Beruflich Dokumente
Kultur Dokumente
Vol.26,aNo. 2,a2013 1
DOI:10.3901/CJME.2013.02.***,availableonlineatwww.springerlink.comwww.cjmenet.comwww.cjmenet.com.cn
ScaleEffects onPropellerCavitatingHydrodynamicand HydroacousticPerformances
withNonuniformInflow
YANGQiongfang
1,
*
,WANG Yongsheng
1
,and ZHANGZhihong
2
1College of Marine PowerEngineering, NavalUniversityofEngineering, Wuhan 430033,China
2CollegeofScience, NavalUniversityofEngineering,Wuhan430033,China
Received April19,2012 revised December 20,2012 accepted December26,2012
Abstract: Considering the lack of theoretical models and ingredients necessary to explain the scaling of the results of propeller
cavitation inception and cavitating hydroacoustics from modeltests to fullscale currently, andtheinsufficientreflection of the nuclei
effects on cavitation in the numerical methods, the cavitating hydrodynamics and cavitation low frequency noise spectrum of three
geometrically similar 7bladed highly skewed propellers with nonuniform inflow are addressed. In this process, a numerical bridge
from the multiphase viscous simulation of propeller cavitation hydrodynamics to its hydroacoustics is built, and the scale effects on
performancesandtheapplicabilityofexistscalinglawareanalyzed.Theeffectsofnoncondensablegas(NCG)oncavitationinception
areinvolvedexplicitlyintheimprovedSauer'scavitationmodel,andthecavityvolumeaccelerationrelated to itscharacteristiclength is
used to produce the noise spectrum. Results show that, with the same cavitation number, the cavity extension on propeller blades
increases withdiameter associated withanearlier shift of the beginning point of thrust declineinduced by cavitation, whilethe three
decline slopes of thrust breakdown curves are found to be nearly the same. The power of the scaling law based on local Reynolds
numberaround0.9Rsectionisdeterminedas0.11.Asforthesmallestpropeller,thepredominanttonalnoiseislocatedatbladepassing
frequency(BPF), whereas 2BPF for the middle and both2BPF and 3BPF for the largest, which shows the cavitating line spectrum is
fullyrelatedtothe interactionbetween nonuniforminflowandfluctuatedcavityvolume.Thepredictedspectrumlevelexceedancefrom
themiddletothelargepropelleris6.65dB atBPFand 5.94dB at2BPF.Sinceitjustdifferslessthan2dBtothe incrementobtained by
empiricalscalinglaw,itisinferredthatthescaleeffectsonthemareacceptablewithasufficientmodelscale,andsodothescalinglaw.
Thenumericalimplementation of cavitating hydrodynamics and hydroacoustics prediction of propellerin big scalein wakehas been
completed.
Keywords: propeller,cavitationinception, cavitationnoise, scaleeffect,cavitationmodel,turbulencemodel
1 Introduction
*
As demonstrated experimentallytothecavitation of3D
hydrofoil and submerged body by KELLER
[1]
, obvious
scale effects including velocity scale and size scale effect
on cavitation inception and cavitation developing level
were presented. As for the marine propeller, SZANTYR
[2]
further addressed the scale effects in detail through
cavitation experiments. According to that, the tip vortex
inception cavitation number differed significantly between
the model with fully scale due to the differences in static
pressuredistribution,inwaterquality(noncondensablegas
content(NCG)), in boundary layer separation and in
velocity and geometry scale. And the propeller cavitation
extentwasalsodifferentevenwithasamecavitationindex.
Todemonstratethepracticalworthinessofthemediumsize
* Correspondingauthor.Email:yqfhaijun2008@126.com
This project is supported by National Natural Science Foundation of
China(GrantNo.51009144)
ChineseMechanicalEngineeringSociety andSpringerVerlagBerlinHeidelberg 2013
cavitation testing facilities in predicting the acoustic
characteristics of a propeller and to provide useful noise
datainfullscaleusingthescalinglawrecommendedbythe
18th ITTC Cavitation Committee, ATLAR, et al
[3]
,
presented cavitation tunnel tests of a model propeller and
the comparison between the extrapolated results to its
fullscale measurements. It was shown that a useful basis
for the propeller noise prediction in fullscale could be
provided by the extrapolation procedure mentioned with
about6dBto18dB discrepanciesin1kHzfrequencyband.
Referringtothescalinglaw,theincrementofthespectrum
levelfrommodeltofullscale is relatedexponentially to the
geometry scale, tip circumferential velocity, cavitation
index, fluid density and reference distance for which the
noiseispredicted. Asapracticalapplication,SHEN,etal
[4]
,
concluded that the exponent value of 0.28 represented the
Reynoldsnumbereffectonpropellercavitationnoiseatthe
critical point with 20 dB below the maximum noise level
for both thefullscale and model scale, through trial noise
measurementsof the fullscalenoiselevelofUSS212class
submarine and its model with a scaleratio of 14.5. Under
YYANGQiongfang,etal:ScaleEffectsonPropellerCavitatingHydrodynamicandHydroacousticPerformanceswithNonuniformInflow
2
this circumstance, the background noise had negligible
effect on cavitation noise due to the welldeveloping
cavitation. Additionally, if the cavitation inception speed
(CIS) at model scale pointed to the silent critical point on
theScurveofnoise,i.e.0dBabovethebackgroundnoise,
the exponent 0.315 would be used to predict the CIS of
fullscale, and the scale effects could be involved
reasonably. It means the exponent 0.315 is related to the
effect ofReynoldsnumber on welldevelopedcavitationof
propeller.
In the framework of single phase viscous simulation to
the scaling of tip vortex cavitation inception noise of
hydrofoil and ship propeller, HSIAO, et al
[56]
, PARK, et
al
[7]
had proposed integrated technical solutions in the
recent past. Exactly, in conjunction with the Reynolds
averaged NavierStokes(RANS) computations for single
phase flow field, the spherical or nonspherical micro
bubble dynamics modelwith or withoutaccounting for the
nuclei size distribution was used to detect the cavitation
inception witha visual or an acoustic criterion. Due to the
unawareness of the predefined threshold, it is difficult to
apply this method directly to the propeller cavitation
inception and cavitating hydroacoustics prediction at
fullscale until now, then there is little prospect that the
scale effects on them could be clearly clarified. According
to SINGHAL, et al
[8]
and GINDROZ, et al
[9]
, both volume
fraction and mass fraction of NCG and the turbulent
pressure fluctuation will affect the cavitation inception
significantly. Likewise, even the freestream turbulence is
one of the main factors contributing to the scale effects on
the inception of cavitation, which has been found by
KORKUT, et al
[10]
this year. Consequently, adding these
effects explicitly into the process of cavitation turbulent
flowsimulationissupposedtobemorereasonablethanthe
single way solution with nuclei effect involved as a
supplement. It is just the original calling to the cavitation
multiphase simulationforcavitatingnoiseprediction.
However, benefiting from the fast and convenient
realizationofthepotentialflowtheory,theinviscidsurface
panel method based on velocity has been widely used in
propeller sheet cavitation simulation and its radiated noise
evaluation in the recent years, including SALVATORE, et
al
[11]
, EKINCI, et al
[12]
, SEOL, et al
[13]
. Following the
realization, the single pulsing spherical bubble radiated
noise theory is mostly used to compute thenoise spectrum
ofcavitatingmonopolesource,likeHU
[14]
andZHANG
[15]
.
From another perspective, SEOL, et al
[13]
and TESTA
[16]
both used a more robust mathematical model, i.e. the
FfowcsWilliamsandHawkingsintegralequationtopredict
thepropellercavitationnoise.Althoughthepredictedsound
pressure level at blade passing frequency(BPF) and its
harmonics were found to be good agreement with that of
experiment results, the NCG effects are still hard to be
carriedintotheprocessofnumericalsimulation.
Currently,manyresearchers'attentionhasturnedbackto
cavitation multiphase simulation for higher accuracy with
the prospect of developing better cavitation model or
depending on model calibration. Since all the multiphase
flow model, turbulent model, cavitation model and phase
change threshold pressure acting as a whole to affect the
cavitation simulation, no optimal composite models acing
as a universal solution to cavitating hydrodynamics in
marine engineering have been brought up along with
integrated validation until now. Due to the popularization
of commercial CFD solvers, the mixture multiphase
modeling method develops relatively fast and acceptably,
and several cavitation models based on homogeneous
multiphase transport equations have proved to be
extremely valuable for marine propeller cavitation flow
recently, like SINGHALs model
[8]
, SAUERs model
[17]
,
KUNZs model
[18]
, and ZWARTs model
[19]
. Among these
models, RHEE, et al
[20]
used the SINGHAL full model
embedded in FLUENT 6.1 code to simulate MP 017
propeller sheet cavitation and predict its thrust breakdown
performance,LINDAU,etal
[21]
usedKunzmodelinvolved
in UNCLEM software for the cavity pattern simulation
and cavitation breakdown performance maps prediction of
the NSRDC 4381 and INSEAN E779A propellers, and
addedanaxialflowwaterjet.Pointingtothesamepropeller
P4381, KIM
[22]
chose the Sauer model within the open
source code OpenFOAM to verify its simulated cavity
extension, and a higher accuracy was shown at last.
Additionally, for analyzing and comparing different
numerical cavitation models to serve as a complement to
experimentsforwaterjetpumpcavitatingflow,OLSSON
[23]
investigated the SINGHAL, KUNZ and SAUERs model
combinedwith e - k RNGturbulentmodelwithturbulent
viscosity modification by density function on sheet
cavitation ofhydrofoilandpump.Itwasshownthat,asfor
the rotary cavitating flow, the Sauer model was most
efficient for global force prediction and the cavitation
breakoff point determination. Furthermore, MORGUT, et
al
[24]
assessed the three mass transfer models, i.e.
SINGHAL,KUNZandZWARTactivatedin ANSYSCFX
12 commercial code on E779A propellers cavity patterns
and global force validation. It was clarified that the two
equation SST turbulence model could guarantee the same
level of accuracy as the computationally more expensive
BSLRSMturbulencemodel.Andthecomputationalresults
obtained from the particular condition, using alternatively
the different welltuned mass transfer models, were very
close to each other. In addition, JI, et al
[25]
, also chose the
default ZWART cavitation model and SST turbulence
model in CFX code to investigate the pressure fluctuation
induced by cavitating propeller with nonuniform inflow.
The amplitudes of first three harmonics of BPF were
acceptably predicted with a maximum difference reaching
to 20%. From these reviews, it is concluded that, with
reasonablephysicalmodelsincludingbothcavitationmodel
andturbulencemodel,addingagoodmeshingstrategyand
an efficient and robust CFD solver, CFX for instance,
reliable numerical cavitating flow can be represented and
CHINESEJOURNALOFMECHANICALENGINEERING
3
visualized.
More recently, accounting for the effects of both mass
fraction and volume fraction of the NCG on cavitation
inceptionatthesametime,YANG,etal
[2628]
introducedan
improved SAUERs cavitation model associated with a
modified SST turbulent model by Compiling Expression
Language(CEL) in CFX 12 solver to investigate the
propeller cavitating turbulent flow and cavitation inception
phenomena. Using these models, the effects of both NCG
and turbulence on cavitation inception and cavitating
bubblegrowthwereconsideredbythe equationsofmixture
density and phasechanging threshold pressure. Since the
pressure distribution was the decisive factor for cavitation
inception in numerical simulation, a rule that when the
cavitation indexreaches toits criticalnumber, the pressure
coefficientdistributionaroundacertainbladesection, i.e.at
theradiusof0.9R,isunaltered hasbeendrawntodetermine
the propeller cavitation inception time. This numerical
processisjustcontrarytothatincavitationtunneltestswith
pressurereduction.Thephysicalgascontentismodeledby
constants of NCG mass fraction and volume fraction. The
critical point of cavitation inception is that, as increasing
cavitationnumber,thepressurecoefficientaroundthegiven
section is nearly superposed to the result in singlephase
noncavitation simulation. Then, the visual or acoustic
criteria of cavitation inception can be achieved by the
relative changing amplitude of the separate L2 norms of
pressure coefficient acting as a similar manner to
verification and validation of a point variable. As a result,
the inception buckets were predicted satisfactorilyfor both
noskewed and highly skewed propellers by multiphase
flow CFD calculation method initially. On this basis, in
order to investigate the effects of thrust loads and cavity
extension on the discrete line spectrum frequency and its
spectrum source level, a method coupling the multiphase
flow cavitation simulation with pulsating spherical bubble
radiated noise theory has been undertaken to predict both
the 5 and 7balded propellers cavitating noise spectrum
by YANG, et al
[29]
. A numerical system to measure the
cavitating hydrodynamics and noise performances of the
shippropellershasbeenconstructed there.
Continuing to demonstrate the viability and importance
ofthismethodforpropellerperformancesassessmentinbig
scale, and describe the scale effects on cavitating
breakdownperformances,cavitationinceptionandradiated
tonal noise at low frequency, three geometrically similar
7bladedhighlyskewedpropellersofdifferentscalesand of
the same nondimensional incoming wake flow are
addressed in this research. According to the numerical
results,the applicability ofthe empiricalscalinglaw willbe
analyzed to give deep insights into cavitation noise
predictionforpropellerinfullscale.
In the following, the numerical models employed are
presented firstly in section 2, and followed by their
calibration and validation in section 3,including cavitation
model,turbulencemodelandcavitatingnoisemathematical
model.Thenthescale effects on cavitatinghydrodynamics
and hydroacoustics of three propellers are provided in
section4. Attheend,ourconcludingremarksaregiven.
2 NumericalModels
2.1 Cavitationmodelandturbulencemodel
According to Refs. [2628], the mixture multiphase
flow model including both the NCG mass and volume
fractioninmixturedensityisusedagain.Boththeeffectsof
nuclei distribution and dissolved air content on cavitation
inception can be considered explicitly in that way. At this
moment,the total numberofmixturephaseisn=3,whichis
more reasonable than the reviewed literatures with two
phases.Sincethevelocityslipbetweenthemainphaseand
secondary phase is rather small for high Reynolds number
and small vapor bubbles, the assumption for no velocity
slipis stillused as the sameasthatin Refs. [1925]. Ona
volumefractionbasis, themixturedensityisexpressedas
) 1 /( ) ) 1 ( (
g l g v v v m
f - - - + = r a a r a r , (1)
where
v
a and
v
r are vapor volume fraction and its
density respectively,
g
a and
g
f are volume and mass
fractionofNCGrespectively,
l
r isthewaterdensity.The
mass transferred through cavitation is modeled by the
transportequationforthevapormassfraction:
) , , 1 (
m
n i f
i
i i
=
r
r
a , (2)
in which subscript i stands for the ith phase. And the
transportequationforthevaporis
A f f
t
= +
) ( ) (
v m m v m
v r r , (3)
where
v
f is vapour mass fraction,
m
v is the mixture
velocity. A is the mass transfer source between water and
vapour, which is usually divided into two items
- +
+ = m m A & & . In which
+
m & and
-
m & stand for vapor
vaporization (bubble growth) and condensation (bubble
collapse) process respectively. The improved Sauer
cavitation model is presented here again. Its mass transfer
ratesarewrittenas
-
-
=
-
- -
=
-
+
), sgn(
3
2 3
), sgn(
3
2
) 1 ( 3
v
l
v
B
v v d
v
l
v
B
v v g p
p p
p p
R
C
m
p p
p p
R
C
m
r
r a
r
r a a
&
&
(4)
where empirical coefficients are 50
p
= C and 01 . 0
d
= C .
Thephasechangethresholdpressure
v
p ismodeledby
YYANGQiongfang,etal:ScaleEffectsonPropellerCavitatingHydrodynamicandHydroacousticPerformanceswithNonuniformInflow
4
) 39 . 0 (
2
1
m sat v
k p p r + = , (5)
in which k is turbulent kinetic energy,
sat
p is the
saturation vapor pressure with 3540Pa. With calculative
experiences, optimum NCG mass fraction and volume
fraction are
6
g
10 1
-
= f and
4
g
10 8 . 7
-
= a
respectively, and the bubble initial average radius is
5 . 1
B
= R m. Explicitly, an earlier cavitation inception is
represented by the combination of mixture density and the
turbulence fluctuation. And the NCG effects are included
directlybythisformulation.SimilartoEq.(5),FRANK,et
al
[30]
adoptedafurtherslightmodificationas
) 1 ( 39 . 0
v m sat v
a r - + = k p p (6)
to apply the pressure turbulence term only in vapor phase
presencesotospeedupthesolutionconvergence.Sincethe
calculated vapor volume fraction with isosurface
5 . 0
v
= a to limit the propeller sheet cavity extension
seems reasonable in reviewed Refs. [2122, 24] except JI,
et al
[25]
and KIM
[22]
with 1 . 0
v
= a , the two modifications
are the sameunderthat occasion.Actually, theappropriate
isosurface boundary depends on the initial cavitation
inception time according to the comprehensive validation
of cavity area of E779A propeller in Ref. [26]. If the
modified phasechange threshold pressure is used, the
cavitation inception is earlier than that being controlled by
theconstantvaporpressureofwhathadbeenusedbyJI,et
al and KIM, so to produce more vapor volume fraction
under steady numerical tests, and a larger isosurface
boundaryisneededsubsequently.
Comparing to the Sauer model, its differences are as
follows:
(1) Introducing the NCG volume fraction to vapor
vaporization term and replacing the density ratio
m l
/ r r
asconstantcoefficientC.
(2)Expressingtheradius of bubbles in denominator asa
function of vapor volume fraction
v
a and replacing the
number of bubbles per unit volume
0
n to the initial
constant 5 . 1
B
= R m.
(3) Determining the initial vapor volume fraction
v0
a
bytheinitialnumberofbubblesgiven.
The improvements made by the author mainly locate at
involving the NCG mass and volume fractions in mixture
density at the same time and determining their reasonable
values after a large number of tests on cavitating flows of
NACA 66 (mod) hydrofoil and propeller NSRDC 4381 in
Refs. [3031]. Following Eqs. (1), (2), (4), (5), the
g
f
directly influences the
m
r , and then to both turbulent
viscosity
t
m and
v
p , so that the cavitation inception
pointisaffected.Atthesametime,
g
a directlydrivesthe
evaporation rates
+
m & to change, and then to change the
cavityextension.
To solve the turbulent cavitating flow, the SST model
with modified wall function is implemented to provide
turbulence closure in this research to overcome the
singularityatseparationpointswherethe nearwallvelocity
approaches zero, so that they can be applied to any fine
mesh element. And a further calibration of the turbulent
viscosityisintroducedas
w r m /
m t
k = (7)
referringtoRefs.[20,32].Intheequation, w isturbulent
vortex frequency. The physical interpretation of this
calibration is in respect of the numerical calculation.
Specifically, witha localrefinement grid topology, Eq. (7)
willenlargetheapplicationscopeoflow Reynoldsnumber
equation to some extent, so to perform better the
applicability of w transport equation to model separate
flowwithcurvaturechangeandtheunsensitivity ofthe e
equationtofreestreamvortexfrequency.
2.2 Cavitatingnoisemathematicalmodel
According to tunnel tests in model scale, propeller
cavitation are mainly divided into three types: tip vortex
cavitation, blade surface cavitation, and hub vortex
cavitation. Of the various types of cavitation, because of
thevortexcavitationremaininginnegativepressureregions
for relatively longer time and collapsing with less
fluctuatedenergy,thesuctionsidesheetcavitationproduces
the highest noise level, and hub vortex cavitation the
least
[33]
.Inotherwords,thesheetcavitationonsuctionside
isthemainnoise sourceof acavitatingpropeller.
In the field of timedomain acoustic analogy for noise
prediction, both the propeller blade thickness rotation and
unsteady sheet cavity volume fluctuation can besimplified
as monopole sources, while the blade surface pressure
fluctuation called loading noise is equivalent to the dipole
source term and the stochastic noise with broadband
characteristics is modeled as a quadrupole source term.
Applying this method, SEOL, et al
[34]
, had proved
numerically that the thicknessnoisecomponent ofpropeller
was negligible comparing to the unsteady loading noise
especially under thenonuniform inflow condition. Due to
the internal low Mach number, the fluctuated volume
radiating noise component will dominate the propeller's
farfieldnoiseintensity oncethecavitationoccurs.Itmeans,
precisely trackingthe cavity volumeperiodicfluctuation is
essentialtodeterminethemainnoisesourceofacavitating
propeller.
As an application of the Ffowcs WilliamsHawkings
equation integral method, Farassat 1A formulation in
timedomain can predict noise from an arbitrary shaped
object in motion without the numerical differentiation of
the observer time
[35]
. When the turbulent quadrupole noise
source is neglected,thesound pressureat observer (x,t)is
the sum of thickness noise p
T
' and loading noise p
L
'
contributions of all source nodes on the object. It is
expressedas
CHINESEJOURNALOFMECHANICALENGINEERING
5
) , ( ) , ( ) , (
'
L
'
T
'
t p t p t p x x x + = , (8)
where the thickness noise component is obtained by
introducing the sound Green's function in unbounded
field
[36]
,
S
R
v
G c
S v v G p
S
n
n n
S
d )) 1 ( (
d
1 1
1
s
'
s
s s
s
'
T
D R R
D
R R
D
D
- - -
-
+
+
-
-
=
r
gr r
& &
&
, (9)
in which the first term in right accounts for the nearfield
while the second for farfield, and the Greens function is
read as
) 1 ( 4
s
'
D -
=
R
G
g
, (10)
in which
'
R is the distance from source to observer,
inflow Mach number c /
s
V =
,
s
V is inflow speed,
2
2
1
1
-
=
g , the source Mach number relative to fixed
reference frame is c /
s
t
=
y
, the radiation vector is
equal to
- = R D
2
g g ,
' 2
/ ) ) ( ( R
+ = r r R g ,
observersource distance y x r - = , n l p = is the force
extractedfrompressure,
n
v isthesurfacenormalvelocity.
Variable t meanstheintegrationatretardedtime.
Thecalculationsofretardedtime t forany pointonthe
blade and its normal velocity
n
v are crucial to the
'
T
p .
Once the sheet cavitation occurs, the normal velocity of
partial acoustic nodes shifts from blade surface to the
cavity surface. With a zero thickness hypothesis of the
cavity in panel surface method, extracting the cavity
normalvelocity isthesameasthatofnoncavitation.While
in the viscous CFD simulation, since the sheet cavity
extension is visualized byisosurface of the vapor volume
fractionmapped on CFDmesh elements,wherein the fluid
variables are extrapolated from the centered nodes rather
then the element surfaces in generic CFD solvers, hence,
the predicted cavity surface is not closely coincident with
the blade surface even with extremely small first height
nodes. At this moment, the discrepancy between the
acousticsourcenodes information output from CFD solver
in every transient timestep with the mesh element on
blades will be a barrier into the hydroacoustic tool. This
problemis beingactivelysought inauthor'steamnow.
Fromanotherpointofview,thepulsatingsphericalnoise
sourceradiatingnoiseisgivenby
2
c
2
l '
d
d
4
) (
t
V
r
r p
p
r
= , (11)
where
c
V isthecavityvolume.Ifthewholeblades'cavity
extensionis equivalent toaspherical source with the same
volume, the farfield radiated noise at any point can be
obtained by this equation. Neither by the surface panel
method nor the viscous CFD calculation, the
periodicfluctuated sheet cavity area versus rotating angles
can be integrated directly in postprocess of the results.
When the spherical hypothesis is used, the cavity volume
can be easily obtained to predict its the sound pressure.
Refs. [1415] just used this simplification in conjunction
with the surface panel method to calculate propeller
cavitation noise. However, the cavities on propeller blades
arefarfrombeingsphericalaccordingtothe experiments in
PEREIAR, et al
[3738]
. The vapor extension (cavity area)
generated by unsteady sheet cavitation of both hydrofoil
and propeller can be represented by a characteristic length
c
l ,andthecavityvolume
c
V isproportionalto
c
l .Their
specific expressions are writtenas
c c
E l = , )) d / d ( 3 ) d / d ( 6 (
d
d
2
c
2 2
c
2
c c
2
c
2
t l l t l l
t
V
+ = . (12)
Asaresult,aimsatpredictingpropellercavitatingnoise,
the fluctuated cavity areais capturedby viscous CFD runs
firstly,andthenthesecondderivativeofthecavityvolume
is obtained byEq.(12)to deduce the sound pressureusing
Eq. (11). It is supposed to be more reasonable than the
sphericalcavityassumptionintheviewofflowpatterns.
3 Validationofthecavitationsimulation
Withablockstructuredandflowadaptivegridtopology,
Fig. 1 shows the surface mesh of propeller E779Aand its
simulatedcavitypatternsatadvanceratio J=0.77and cavity
areasunderlightly,moderatelyandheavilycavitationlevel
conditionsby theimprovedSauermodelandmodifiedSST
model above. The cavitation index s based on inflow
speed and the rotating speed cavitation index
n
s are
introduced into the numerical calculation to control the
pressure
out
p on the outlet surface of numerical domain
afteractivatingthecavitationmodel,whicharedefinedas
2
s l
v out
5 . 0 v
p p
r
s
-
= ,
2
l
v out
) ( 5 . 0 nD
p p
n
r
s
-
= , (13)
where
s
v isincomingflowvelocity,nandDarepropeller
rotatingspeedanddiameterrespectively.As
n
s decreases,
the boundary pressure changes to represent the
pressuredecreasingincavitationtests.Inthefigure,
c
A is
thecavityareaandnotethat
0
A isthebladefaceareafor
r/R0.3 with respect to the experiment
[39]
. Under
moderatelycavitationlevelcondition( 25 . 0 / 1 . 0
0 c
< A A ),
YYANGQiongfang,etal:ScaleEffectsonPropellerCavitatingHydrodynamicandHydroacousticPerformanceswithNonuniformInflow
6
the simulated cavity area is very close to the experiment,
wherein a little small for the lightly cavitation level
( 1 . 0 /
0 c
< A A )duetothewallroughnesseffectandbigger
than the measurement for heavily condition
( 5 . 0 / 25 . 0
0 c
< A A ) because of the bubble cavitation
area being included. The numerical domain and boundary
conditions of this test are shown in Fig. 2. Uniform
incoming velocity and averaged static pressure arelocated
on inlet and outlet respectively. The secondorder upwind
schemeisusedfor theconvectiontermandhighresolution
option for turbulence numeric accuracy in the solving of
governingequations.
783 . 1 =
n
s 069 . 0 /
0 c
= A A 082 . 2 =
n
s 045 . 0 /
0 c
= A A 5 . 2 =
n
s 024 . 0 /
0 c
= A A
2 . 1 =
n
s 237 . 0 /
0 c
= A A
77 . 0 = J
5 . 1 =
n
s 105 . 0 /
0 c
= A A
0 . 3 =
n
s 011 . 0 /
0 c
= A A
0 . 1 =
n
s 168 . 0 /
0 c
= A A
n
s
0
Exp. _ 71 . 0 = J
Exp. _ 77 . 0 = J
Exp. _ 83 . 0 = J
Exp. _ 88 . 0 = J
cav. _ 3 G _ 71 . 0 = J
cav. _ 3 G _ 77 . 0 = J
Fig.1. Validationofthepropellersheetcavitation
~6D ~2.8D
~.65D
Innerrotatingsubdomain
Dummyshroud
Inlet
s
v
Fixedsubdomain
~4D
Outlet
out
P
Noslipwall
~1.5D
Shaft
Fig.2. Numericaldomainandboundaryconditions
Using the inception rule when
i
s s > , the pressure
coefficients distribution of blade tip section is relatively
unaltered, the predicted visual tip vortex cavitation
inception number is proved to be an excellent agreement
with the experiment, which can be seen in Ref. [26] in
detail. From the qualitative and quantitative comparison,it
isseenthat,withtheaidofproperrefinementgridtopology
around the blade surface, the adopted cavitation model
combined with the modified turbulence model are
extremely valuable for sheet cavitation simulation under
the level of moderately cavitation, and also be able to
capturewellthebeginningpointofcavitationinceptionand
developingarea ofcavitation.
4 Scaleeffectsanalysis
4.1 Scaleeffectsoncavitationhydrodynamics
In the following, all the simulations are undertaken on
three similar 7bladed propellers with diameter 250 mm,
500mmand1 000 mmrespectively.Forsimplicitytheyare
named as small, middle and big propeller. Their single
passage numerical domains and corresponding hexed
structure meshes are all completed by procedural
realizations presented in Ref. [40]. Note that, the mesh
nodes density on leading edgeregion, trailing edgeregion,
tip section area and blades surface are all localrefined
gradually with diameterincrease.The surface mesh details
areseeninFig.3.Inordertominimizetheeffectsofmesh
quality differences, the mesh minimum determinant
indexesofthreepropellersareallabove0.2associatedwith
closemeshdensityandaverageYplusdistributiononblade
surfaces. The total number of mesh nodes in three single
passage domains is controlled with a grid refinement ratio
7 . 1
G
= r . For decreasing the numerical errors induced by
variables interpolation between periodic interfaces with
unmatching mesh nodes, the fullpassage numerical
domainsofthreepropellersare includedinthecalculations.
The numerical domain and boundary conditions in
noncavitationsinglephaseRANScalculationarethesame
as that in Fig. 2. Fig. 4 shows the calculated open water
characteristics of three propellers. This figure also shows
the results of the model tests for the small propeller. The
agreementis seen to excellent good again. Under the same
CHINESEJOURNALOFMECHANICALENGINEERING
7
advance ratio, the thrust coefficient
t
K increases but
torque coefficient
q
K decreases with larger Reynolds
number of bigger geometry scale. So the derived open
water efficiency increases obviously with diameter
associated with a smaller increase rate further away the
designpoint.In thefigure, thevariablesaredefinedas
nD
v
J
s
= ,
4 2
t
D n
T
K
r
= ,
5 2
q
D n
Q
K
r
= ,
q
t 0
0
2 K
K J
= h ,(14)
where T and Q are axial thrust and break torque
respectively. Subscript 0 stands for the uniform inflow
condition, and subscripts s, m and b stand for the small,
middle and big respectively. The rotating speeds of three
propellers are set as 20
s
= n r/s, 15
m
= n r/s, 10
b
= n r/s
to match the tunnel tests. So the Reynolds number
n
Re
basedonrotatingspeedanddiameterdiffersbyanorderof
magnitude between the small and big propeller, which can
beenlargedsubsequently tofullscaleanalysis.
Gridtopology
= =
n b
, 00 0 1 N D = =
n m
, 500 N D = =
n s
, 250 N D
Hexedmeshdeterminantdistribution
201978 mm 342652 mm 590538 mm
Fig.3. Gridtopology,meshdeterminantandsurfacemesh
detailsof allscaledpropellers
J
t
K
q
10K
0
h
0
_ mm 250 h D
0
_ mm 500 h D
0
_ mm 1000 h D
,
1
0
K
1
0
K
T
h
r
u
s
t
c
o
e
f
f
i
c
i
e
n
t
t
o
r
q
u
e
c
o
e
f
f
i
c
i
e
n
t
1
0
K
s
Fig.7. Thrustandtorquebreakdownperformancesof propellers
with uniformandnonuniforminflow
6 . 0 _
0 b
= J D
403 . 0 _ p b = J D
6 . 0 _
0 s
= J D 403 . 0 _
p s
= J D
6 . 0 _
0 m
= J D 403 . 0 _
p m
= J D
0039 . 0 / , 0 . 4 ,
0 c b
= = A A D s 0054 . 0 / , 75 . 3 ,
0 c m
= = A A D s 0114 . 0 / , 5 . 3 ,
0 c s
= = A A D s
s
0
403 . 0 , 250
p s
= = J D
c z /
P
r
e
s
s
u
r
e
c
o
e
f
f
i
c
i
e
n
t
(b)
5
4
3
2
1
0
1
2
3
4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
5 . 3 = s 75 . 3 = s 25 . 3 = s
p
c z /
5 . 3 = s
0 . 3 = s
5 . 2 = s
25 . 3 = s
Nocav.
0 . 4 = s
75 . 3 = s
5 . 4 = s
403 . 0 , 500
p m
= = J D
Nondimensionalchordlength
mm
5 . 3 = s
0 . 3 = s
5 . 2 = s
25 . 3 = s
0 . 4 = s
75 . 3 = s
0 . 5 = s
5 . 3 = s 0 . 4 = s 75 . 3 = s
403 . 0 , 1000
p b
= = J D
p
c z /
P
r
e
s
s
u
r
e
c
o
e
f
f
i
c
i
e
n
t
Fig.10. Thrustbreakdownperformancesof4381propeller
Following the predicted inception number, if the effect
on cavitation inception number is related to the Reynolds
number, the cavitation index
n
s based on rotating speed
canbededuced from s .Exactly,theinitialindexesbased
onrotatingspeedare =
i n
s 1.44,1.62and1.8respectively
for the three propellers. When the local Reynolds number
based on the total velocity and chord of 0.9R section is
introduced, the flow parameters of all scales are shown in
Tab.1. The localReynoldsnumberisdefined as
n n / ) 9 . 0 ( / Re
9 . 0
2 2
a 9 . 0 9 . 0 9 . 0 R R R R
c nD v c v + = = , (15)
where
R
c
9 . 0
is chordof 0.9R section, n is viscosity,
a
v
is the axial velocity component. If the classical power law
relationship between the Reynolds number and cavitation
inceptionnumberisused,
g
s Re
i
n
, (16)
the powervalueof g equalsto0.11,whichis smallerthan
thevalueof0.22calculatedbyHSIAO,atal
[5]
.However,it
is in a satisfactory accordance with the power 0.12 found
numerically in Ref. [41] for a highlyskewed 5bladed
propeller.
Table 1. Comparison of parameters of three propellers
Levelrank Small Middle Big
Diameter D/mm 250 500 1000
Rotatingspeed n/r/s 20 15 10
Inflowspeed va /m/s 2.016 3.024 4.032
Chord C0.9R /mm 694.8 1389.6 2779.2
ReynoldsnumberRe0.9R 9.8510
6
2.9610
7
7.8810
7
ReynoldsnumberRen 1.2410
6
3.7210
6
9.9310
6
Inceptionnumberi 4.0 4.5 5.0
Inceptionnumberni 1.44 1.62 1.8
4.2 Scaleeffectsoncavitatinghydroacoustics
As mentioned in section 2.2, the periodicpulsating
cavity area or the cavity volume determines the low
frequency cavitating noise spectrum, including its tonal
componentsandspectrumlevel.Thetimedependentsignal
can be obtained by cavitation transient simulation. In
numerical tests, the iterative timestep of three propellers
are set as
4
s
10 78 . 2
-
= Dt s,
4
m
10 7 . 3
-
= Dt s and
4
b
10 56 . 5
-
= Dt s respectively, which are all associated
YYANGQiongfang,etal:ScaleEffectsonPropellerCavitatingHydrodynamicandHydroacousticPerformanceswithNonuniformInflow
10
with2degreesofbladerotation.Accordingtothesampling
theorem, their relating effective maximum frequencies are
1800Hz,1350Hzand900Hz.Afterrunningforfivecycles,
alltheflowvariablesextractedforanalysisareoutputfrom
thefifthrevolution.
Fig. 11 shows cavity area fluctuation versus azimuth
angles under the condition of 5 . 2 , 403 . 0
p
= = s J for
threepropellers,theirfluctuationinfrequencydomainwith
nondimensional frequency n f St / = are also shown in
thisfigure.Itseemsthattwopeaksareappearedinthetime
domainbothforthemiddleandbigpropeller.Regardingto
the frequency domain, only the axial passing frequency
(APF)andBPFtonalcomponentsexistfor smallpropeller,
while both BPF and 2BPF line spectrum dominate the
middlepropeller'ssignal.ExceptingfortheBPFand2BPF
components, the BPF harmonics stretching to 5 are still
obvious for the big propeller. Besides, the fluctuating
amplitude of cavity area increases significantly with the
geometryscale.
n f St / =
3
10
-
0 10 20 30 40 50 60 70 80
0
0.5
1.0
1.5
2.0
2.5
BPF 2BPF
3BPF
4BPF
5BPF
250 s = D
500
m
= D
000 1 b = D
(b) Frequencydomain
0.6
0.61
0.62
0.63
40
80
120
160
200
240
280
320
360
/ q
0
c
/
250 _ / 3 . 1 s 0 c = D A A
500 _ / 1 . 1 m 0 c = D A A
1000 _ / b 0 c = D A A
Circumferentialangleofbladeone
(a)Timedomain
mm
mm
mm
Nondimensionalfrequency
mm
mm
mm
Fig.11. Cavityareafluctuationunder 5 . 2 , 403 . 0
p
= = s J
After the acceleration of cavity volume is obtained, the
cavitating low frequency noise spectrum can be plotted by
substituting its second derivative into Eq. (11). Fig. 12
shows the calculatednoise of all scales undernonuniform
inflow condition. The noise spectrum predicted by
spherical cavity hypothesis are also shown in it for
comparison.Thesourceobserverdistanceofsmall,middle
and big propeller is 1
s
= r m, 2
m
= r m, and 4
b
= r m
correspondingtothesamerelativedistance.Itisfoundthat
the sound pressure calculated by spherical cavity is bigger
than that by the cavitation characteristic length for all the
scales. Exactly, the amplitude exceedance at the
predominatedBPFfrequencyis3.22dB,2.55dBand2.17
dB for the small, middle and big propeller respectively. It
means the predicted difference between these two
approaches increases inversely with the geometry scale.
Using the cavitation characteristic length, the predicted
cavitating spectrum level at BPF frequency is 153.92 dB,
173.15dBand 179.81dBre. 1Paand 1 Hz forthethree
propellers. At the same time, the noise increment at BPF
frequency from small to middle propeller is 19.23 dB and
33.02 dB at 2BPF. However, the noise enhancement from
themiddletobigpropellerisonly 6.65dBatBPFand5.94
dB at 2BPF, which differs markedly to the conclusion of
radiating equivalent sound pressure level from different
scaled airpropeller measured with uniform inflow and the
same relative distance in wind tunnel in Ref. [42]. One
reasonisthedifferenceofincominginflow,andtheotheris
the scale effect of developing cavitation on noise. On this
point, the difference of tonal component frequencies for
threepropellersisjust attributedto the interactionsbetween
different cavitation level with a similar incoming flow. In
detail, the predominate line spectrum of small propeller is
located at BPF, while the 2BPF for the middle and both
2BPF and 3BPF for the big. Additionally, it also draws a
conclusion that, with the same scale ratio and a similar
observer distance, the scale effects are weakened a lot by
theincreaseof modelscale.
250 = s D
c
V
c
V
CHINESEJOURNALOFMECHANICALENGINEERING
11
4
10
500
m
= D
c
V
c
V
S
p
e
c
t
r
u
m
l
e
v
e
l
S
L
/
d
B
P
r
e
s
s
u
r
e
f
l
u
c
t
u
a
t
i
o
n
p
/
P
a
4
10
000 1 b = D
c
V
c
V
S
p
e
c
t
r
u
m
l
e
v
e
l
S
L
/
d
B
P
r
e
s
s
u
r
e
f
l
u
c
t
u
a
t
i
o
n
p
/
P
a