamwise component U is shown in gure 21, the velocities
are plottedrelativeto theincomingvelocity. Fromgure21onecansee thatthe
velocityinthewakeexceedsthereferencefreestreamvelocity. Thisindicatesthere
is a speedup eect caused by the cylinder. The average free stream velocity is
9 . 7 m/s
whiletheaveragevelocityin the wakeproleis9 . 3 m/s
. A loweraveragevelocityin thewakemeansthat thepresenceof thecylindercausesablockageof
theow. Theareaof thecrossection traverseddoeshowevernot covertheentire
crossection. Since thevelocityin thewakeproleishigher closeto thewall, this
means that the average velocity in the wake is higher than
9 . 3 m/s
. By simplyestimating the rest ofthe wake prolefrom the highestmeasuredvelocityin the
wake,the averagevelocityin the wake isfound to be
9 . 66 m/s
. This means thatblockage eects are neglectable, and that areference velocity of
9 . 7 m/s
may bereasonable. BasedonthereferencevelocitytheReynoldsnumber,
Re D
,iscalculatedto be30717.
Asthecylinder wakeisanevolvingow,V isnotexpected tobezero,due to
the constraintsof continuity. Fora perfectly symmetrical wake V is expected to
bezeroin thecenter oftheowandnegativeoverthesymmetrylineandpositive
below. TheproleofV isplotted in gure22, alongwith thefreestreamprole.
Ideallythefreestreamconditionsshouldbezero. Duetomisalignmentoftheprobe
anosetfrom zerowouldnotbeunexpected, thevelocitydoeshowevervaryover
0.75 0.8 0.85 0.9 0.95 1 1.05 1.1
−3
−2
−1 0 1 2 3
U/U r ef [−]
y/D cyl. [−]
U wake /U ref U freestream /U ref
Figure21: NormalizedproleofUin cylinderwakeforRe=30717,x/D=10
thecrossection. Whetherthis isatruepropertyof theowornotisunclear,the
rangeofvariationinVequals
2 . 4 deg
probepitchrelativetothereferencevelocity.Ifthevelocityproleiscorrectedbysubtractingthelocalfreestreamvelocity,the
prolewillbeasin gure23. Thiscorrectionassumesthattheowvelocities can
besuperpositioned. The resultlooksmorelikewhatonewouldexpect,but when
comparedtotheresultsofothersi.e. OngandWallace[7]theshapeofthevelocity
proleisnotaperfectmatch.
The velocity component in the z-direction is plotted in gure 24 along with
the freestream measurements. The samevelocity prole correctedfor freestream
conditionsisplottedin gure23.
Thefreestreamvelocityvariationcannotbeexplainedbythenitesizeofthe
probeandit is notaconstantoset asayawangle would give. The variationof
Winfreestreamconditionsishoweverrathersmall,itequalsapproximately
2 ◦
. In
thewakeofthecylinderthevariationofWis sligtlylarger.
Itisnotstraightforwardtounderstandhowthetimevaryingvelocitygradients
Itisnotstraightforwardtounderstandhowthetimevaryingvelocitygradients