1.2 High-Speed Simulations Assessment with Wind T unnel Data and
1.2.4 Inidene Eet
A high-speed run inthe windtunnel ampaign hasbeenhosento ompare
HOST-MINTsimulations withexperimental dataand CFDuRANSomputationsinelsA.
Table1.3presentsasummaryofthemainowonditions. Simulationsareperformed
at experimental thrust levels by hanging the blade pith in HOST-MINT and in
elsA simulations.
Propellerperformane. Anumberofparametersareomparedhereafter: thrust
oeient(
C
TH(a)Loadtermsdenition (b)Momenttermsdenition
Figure1.10: Denition ofloads and moments
Notie rst the non-zero 1P loads modulus at
0 ◦
inidene in the wind tunneldata. Asithasbeenexplained byOrtun[Ortun 2012℄thismayomefrom anerror
inthe balanesalibration. Hene,inthepresent study,1P loadsat zero inidene
have been onsidered as a onstant oset in the measurement of the balanes for
all thetest ases. Theplotted errorbars showthe upper and lower limit ofthe 1P
loads modulus andphase whenonsidering this oset.
ComparisonsinFig.1.11showagoodagreement between isolated HOST-MINT
simulations and wind tunnel test datain termsof thrust and power. However, we
an also notie an underestimation of the 1P load oeient (around
40%
) and aslight overestimation of the 1Pphase lag(from
5
to15 ◦
).When installation eets are added in HOST simulations, an underestimation
of thrust is obtained. Consequently, the pith angle should be modied by
+0.7 ◦
in order to obtain the experimental thrust levels. In those ases, as shown in the
upperleftplot inFig.5.1,thepoweris overestimated byabout
10%
.Several reasonsan explain these powerosets. First,only theoldCAD blade
shape was available for simulations, thus bladedeformationswerenegleted in
sim-ulations. Seond, asit has been explained in the Methods and Tools Chapter, the
tehnial report indiated that the rotating balanes usedin the wind tunnel tests
mayunderestimatetheshaftpower. Third,inordertoreahexperimentalthrust
lev-els,aninreaseof
0.7 ◦
inpithanglewasneessaryinthesimulations. Changingthe bladepithleadstoaninreaseinthebladeinideneandthustoaninreaseintheseparated boundary layer near blade tips, whih are aerodynami onditions more
diult to be preditedorretly with RANSsimulations or withtwo-dimensional
airfoil data. Finally, neither in elsA CFD simulations nor in HOST airfoil data,
laminar-turbulent transition eets aretaken into aount.
Incidence [deg]
(a)Thrustoeient,
C
THIncidence [deg]
(b)Poweroeient,
C
PWIncidence [deg]
1 P L o a d C o e ff ic ie n t, C 1 P
0 0.5 1 1.5 2 2.5 3
1 P M o m e n t C o e ff ic ie n t, C M 1 P
0 0.5 1 1.5 2 2.5 3
(e)1Pmomentoeient,
C
M1PIncidence [deg]
1 P M o m e n t P h a s e L a g , M 1 P
0 0.5 1 1.5 2 2.5 3
Figure 1.11: Inidene eet on propeller performane and 1P loads. Comparison
between wind tunnel data, CFD uRANS omputations and HOST-MINT
simula-Propellerplane
Spinner
α ∞
V ∞ α ind α V ⊥
α = α ∞ + α ind
Figure1.12: Spinner eet sheme(side and front view)
islosertowindtunnelmeasurementsandverylosetoCFDsimulations. Asshown
bytheshemeinFig.1.12,onsideringtheperturbationveloitiesaroundthespinner
gives an additional indued inidene to theairow, whih inreases the inidene
asymmetrybetween downward and upward movingblades, andsoinreases the1P
load modulus.
When regarding 1P phase lag, while CFD omputations predit values within
theerror bars for all the inidenes,it an be notied thatHOST simulationstend
to overestimate thesephase lagsby
5 ◦
to10 ◦
.A very good agreement in terms of 1P moment oeient an be observed
be-tweenisolated HOSTsimulationsandexperimentaldata. However,whenaddingthe
installation eets, HOST-MINT tend to overestimate the norm of these in-plane
moments. On the other side, signiant osets areobserved in1P moments phase
lag. Even though,the levelsof mismatharesimilarto thoseobtained inelsA
om-putations. These mismathes in 1P moment preditions may ome from the fat
that thelifting-lineapproah is lessaurate inthe predition ofloads lose to the
bladetips. Astheontribution ofagivenbladesetionon1Pmomentsorresponds
to theamplitude ofthrust osillations multipliedby theleverarm, a slight errorin
thethrustpreditionnearbladetipsmayindueimportant1Pmoment mismathes.
Simulations onsidering installation eets and the hub model give results in
terms of thrust and power that are similar to the results of isolated simulations.
When the hubmodel isativated,no blade pith modiationis needed.
Notiethe important ontribution to the1Ploadnorm, when omparedto isolated
or installed HOSTsimulations. An overestimation ishowever observed in
ompari-sonwithexperimental data(around
+17%
). Ontheother side,a slight inreasein1P phase laganbe observed (
+2 ◦
).Finally,1Pmomentoeientpreditionsarenotsatisfatorybothintermsofnorm
and phaselag.
Blade loading. In order to better understand the previous omparisons, load
Relative Radius
A m p li tu d e d /d r P h a s e l a g d /d r
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Figure 1.13: Blade load distribution modes in Case No. 1149. elsA omputations
are ompared to three HOST simulations: isolated, with installation eets, and
with installationeets andhub model.
this omparison isthe spanwise gradient of thethrust oeient
∂τ /∂ξ
:∂τ
∂ξ = ∂T
∂r
R
ρ ∞ N 2 (2R) 4
(1.3)Figure 1.13shows theload distributionomparison between elsA and thesame
three HOST-MINT simulations exposed in the previous paragraph. Only the two
rst modesof the Fast-Fourier Transform areshown: zeromode(mean value) and
rst mode (one per revolution, 1/rev). All HOST-MINT simulations give similar
results in terms of mean blade loading (left-hand side plot), and predit orretly
the load evolution for theentral part of the blade. However, an underestimation
anbenotiedfromthe
0.7R
spantothetipoftheblade. Thisloadunderestimation generates the already observed thrust mismathes. Moreover, notie that the loadnearthebladerootisnotwellpreditedbyisolated andinstalled HOSTsimulations,
while the hub eets gets loserto elsA.
When fousing onthe 1/rev mode,more important dierenes arenotied. On
oneside,thesolidredlineandthesymbolsarounditrepresentthenormofthe1/rev
mode, whih an belinkedto the normof 1P loads. Ontheother side, thedashed
redlineandthesymbolsarounditrepresentthephaselagofthe1/rev mode,whih
anbelinked tothe phaselagof1Ploads. Wheninstallationeetsareonsidered,
the1/revnorm getsloserto elsAresults,while keepingalmostthesame phaselag
thaninisolated simulations. However,important mismathesareinduedwhenthe
hub model isonsidered. Notie theoverestimation with respet to CFDresults of
both the modenorm and the mode phaselag. Theimplementedhubmodel,whih
should have only an important eet near the blade root, impats in fat all the
Table 1.4: Rotationalspeed sweep inAPIANexperimental high-speed tests
Test No. Mah No. Inidene RPM J
[℄ [
◦
℄ [min