DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.

org
doi:10.14355/daoe.2016.05.001

UnsteadyWakeRollupModelingUsinga
MollifierBasedFilteringTechnique
GerasimosK.Politis
Professor,DepartmentofShipandMarineHydrodynamics,SchoolofNavalArchitectureandMarineEngineering,
NationalTechnicalUniversityofAthens,HeroonPolitechniou9,Zografos,Athens,Greece
polit@central.ntua.gr

Abstract
3DBEMformulationsofunsteadyliftingflowproblemsareparadigmsoffreeboundaryproblems.Thefreeboundaryentersin
theBEMformalismasthenecessaryspatialsupportofthevelocitydiscontinuity,generatedatthetrailingedgesofthelifting
bodies.InthecontextofaBEM,thisvelocitydiscontinuity,alternativelytermedashearlayer,issimulatedbyasurfacedipole
distributionoritsequivalentsingularvortexsheet.Freevortexsheetsareinherentlyunstableandamenabletotwowellknown
instabilities:theKelvinHelmholtzinstabilityandtherollupoftheirfreeedges.Asaresult,theyshowchaoticbehaviorwith
the passage of time. On the other hand, experimental evidence on flows around lift producing devices shows that the wake
vorticityisorganizedinspecific,problemdependentrolluppatterns.Inthispaperwepresentafilteringtechnique,whichby
introducing artificial viscosity to our problem, suppresses the smaller scale instabilities, leaving the larger scale organized
vorticestodeterminetherolluppatternofthefreewakevorticity.Flowsimulationsarepresentedforasteadilyadvancingwing,
a biomimetic wing at two different Strouhal numbers, a naval propeller at two different advance coefficients and a novel
flexibleoscillatingductpropulsiondevice.Theeffectofthefilteringparameterstotheshearlayerwakerolluppatternaswellas
tothecalculatedunsteadyforcesandmomentsisshownanddiscussed.
Keywords
UnsteadyShearLayerDynamics;UnsteadyWakeModeling;LiftingFlowsaroundFlexibleBodies;BoundaryElementMethod

Introduction
Boundary element methods (BEM) have been long applied for the solution of flow problems around systems of
streamlinedliftingornonliftingbodies.IntheexcellentbookofKatz&Plotkin(2001)[10],almostallaspectsand
the state of the art of modern BEM formalisms, for fluid flow problems, are presented and discussed. BEM
formulationofflowproblemsaroundliftingbodiesisaparadigmofafreeboundaryproblem.Thefreeboundary
entersasthespatialsupportofthetangentialvelocitydiscontinuity,generatedatthetrailingedgesofthewings.In
thecontextofaBEM,thistangentialvelocitydiscontinuityissimulatedbyadipoledistributionoritsequivalent
singular vortex sheet. Typical remedies for the determination of this free boundary are: (a) the prescribed wake
shapemethodor(equivalently)thefrozenwakemodelbywhichthegeometryofthefreeboundaryissomehowa
prioriselected,(b)thewakerelaxationmethodbywhichaninitialsupportinggeometryisassumedandacorrector
method based on Helmholtz vortex theorems is applied to obtain the final geometry, and (c) the time stepping
algorithmbywhichthesupportinggeometryisderivedbyapplyingthefreeshearlayerkinematicanddynamic
boundary conditions in a step by step basis, Politis (2004) [18], Wu et al. (2006) [35]. Remedies (a) and (b) are
applicableineithersteadyflowproblemsorinunsteadyflowproblems,wheretheunsteadyflowfielddoesnot
affectconsiderablythefreeshearlayerposition.Remedy(c)ismoregeneralandcanbeappliedwithoutlimitations
tothesolutionofanysteadyorunsteadyflowproblem.Unfortunately,applicationofremedy(c)requirescoping
withtheinherentfreesingularvortexsheetinstabilities.Morespecifically,itiswellknownthatasingularvortex
sheet is amenable to chaotic behavior due to a KelvinHelmholtz instability. Furthermore, the edge of the sheet
inducesinfinitevelocitiestoitself,factwhichisrelatedwiththetipvortexrollup.Finally,asingularvortexsheet
induces a nonphysical tangential fluid velocity jump in its vicinity, notexistent in a viscous environment. An
extensive discussion of the free vortex sheet instabilities, their chaotic behavior and corresponding simulation
methodscanbefoundinthereviewbySarpkaya(1989)[26].ThesamesubjectisdiscussedinthebooksbySaffman
(1992)[25],Marchioro&Pulvirenti(1994)[12],Bajer&Moffatt(2004)[2]andWuetal.(2006)[35].Furthermore,alot
ofexperimentalandtheoretical/numericalworkshavebeendoneonthesubject,suchas:Winckelmans&Leonard

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

(1993)[33],Ramsey(1996)[22],Riley&Lowson(1998)[23]inconnectionwithdeltawings,Wood&Grace(2000)[34],
Krasny et al. (2002)[11] in connection with the vortex blob numerical simulation method, Dong et al. (2006)[9],
Muijres & Lentink (2007)[14], Parker et al. (20072008)[1617], Borazjani & Sotiropoulos (2008)[3], Wang et al.
(2008)[32]andBohl&Koochesfahani(2009)[4],tomentionsomeofthem.
Correct prediction of the dynamic evolution of the free shear layers is of high importance in many physical
problems.Forexample,inbiomimeticflows,thebirdorfishwingperformslargeoscillatory(flapping)motionsin
comparisontoitsparalleltranslation,producingahighlydeformedshearlayerinthevicinityofthetrailingedge,
whichaffectsthewingflowfield.Similarly,insteadyorunsteadynavalpropelleroperation,thefreeshearlayers
arethemainmechanismsofbladetobladeinteractionandintroductionofthehistoryeffects.Forhigherloadings,
theshearlayersarelimitedinsmallregionsbehindthebiomimeticwingsorthepropellerdisk.Asaresult,astrong
amalgamationofthefreeshearlayersoccursandtheproblemofthemutualinteractionofneighboringshearlayers
issuperimposedtothatoftheirselfaction.
InPolitis(2004)[18],wehaveformulateandsolvetheproblemofflowaroundapropellerperformingageneral
unsteady motion. A potential based BEM has been used to represent the propeller blades while a time stepping
algorithm has been used to simulate free shear layer dynamics. During the period 20022008, this formulation,
initiallydevelopedtotreatpropellerproblems,hasbeenexpandedtoincludethesimulationofany3Dunsteady
incompressible nonviscous flow problem around an arbitrary system of interacting nonlifting/lifting,
rigid/flexiblestreamlinedbodies.Byflexiblewemeanpseudoflexible;thatisbodydeformationisdeterminedby
theuserandnotthroughcouplingwiththeelasticbehavior.Thedevelopedcodehasalsobeenenrichedwithtwo
new nonlinear Kutta conditions of the pressure type, in addition to the initially used Morino type Kutta. The
corresponding formulation is discussed in Politis (2009, 2011) [1920] where comparisons with experimentally
measuredforcesforbiomimeticsystemsandpropellersarealsopresented.
In this paper, we concentrate on the methodology used for the numerical determination of the free shear layer
dynamicsandmorespecifically,tothefilteringtechnique,whichsuppressesthesmallerscaleinstabilitiesleaving
the larger scale dominant vortices to specify the dynamic evolution of the free shear layer surfaces. The method
workssuccessfullyforanyflowcase,includingthatoflargeunsteadymotionsofinteractingliftingrigid/flexible
bodies.Afterapresentationofthemethod,wegiveanumberofapplicationsasfollows:(a)aburststartingsteadily
advancing3Dwing,(b)abiomimetic3DwingoperatingattwodifferentStrouhalnumbers,(c)aburststarting
steadily advancing and rotating naval propeller operating at two different advance coefficients, and (d) a novel
FOD(FlexibleOscillatingDuct)propulsionsystem.Thelastapplicationshowshowourmethodcanbeusedasa
tool in analyzing and/or design of novel hydrodynamic systems based on the new Electroactive Polymer (EAP)
materials,whichcanactivelydeforminafluidenvironment,simulatingtheactionofmuscles,Cohened.(2004)[5].
Inallcases,weshowtheeffectofthefilteringparameterstotheobtainedshearlayer3Dpatterns.Itisshownthat
for a range of the filtering parameters the free shear layers are organized in concrete 3D space patterns,
characteristic to the problem considered. We thus arrive at the notion of the attracting wake configuration of a
problem.Furthermore,calculatedunsteadyforcesandmomentsarealmostindependentofthefilteringparameters,
even in chaotic shear layer pictures, as far as the shear layer geometry lies within the attracting configuration.
Finally,itisshownthatincertainflowproblems,theoldassumptionofafrozenwakemodelresultsinunsteady
forcesandmomentscomparabletothatusingthemoreelaboratefreewakemodel.
Formulation
We start with a brief introduction of the main theoretical & numerical aspects of the BEM, necessary for the
descriptionofthefilteringtechnique.AmoredetaileddiscussioncanbefoundinPolitis(2011)[20].
GeometricConsiderations
We build the geometry of complex systems of bodies using surface patches. Each patch consists of a number of
bilinearquadrilateralelements.Twotypesofpatchesareallowed:(i)liftingpatchesand(ii)nonliftingpatches.By
combiningpatcheswebuildliftingand/ornonliftingbodies.Fortheformercase,theuserhastodeterminetheline

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

offlowseparationinthesurfaceofeachliftingpatch.Thetotalsystemsurfaceattimetisdenotedby SB (t ) .Total
systemKuttastripsurfaceattimetisdenotedby SK (t ) .Totalsystemfreeshearlayersurfaceattimet,excluding
Kuttastrips,isdenotedby SF (t ) ,Politis(2011)[20].
VelocityandPotentialRepresentationTheorems
We use an inertial (build in earth surface) frame of reference for the definition of velocities (body or fluid). A
correspondingcoordinatesystem(assumedCartesianorthogonal)isdenotedbyOXYZ.
Asaresultofthe(known)unsteadymotionofoursystemofbodies,intheregionouterto SB(t ) SK (t ) SF (t ) ,a
velocity (perturbation) potential exists which, at each time step, is expressible through its traces , on the
boundarypoints Q SB (t ) SK (t ) SF (t ) .
Introduce:

nr
r 3 dS (1)
SF ( t )

r
r
1
r
1
r
1
1
1
dl r

H ( P)
( n ) 3 dS
(n ) 3 dS
3 dS
3 dS
3 (2)
4 SB( t )
4 SB(t )
4 SK(t )
4 SF( t )
4 L( t )
r
r
r
r
r

where: P is the evaluation point (or control point) for either F or H , n is a unit normal vector at the boundary
1
F ( P)
4

n
1
dS

r
4
SB ( t )

nr
1
3 dS

r
4
SB ( t )

nr
1
3 dS

r
4
SK ( t )

integration point Q SB (t ) SK (t ) SF (t ) showing inside the flow region, r QP , r QP , is the dipole

intensitywithsupporton SK (t ) SF (t ) and thecorresponding(to )surfacevorticityintensitygivenby:

(3)

n (4)
Finally, L(t ) (integrationregionofthelastlineintegralintherighthandsideofequation(2))isthefreepartofthe
lineboundingthefreeshearlayers,definedby: L(t ) ( SK (t ) SF (t )) ( SK (t ) SB (t )) .
Withtheaidofrelations(1),(2)representationtheoremsfor , become:

( P) F ( P)

P outer to ( SB (t ) SK (t ) SF (t )) (5)
( P ) H ( P )

2 ( P) F ( P)

P SB(t )

1
1
1

( P) (n ) n (n ) n H ( P)
( P ) H ( P )
(6)
2
2

1
2

( P) ( P) F ( P)

,
P SF (t )

, ( P ) ( P ) n ( P ) H ( P )
(7)

1
2

, ( P) ( P) F ( P)

andsimilarlyfor SK , .Inrelation(7),thesuperscripts ( , ) denotethetwosidesofthefreeshearlayersurfaces,

whiletheunitnormal n isdirectedfrom()to(+).Noticethatsomeofthesurfaceintegralsfor F or H inrelations

(6) or (7) contain strong surface singularities of the Cauchy type. Thus, their meaning is realizable only in the
principalvaluesense,Mikhlin(1965)[13].
TheIntegralEquation

Let vA denotesthe(known)velocityoftheboundarypoint A SB (t ) and n aunitvectornormaltobodysurfaceat

A withdirectionpointingintotheflowregion.Then,thenoentranceconditionat A hastheform:

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

n vA n

(8)

Substituting(8)tothefirstofequations(6)andusing(1)weget:



n vA
1
1
1
1
1
n r
n r
( P)
3 dS
3 dS
dS

2
4 SB ( t ) r
4 SK ( t ) r
4 SB ( t ) r
4

nr
r 3 dS , P SB(t ) (9)
SF ( t )

ThisisasecondkindFredholmtypeCauchysingularboundaryintegralequationforthedeterminationof and

onpointsof SB (t ) and SK (t ) respectively.Intherighthandsideof(9),thefirsttermisaknownintegral(asfar

as the motion of the system of bodies is known). The second term is an integral over the free boundary SF (t )
knownfromthesolutionoftheproblematprevioustimesteps.Theunknownsinthelefthandsideof(9)arethe
potential on SB (t ) and the dipole intensity on SK (t ) . For their determination, the additional required
conditionistheKuttaconditionattheseparationlines(trailingedgeincaseofawingfloww/oseparation).
TheKuttaCondition
Letthepoint A SB (t ) .Let

d
denotethetimederivativeforanobserverbuilttothepointAofthemovingbody
dt A

andlet vA denotetheknownvelocityof A .ThenunsteadyBernoulliequationtakestheform:

p p

d
dt

1
2 1
vA vA2 (10)
2
2

According to a pressure type Kutta condition, as we approach the trailing edge point from either pressure side
(superscript +) or suction side (superscript ), the pressure should be continuous, i.e.: p p . Using(10), this
becomes a quadratic (nonlinear) relation between , , , . Assuming steady linearized flow, Bernoulli
equationdegeneratestothefamousMorinocondition: approaching L(t) ( )approaching L(t) whichisalinearequationin
from shear layer

from body points

, .EitherformofKuttaconditionscanbehandledbyourcode.
Shear layerdynamics. Kinematic and dynamic conditions on a free vortex sheet expressed in terms of the dipole
intensityofthesheetresultinthefollowingequation,Politis(2004)[18],Wuet.al.(2006)[35]:

D
0 (11)
Dt

where D / D t denotes a material derivative for based on the mean perturbation velocity v of the shear
layer:

orindevelopedform:

v
H ( P) (12)
2

dl r
r 3 (13)
L(t )

Equation (11) informs us that the dipole surface SF (t ) , supporting the dipoles with intensity ( , ) , is travelling
1

v ( P )
4

r
1

(
n
)
dS

r
SB ( t )

r
1

(
n
)
dS

r
SB ( t )

r
1
dS

r
SK ( t )

r
1
dS

r
SF ( t )

withvelocity v ,where , denotesasetofcurvilinearsurfacecoordinatesforthepointson SF (t ) .Thus,ifa

surfaceexistsattime t ,wecanfinditsnewposition,attime t dt ,bydeformingitby v dt .

CalculationofForces,MomentsandPower

Let F (t ) denotethetotalforceexertedbythefluidtothebodyat A SB (t ) .Thisforceisasumofpressureforces

normal to the body (denoted by P(t ) ) and viscous forces tangential to body (denoted by D(t ) ). Thus:

F (t ) P(t ) D(t ) .Pressureforcescanbecalculatedfromequation(10)whiledragforcescanbefoundassuminga

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

localat A viscousdragcoefficient.Assumingareferencepointformoments rM (t ) ,thetotalinstantaneoussystem

force/momentcanbecalculatedbytheformulas:

Fsystem (t ) F (t ) dS , M system (t )
SB ( t )

rM (t ) F (t ) dS (14)

SB ( t )

Thelastformulascanbeappliedtothewholesystem(use SB (t ) asintegrationregion),ortooneormorebodiesof
ourconfiguration(useasubsetof SB (t ) asintegrationregion).Inpropulsionproblems,thereisalwaysapreferable

instantaneous direction in which the system moves. Let V (t ) the instantaneous velocity of the system along this

direction.Thentheinstantaneoususefulpowerofthesystemisgivenby: EHP (t ) V (t ) Fsystem (t ) ( EHP : Effective

Horse Power). The net instantaneous power interchange of our system with the environment is given by:

F (t ) vA (t ) dS .Thenthepowerprovidedtothesystem,termedDHP(DeliveredHorsePower),isgivenby:
SB ( t )

DHP (t )

F (t ) v A (t ) dS EHP (t ) (15)

SB ( t )

Theratio EHP(t ) / DHP(t ) definestheinstantaneousefficiencyofoursystem.Noticethatthepreviousformulascan

beappliedequallywelltoeitherrigidorflexiblebodies.
DiscretizationandSolution
Subdivide SB (t ) into N B elements.Subdivide SK (t ) into N K elements.Subdivide SF (t ) into N F elements.Fournode
quadrilateralelementshavebeenusedforthesubdivisionofbodyandshearlayerboundaries.Assumepiecewise

constant and n for all elements on SB (t ) . Assume piecewise constant for all elements on SK (t ) SF (t ) .
Denotetheseconstantvaluesby i , i ( n i ), i wheretherangeofindex(i)isadaptedaccordingly.Withthe
aidofthepreviousassumptions/notationintegral,equation(9)becomes:
1
i Bi , j j Bi , j j
2
j 1, N B
j 1, N K

1
Ai , j
4

Ej

j 1, N B

dS
1
, Bi , j
QPi
4

Ej

Bi , j j (16)

n (Q) QPdS
i

Ai , j j

QPi

j 1, N F

(17)

and E j denotes the surface of the j th element from either SB(t ), SK (t ), SF (t ) and Pi denotes the i th control point
(centroid of Ei ) on SB (t ) . Relation (16) applied at the N B centroids of the body elements, gives N B linear
equations for the determination of the j , j . The N K additional equations required for the calculation of j are
takenfromthesatisfactionoftheKuttaconditionon SK (t ) SB (t ) (i.e.trailingedges).Inourcodeweimplement
threealternativeformsoftheKuttacondition:(i)ThefirstalternativeconsistsofalinearMorinocondition,which
in discretized form becomes: i i i ( i and i denote element numbers on body, neighboring to trailing
edge from different sides and i denotes element number on kutta strip, neighboring to the same point of the
trailing edge), (ii) The second alternative consists of a nonlinear pressure type Kutta in the form of equation
p p with the additional conditions that the downstream length of the KuttaStrip element is calculated
according to the total velocity there and the kuttastrip element is either tangential to blade face or blade back
accordingtothelocalvalueofcirculationatthisspanwiseposition,(iii)Thethirdalternativeconsistsofamixed
type Kutta i.e. partly Morino and partly Pressure type. We have found that at higher loadings, a pressure type
Kuttaatbladetipsistoostrongandoccasionallyleadstoadestructionoftheshearlayergeometryatthoseregions.
Thus,wehavedecidedtointroduceathirdalternative,themixedKutta,i.e.aMorinoconditionatthetipsofthe
separation line and a pressure type Kutta at all other points. If alternatives (ii) or (iii) have been selected, the
resultingsystemofequationsisnonlinearanditissolvedbyusingaNewtoniterationmethod,withstartingvalue
takenfromaMorinotypeKutta(firstalternative).

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

Withthepreviousdiscretizationschemeinmind,thediscretizedformulaforthecalculationof v atpointson
SF (t ) ,neededfortheapplicationoftheshearlayerdeformationscheme,becomes:

vi Ci , j j Di , j j Di , j j Di , j j (18)
j 1, N B

j 1, N B

j 1, N k

j 1, N F

where:

1
Ci , j
4

Ej

QPi dS
QPi

1
, Di , j
4

l (Q ) QPi dl
QPi

E j

( l aunittangentvectorto E j )(19)

For the derivation of formula for Di, j the equivalency of a constant dipole element with a line vortex at its

boundary has been used, Politis (2004) [18]. Furthermore, for a bilinear boundary element E j , its boundary E j
consistsoffourstraightlines.Thus,thecorrespondinglineintegralinequation(19)canbecalculatedanalytically.
In detail: let E j AB BC CD DA ( A, B, C , D denote the nodes of element E j ). Then:

Di , j Di , j Di , j Di , j Di , j . The contribution Di , j to Di, j from the side AB (and similarly from sides:
AB

BC

CD

DE

AB

BC , CD , DA ),tothecontrolpoint P ,isgivenbytherelation:

PA PB
(cos cos ) (20)
AB
4 h
PA PB

where AB , AP , BA, BP and h denotesthenormaldistancefromcontrolpoint P totheline AB .

Di , j

Solutionofourproblemisimplementedbyatimesteppingalgorithmasfollows:
Ateachtimestep:
a.

Findthenextgeometricpositionofthesystemofbodies.

b. Generate corresponding Kutta strips, for the case of lifting bodies, introducing thus the extra unknowns
requiredfortheKuttaconditionsatisfaction.
c.

Solvethesystemconsistingofthenoentrance,equation(16),andKuttaconditions.Incaseofpressure
typeKutta,aNewtoniterationisusedatthisstep.

d. Deform the free shear layers to their new positions, by applying a special filtering technique (to be

discussedlater)tocalculate v ,equation(18).
e.

Output results (pressures, forces, velocities, position of free shear layers) at this time step, in a form of
TECPLOTcompatiblefiles.

f.

Proceedtothenexttimestepandrepeatthecalculationstartingfromstep(a).

The Filtering Technique

Fromthephenomenologicalpointofview,thephysicalmechanismsofeithertheselfactionofafreeshearlayeror
the mutual interaction between neighboring shear layers, are mainly viscous and as such are excluded from the
modeling capabilities of a rigorous BEM formulation based on the representation theorems for potential and
velocity.
Morespecifically,determinationofeithertheshearlayerselfmotionoritsinteractionwiththeneighboringshear

layersrequiresevaluationofthevelocity v ,equation(18),atcontrolpoints P SF (t ) .Inourformulationweuse

ascontrol points, the grid nodes of the movinggrid representing the freeshear layerat time t . Considering the

terms in the right hand side of formula(18), we observe that only the last two contribute to v in a singular
manner. This singular manner is explicitlyexpressed byrelation (20). A heuristic proposalfor the elimination of
thesingularityisofferedbyeithertheLambOseenortheBurgersvortices,whicharespecialexactsolutionsof

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

theNavierStokesequations.Inbothcases,theazimuthalvelocity u ofavortexfilamentisgivenbytheequation,
Wuetal(2006)[35]:
r

u
(1 e R ), p 2 (21)
2 r

where: denotestheintensityofthevortexfilament, r theradialdistanceofthecontrolpointfromthecoreofthe

filament, c is a constant characterizing the range of action of the term (1 e

r
c
R

) and R 2 t is a time

dependent characteristic length used to make the action distance r nondimensional ( denotes the kinematic
viscosityofthefluidand t theageofthevortex).
Comparing formula (21) with formula (20) applied to the case of an infinite singular vortex line
( 00 , 00 cos cos 2 ),weobservethattheyaresimilarexceptforthemultiplicativefactor:

m(c, p, r / R) (1 e

r
c
R

), p 2 (22)

whichmollifiesthesingularityat r 0 ,substitutingitbyaviscousvortexcoreoffinitediameter,controlledby
p

r
the parameters: c, p, R . Rearranging the terms in (22) and taking the logarithms we get: c ln(1 m) / .
R
Assuming a characteristic intensity m 0.5 at r / R 1 , the mollifier constant c can readily been calculated as:
c ln(0.5) 0.69314718 independentof p .Inthiscase,themollifierfunctionisuniquelydefinedbyitsradius R

anditspower p ( p 2 forvortexfilaments).
Formula(21),for 2 andpreselected c 0.69314718, p 2, R 1 ,isshowninfigure1.Formula(22)isshownin
figure2forthreedifferentvaluesoftheparameter p 2, 3, 4 andwith c 0.69314718 .
Notice that mollifier functions have already been introduced in vortex particle flow simulation methods,
Winckelmans&Leonard(1993)[33],Krasnyet.al.(2002)[11].

0.60

0.50

m(p=2)/r

0.40

0.30

0.20

0.10

0.00
0.00

0.50

1.00

1.50

2.00

2.50

3.00

FIGURE1.AZIMUTHALVELOCITYOFALAMBOSEENVORTEX. 2 , c 0.69314718, p 2, R 1

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

1.20

1.00

0.80
m(p=2)
m(p=3)
m(p=4)

0.60

0.40

0.20

0.00
0.00

0.50

1.00

1.50

2.00

2.50

3.00

r/R

FIGURE2.MOLLIFIERFUNCTION, c 0.69314718, p 2,3, 4

Withtheaidofthemollifierfunction(22),formula(20)becomes:

Di , j

where Di , j

AB

(1 e

h
c
R

PA PB
)
(cos cos ) , p 2 (23)
4 h
PA PB

denotesthemollifiedequivalentto Di, j .Consideringformula(18)weobservethatthe Di, j ismetat

AB

thesecond,thirdandfourthtermsinitsrighthandside.Sincethecontrolpointfor vi belongsto SF (t ) ,onlythe

thirdandfourthtermsresultinsingularbehaviorandthushavetobemollified(i.e.use Di , j insteadof Di, j ).This

AB

observationiscorrectforallcontrolpointson SF (t ) withtheexceptionoftheregionof SF (t ) inthevicinityofthe

Kuttastrip(i.e.adjacenttothetrailingedges).Inthisregion,theinducedvelocityisstronglyaffectedbyboththe
freeshearlayerandthebodyelements.Byapplyingafilteringprocesssolelytothefreeshearlayerdipoleelements,
an asymmetry in the induced velocities in the vicinity of the Kutta strip can occur, which can trigger the
breakdownoftheshearlayergeometryatthosepoints.Thisdangerismorepronouncedatthefirstfewtimesteps
ofaburststartingsimulation,whereastrongstartingvortexiscreatedinthevicinityofthetrailingedges.Toavoid
the numerical instability, the second term in the right hand side of equation (18) (i.e. the induced velocity
contribution from body potential) has to be mollified as well. Furthermore, for the same symmetry reasons, the
mollifier function has to be smooth (or at least continues) in its extended support: SB(t ) SK (t ) SF (t ) . In our
formulation, we satisfy this condition by taking the body mollifier radius R to be constant for all body elements
withavalueequaltotheshearlayermollifierradius R atthetrailingedgeregion.
Itisawellknownfactthatviscousdiffusionincreasesthediameterofthevortexcoreofaviscousvortexfilament.
This is the physical content of the relation R 2 t introduced previously. Thus, older vortices have larger
viscous diameters (or action ranges). According to the time stepping algorithm, a new spanwise strip of ring
vorticesisaddedtoeachshearlayerateachtimestep.Thus,differentspanwisevortexstripshavedifferentages.
Inourformulation,wetakeheuristicallythiseffectintoconsideration,byintroducingagedependentmollifierradii
R .Thisplaysacrucialroleindelayingtheonsetofchaosintheolderpartsofthefreeshearlayers,especiallyin
cases with higher hydrodynamic loadings. A measure of the age of a spanwise strip can be provided by the
number of strips inbetween this strip and the trailing edge. From the experience gained from the practical
applicationsofthemethod,weknowthatpropervaluesfor R are:0.5%to15%ofthediameteroftheconvexhull

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

(denoted by DCH ), containing our system of bodies, depending from the hydrodynamic loading, the body
geometricdetailsandthemotiondetailsofoursystem.
Inthecaseofaclosecontactofthefreelymovingshearlayerwithabodysurface(forexampleinakaplantype
ductedpropeller,thebladeshearlayerspassveryclosetotheductsurface)numericalinstabilitiesmayoccurdue
to the zeroing of the denominator in the respective formulas for the calculation of velocities and/or potentials,
equations(18)and(16),whichcanleadtoanintrusionoftheshearlayerinsidethebody.Toavoidsuchaninstance,
the developed code contains an algorithm by which user controlled elastic shields are applied to the bodies,
avoiding thus the intrusion of a shear layer node inside the body. This shield methodology is planned to be
discussedinafuturepaper.
Applications
TheShearLayerGeometryofaBurst(Step)StartingWing
Thefirstapplicationconcernstheflowaroundaburststartingwing.Thewinghasaspan s 2 m ,achordatmid
span c 1m andachordattip ctip 0.3m .ThewinghasanXskewbackof30degrees,measuredatthe c / 3 span
wise line from the leading edge.A NACA 0012 section has been used. Calculations have been performed for an
advance velocity V 5m / s at an angle of attack a 200 , using a blade grid with 40 spanwise elements and 16
chordwiseelementsforeitherfaceorback.Weuseanequallyspacedspanwiseandacosinespacedchordwise
subdivision.Total simulation time is 1.0s subdividedin 84 time steps,with each time step equal to t 0.0119 s .
The previous selections result in an Xextension of the free shear layer, approximately equal to 5 chord lengths.
Furthermore, DCH 2m for this example. A mixed type Kutta condition has been used in this simulation and all
othersimulationsexceptthelast(FOD),inwhichapressuretypeKuttahasbeenused.
Simulations have been performed for the following wake models (WM): (a) a free WM with mollifier radius
R / DCH 0.005 ,(b)afreeWMwithmollifierradius R / DCH 0.07 ,(c)afreeWMusingatimedependentmollifier
radius R(t ) , and (d) a frozen WM. In case (c) the mollifier radius varies linearly between time steps, with the
followingbreakpoints( its denotestheorderoftheshearlayerspanwisestripwithregardtothetrailingedge):
(its, R(t ) / DCH ) (1,0.005),(21,0.03),(84,0.1) .Incase(d),thesupportofthefreeshearlayerdipolescoincideswiththe
traceintimeofthetrailingedge(i.e.locatedattheundisturbedflowposition).

Figure 3 shows the free shear layer geometry for WM (a) colored with the dipole intensity . Since n ,

equation(4),theconstant lines,showninthefigure(andallsimilarfigurestofollow),aretangenttothesurface

vorticityintensityvector .Thus,figure3representsacharacteristicquadrilateralringvortexwiththreefreesides,

consistingofthetwotipvorticesandthestartingvortex,andaboundsidecoincidingwiththewingbody.Aswe
movedownstreamfromthewingtrailingedge,thetipvortexrollupgraduallydevelops,whileapproachingthe
junction point with the starting vortex instabilities appear. By increasing the mollifier radius to R / DCH 0.07 ,
WM(b),thetipvortexinstabilityissuppressedbutwithasimilareffecttothetipvortexrollup,figure4.Toallow
asimultaneousdevelopmentofthetipvortexrollupwithasuppressionofthestartingvortexinstability,atime
dependentfiltercanbeused.Correspondingresultsareshowninfigure5.Withthetimedependentfilter,WM(c),
the starting vortex instability has successfully been limited, without affecting the tip vortex rollup at the region
nearthewingtrailingedges.
Comparing the 3D views shown on figures 3, 4 and 5, we conclude that irrespective of the selected filters, a
definiteshearlayer3Dpatternexistsintheformofa3Dringvortex.Abettersenseoftheeffectofthefiltersto
theshearlayerpatterncanbeobtained,ifwecomparetheslicesoftheshearlayerswithaplanenormaltotheX
axisfordifferentmollifierradii.Figure6showssuchacomparisonfortheplanenormaltotheXaxisatthepoint
x 2m (with regard to the global inertia coordinate system). Solid line refers to WM (a), dashed line refers to
WM(b)anddasheddotlinetoWM(c).
Attemptstosimulatetheshearlayergeometryofasteadilyadvancingwing,usingsingularvortexsheets,hasbeen
madeinthepastbyotherresearcherswithresultssimilartothatpresentedhere.WerefertotheworksofRamsey
(1996)[22]andWood&Grace(2000)[34].

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

Figure7presentscomparisonsofthevertical(alongYaxis)force,dividedbyfluiddensity ,obtainedusingeither
ofWM(a),(b),(c)or(d).Thecurvesinfigure7arelabeledasfollows:WM(a):0.005,WM(b):0.07,WM(c): R(t) ,
WM(d): frozen . The results for the vertical force are practically the same for WMs: (a), (b) and (c). Those
predictions differ from the corresponding frozen wake model by about 0.6% . Thus, a frozen or a generalized
wake model can be an attractive alternative to the more elaborate free wake model, for predicting forces in a
steadily advancing wing. This explains the success of the traditional lifting line or lifting surface theories in
predictingthesteadyliftforwingsusingeitherafrozenorageneralizedwakemodel.

FIGURE3.FREESHEARLAYEROFABURSTSTARTINGWINGFORWM(A).WINGCOLOREDWITHPOTENTIALINTENSITY.SHEAR
LAYERCOLOREDWITHDIPOLEINTENSITY.

FIGURE4.FREESHEARLAYEROFABURSTSTARTINGWINGFORWM(B).
TIPVORTEXROLLUPPARTIALLYSUPPRESSED.STARTINGVORTEXINSTABILITYSUPPRESSED.

10

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

FIGURE5.FREESHEARLAYEROFABURSTSTARTINGWINGFORWM(C).

FIGURE6.SLICEOFTHEFREESHEARLAYERAT x 2m .SOLIDLINE:WM(A),DASHEDLINE:WM(B),DASHEDDOTLINE:WM(C)

FIGURE7.VERTICALFORCEOFABURSTSTARTINGWING.PREDICTIONSWITHDIFFERENTWMS

11

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

TheShearLayerGeometryofaBiomimeticWing
Biomimetic wings are the subject of extensive investigations, since they are ideally suited in converting
environmental(atmosphericorseawave)flowenergytousefulthrust,succeedingthuspropulsiveefficiencieswell
overone.AnextensivecollectionofresearchworksintheareacanbefoundinthebookofTayloretal.(Eds.)(2010)
[30].ThesubjectofaerodynamicsofnaturalflyersisdiscussedinarecentbookbyShyyetal.(2008)[27].Reviews
onthesubjectcanbefoundinthepapersofTriantafyllouetal.(2000)[31]andRozhdestvensky&Ryzhov(2003)
[24]. Experimental and numerical simulations of various biomimetic flow problems can be found in: Dong et al.
(2006)[9], Muijres & Lentink (2007)[14], Parker et al. (20072008)[1617], Wang et al. (2008)[32], Borazjani &
Sotiropoulos (2008)[3], Bohl & Koochesfahani (2009)[4], Andro & Jacquin (2009)[1], to mention some of the most
recentworksonthesubject.
Due to the obvious importance of biomimetic flow problems, we have selected it as our next application in a
biomimeticwing.Inthestudyofnaturalflyersandswimmersincruisingcondition,itisfoundthattheStrouhal
number(whichcanbeconsideredasameasureoftheirhydrodynamicloading),definedby:
2 f h0
(24)
V

iswithintherange 0.2 St 0.4 ,Triantafyllouetal.(2000)[31],Rozhdestvensky&Ryzhov(2003)[24],Tayloretal.

(2003)[29]. In relation (24), f denotes the common frequency of heaving or pitching, h0 denotes a characteristic
St

heaveamplitudeand V theadvancevelocity.Ontheotherhand,manmadebiomimeticwingscanbedesignedto
operate in much higher hydrodynamic loadings (or Strouhal numbers). For those reasons, we have decided to
presentflowsimulationsattwodifferentStrouhalnumbers: St 0.5 and St 1.0 .Inbothsimulations,alldatahave
beenkeptconstantexceptfrequency f whichhasbeendoubledforthesecondcase.
Casewith St 0.5
The selectedwing geometry and BEMdiscretization is thesameas that used in the previous example of a burst
starting wing. The biomimetic wing advances with a velocity V 5.0m / s , performs a heaving oscillation with
amplitude h0 0.5m , a pitching oscillation with amplitude 25.0 deg and a phase angle of 90.0deg (with respect to
heaving) and a common (for both heave and pitch oscillations) frequency f 2.5rps . The pitching axis has been
selectedat33%fromtheleadingedgeatmidspan.Thetotalsimulationtimeis 0.8s ortwoperiods.60timesteps
havebeenusedperperiodor120timestepstotal.ThepreviousdataresultedinaStrouhalnumber: St 0.5 .
Figure8showsthefreeshearlayergeometryforamollifierradius R / DCH 0.01 (where DCH 2m aspreviously),
coloredwiththesurfacedipoleintensity .Anextensiveinstabilityoftheshearlayerisobserved:(a)atitsedges,
and (b) at its spanwise strips which correspond to a sign change of the sectional angle of attack (defined with
respecttothetotalundisturbedfluidvelocity).Byincreasingthemollifierradiusto R / DCH 0.07 ,theinstabilityis
suppressedinthemajorpartofthefreeshearlayerinfigure9.Fromfigures8and9,weobservethatirrespectiveof
thefilterused,theshearlayersoccupyacertainregionofthe3Dspacewithawellrecognizedpattern.Forabetter
confirmationofthisconclusion,wecomparetheslicesoftheshearlayerswithplanesnormaltoeithertheXorthe
Zaxisinfigures10and11.Deformationoftheshearlayerscanbeusedasavisualizerofthelargescalevortices
createdbythemotionofabiomimeticwing.Usingthisdeformation,weconcludethatthewakeofabiomimetic
wing consists of a train of oblique vortex rings which produce a repeated Sshaped 3D wake pattern (reverse
KarmanVortexStreet).Wecanarriveatthesameconclusionifweusetheconstantdipoleintensitylines(usedto
color the shear layers) in figures 8 and 9, to visualize the surface vortex lines. This Sshaped attracting
configurationinfigure10ischaracteristicforallthrustproducingbiomimeticwingsandhasbeenexperimentally
observedbymanyauthors,Triantafyllouetal.(2000)[31],Muijres&Lentink(2007)[14],Parkeretal.(20072008)[16
17],Zhang&Zheng(2009)[36].
Figures12and13presentcomparisonsoftheunsteadybiomimeticwingforcesobtainedusingafreeWMwiththat
obtained with a frozen WM. For the free WM, the following mollifier radii have been used: R / DCH 0.01 ,

R / DCH 0.07 .Curvelabelingfollowsthatintroducedintheexampleofaburststartingwing.Itisshownthatthe

12

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

free WM results in practically the same X and Y forces, irrespective of the selected R / DCH . Those forces differ
from that obtained using a frozen WM only slightly. For example, at t 0.62 s , which is near the fx / , fy /
maximumvalues,thedifferencebetweenthefrozenwakeandthefreewakethrustforce fx / islessthan 1% .The
correspondingdifferencefor fy / is 2 3% .Noticefinallythataccordingtooursignconventions,thethrustforce
infigure12isnegative.Thisisaresultofthemotionofthewingwhichhasbeenselectedinthenegativedirection
(with respect to the inertia coordinate system XYZ and to the fact that we have define positive forces as that
exertedtothebodybythesurroundingfluid).

FIGURE8.FREESHEARLAYEROFABIOMIMETICWINGAT t 0.8 s WITHEXTENSIVEVORTEXINSTABILITIES.WINGCOLORED

WITHPOTENTIALINTENSITY.SHEARLAYERCOLOREDWITHDIPOLEINTENSITY. St 0.5, R / DCH 0.01

FIGURE9.FREESHEARLAYEROFABIOMIMETICWINGAT t 0.8 s WITHINSTABILITIESSUPPRESSEDINTHEMAJORPARTOFTHE

FREESHEARLAYER. St 0.5, R / DCH 0.07

13

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

FIGURE10.SLICEOFTHEFREESHEARLAYERAT
t 0.8s, z 0.0m .REVERSEKARMANVORTEXSTREETWAKE

FIGURE11.SLICEOFTHEFREESHEARLAYERAT
t 0.8s, x 1.8m .HAIRLINE: R / DCH 0.01 ,BOLDLINE:

PATTERN.HAIRLINE: R / DCH 0.01 ,BOLDLINE:

R / DCH 0.07

R / DCH 0.07

FIGURE12.THRUST(XAXIS)FORCEOFTHEBIOMIMETIC
WINGPREDICTIONSWITHDIFFERENTWAKEMODELS.

FIGURE13.VERTICAL(YAXIS)FORCEOFTHEBIOMIMETIC
WINGPREDICTIONSWITHDIFFERENTWAKEMODELS

Casewith St 1.0
This example is similar to the previous one, except frequency, which has been doubled i.e. f 5.0rps . The total
simulationtimeforthiscaseis 0.4s ortwoperiods.60timestepshavebeenusedperperiodor120timestepstotal.
Figure14showsthefreeshearlayergeometryforamollifierradius R / DCH 0.01 .Weobservethattheinstabilities
shownforthelowerStrouhalinfigure8havebeenstrengthened.Bychangingthemollifierradiusto R / DCH 0.07 ,
theinstabilityisagaineliminatedinthemajorpartofthefreeshearlayerinfigure15.Fromfigures14and15,we
(again)observethat,irrespectiveofthefilterused,theshearlayersshowawellrecognizedpattern.Thispatternis
alsoshownintheslicesoftheshearlayerswiththeplanesnormaltoeithertheXortheZaxisinfigures16and17.
Usingthedeformationoftheshearlayerstovisualizethelargescalevortices,wearrive(again)totherepeatedS
shaped wake pattern or reverse Karman Vortex Street. Comparing figure 9 with figure 15, we observe that by
increasingtheStrouhalnumber(andconsequentlythehydrodynamicloading)thetrainoftheobliquevortexrings
hasbeenlimitedtoaregionclosertothewingbody.

14

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

Figures 18 and 19 present comparisons of the unsteady biomimetic wing forces obtained using a free WM with
either R / DCH 0.01 or R / DCH 0.07 ,withthatobtainedwithafrozenWM.ItisshownthatthefreeWMresultsin
practically the same X and Y forces, irrespective of the selected R / DCH . Those forces differ from that obtained
using a frozen WM. For example, at t 0.32 s , which is near the fx / , fy / maximum values, the difference
betweenthefrozenwakeandthefreewakethrustforce fx / is 6 7% .Thecorrespondingdifferencefor fy / is

19 20% . Thus, by increasing the Strouhal number (equivalently the hydrodynamic loading), the differences
between the frozen wake and the free wake become more pronounced. This is an expected consequence of the
shearlayerlimitationinaregionclosertothebodysurfaceastheStrouhalincreases.Sincebirdsandfishesoperate
at: 0.2 St 0.4 ,itseemsthatthesimplerfrozenWMcanbeusedinforcepredictionsinsteadofthemoreelaborate
freeWM.

FIGURE14.FREESHEARLAYEROFABIOMIMETICWINGAT t 0.4 s WITHEXTENSIVEVORTEXINSTABILITIES. St 1.0, R / DCH 0.01

FIGURE15.FREESHEARLAYEROFABIOMIMETICWINGAT t 0.4 s WITHINSTABILITIESSUPPRESSEDINTHEMAJORPARTOF

THEFREESHEARLAYER. St 1.0, R / DCH 0.07

15

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

FIGURE17.SLICEOFTHEFREESHEARLAYERAT
t 0.4 s, x 0.21m HAIRLINE: R / DCH 0.01 ,BOLDLINE:

FIGURE16.SLICEOFTHEFREESHEARLAYERAT
t 0.4s, z 0.0m .REVERSEKARMANVORTEXSTREETWAKE
PATTERN.HAIRLINE: R / DCH 0.01 ,BOLDLINE:

R / DCH 0.07

R / DCH 0.07 .

FIGURE18.THRUST(XAXIS)FORCEOFTHEBIOMIMETIC
WINGPREDICTIONSWITHDIFFERENTWAKEMODELS.

FIGURE19.VERTICAL(YAXIS)FORCEOFTHEBIOMIMETIC
WINGPREDICTIONSWITHDIFFERENTWAKEMODELS.

TheShearLayerGeometryofaBurstStartingNavalPropeller
Ournextapplicationconcernsestimationofthesteadyperformanceofburststartingnavalpropeller.Thepropeller
is 3bladed with a diameter D 5.7 m , a pitch ratio P D 1.1 , a blade area ratio AE / A0 0.55 and its detailed
characteristics are similar to those of a Bseries propeller, Oosterveld & Oossannen (1975)[15], but with slightly
increasedtipskew.Calculationshavebeenperformedusingabladegridof24radialelementsand16chordwise
elementsforeitherfaceorback.PropellerhubisincludedinourBEMmodeling.
Calculations are performed at two propeller advance coefficients: J V 0.2 and J 0.7 . The first concerns
f D

operationataheavilyloadedpoint,whilethesecondatapointwithalighterload,closertopropellerdesignpoint.
Casewith J 0.2
Thepropellermoveswithavelocityofadvance V 2.28m / s androtateswithafrequency f 2rps or 120rpm .Flow
simulationlastsfortwofullpropellerrevolutionswithapropellerangularstepof6degrees.Thisresultsinatotal
simulationtimeof 1.0s and60timestepsperpropellerturnor120timestepstotal.

16

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

In our first simulation, we have selected a mollifier radius R / DCH 0.01 where DCH 5.7m . Unfortunately, the
executionofthecomputercodeforthiscasehasautomaticallybeeninterruptedat t 0.21s ,withtheerrormessage:
ahighlydeformedelementhasbeendetected.Thismessagehastodowithafailureoftheadaptivequadrature
overanelementandwiththelimits(accuracyandmaximumnumberofintegrationpoints)setbytheuserofthe
code.Figure20showsthefreeshearlayergeometryattime t 0.21s ,whichcorrespondstothelastinstancebefore
theautomaticcodeinterruption.Anextensiveinstabilityoftheshearlayergeometryisobservedatthejunctionof
the starting vortex with the hub vortex, which explains the automatic code interruption with the corresponding
error message. By increasing the mollifier radius to R / DCH 0.05 , the instability is suppressed allowing thus the
finishingofthesimulation.Figures21and22showthe3DandYZprojectedviewsofthefreeshearlayerwhen
R / DCH 0.05 . From the transparent views, it is shown that at the rear downstream part of the free wake some
instabilitystilloccurs.Byfurtherincreasingthemollifierradiusto R / DCH 0.1 ,thefreewakeinstabilityisalmost
fullyeliminatedinfigures23and24.Fromfigures21to24,weobservethat,irrespectiveofthefilterused,theshear
layers always occupy a certain region of the 3D space with a wellrecognized pattern similar to the shape of a
mushroom.Figures25and26compareslicesoftheshearlayerswithplanes,whicheithercontaintheXaxisorare
normal to it. Using the deformation of the shear layers as a visualizer of the large scale vortices created by the
propeller,weconcludethatthewakeofapropellerevolvesasaresultofaredistributionofboththestartingand
thetipvortices.Thestartingvorticesaremainlyresponsibleforthecreationofangularmomentuminthepropeller
race, while the tip vortices are connected with the creation of an axial momentum. With the passage of time the
strongtipvortices,interactwiththestartingvortices,increasingtheirdiametersandproducingamushroomtype
downstream wake pattern. This mushroom vorticity pattern strengthens with time, accelerating the flow in the
propellerdiskandcontractingtheflowdownstreamthepropellerdisk.Ournumericalfindingsareexperimentally
confirmedintheworkofStettler(2004)[28].
Figure27presentscomparisonsoftheburststartingpropellerthrustforcesobtainedusingeitherafreeWMora
frozenWMwithexperimentalresults.ForthefreeWM,thefollowingmollifierradiihavebeenused:

R / DCH 0.01, 0.03, 0.05, 0.07, 0.10

Curvelabelingfollowsthatintroducedintheexampleofaburststartingwing.Itisshownthatwiththeexception
ofaninitialtransientperiod,thefreeWMresultsinpracticallythesamethrustforce,irrespectiveoftheselected
R / DCH . This force is very near to the experimental measurements, Oosterveld & Oossannen (1975) [15]. On the
other hand, very large differences exist with the predictions obtained using a frozen WM. At the end of the
simulation,thosedifferencesareoftheorderof30%withatrendtoincreasewiththepassageoftimeinfigure27.

FIGURE20.FREESHEARLAYEROFAPROPELLERAT t 0.21s .EXTENDEDSHEARLAYERINSTABILITIESOCCURATTHEJUNCTION

OFTHEHUBVORTEXWITHTHESTARTINGVORTEX.BLADESCOLOREDWITHPOTENTIALINTENSITY.SHEARLAYERSCOLORED
WITHDIPOLEINTENSITY. J 0.2, R / DCH 0.01 .

17

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

FIGURE21.3DVIEWOFAFREESHEARLAYERAT t 1.0 s .

FIGURE22.YZPROJECTEDVIEWOFAFREESHEARLAYER
AT t 1.0 s . J 0.2, R / DCH 0.05 .

J 0.2, R / DCH 0.05 .

FIGURE23.3DVIEWOFAFREESHEARLAYERAT t 1.0 s .

FIGURE24.YZPROJECTEDVIEWOFAFREESHEARLAYER
AT t 1.0 s . J 0.2, R / DCH 0.1

J 0.2, R / DCH 0.1 .

18

FIGURE25.SLICEOFTHEFREESHEARLAYERWITHA
PLANECONTAININGTHEZAXISAT t 1.0 s .

FIGURE26.SLICEOFTHEFREESHEARLAYERAT:
t 1.0 s, x 0.0m .HAIRLINE: R / DCH 0.05 ,BOLDLINE:

HAIRLINE: R / DCH 0.05 ,BOLDLINE: R / DCH 0.1

R / DCH 0.1

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

FIGURE27.PROPELLERTHRUSTDIVIDEDBYFLUIDDENSITY ASFUNCTIONOFTIME, J 0.2 .PREDICTIONSWITHDIFFERENT

WMSANDCOMPARISONWITHEXPERIMENTS.

Casewith J 0.7
This case is similar to the previous one in all aspects except velocity of advance which has been increased to
V 7.98m / s . Again flow simulation lasts for two full propeller revolutions with a propeller angular step of 6
degrees.Thisresultsinatotalsimulationtimeof 1.0s and60timestepsperpropellerturnor120timestepstotal.
Figure 28 shows the free shear layer geometry for a mollifier radius R / DCH 0.01 . Due to the reduced loading
( J 0.7 ), instabilities in the free shear layer have been limited at the outmost (downstream) parts of it. By
increasingthemollifierradiustoeither R / DCH 0.05 or R / DCH 0.10 ,theinstabilitiesaresuppressedinfigures29
and30respectively.Fromfigures28to30,weobservethatthemushroomwakepatternremains.Figures32and
33compareslicesoftheshearlayerswithplaneswhicheithercontaintheXaxisorarenormaltoit.
Figure31showsthepropellerwithafrozenwakemodel.Comparingfigures2830with31weobservethat,atleast
near the trailing edge, the frozen wake model is a good approximation for the free wake model at this advance
coefficient.
Figure 34 presents comparisons of the thrust forces obtained using either a free WM or a frozen WM with
experimentalresults.ForthefreeWM,thefollowingmollifierradiihavebeenused:

R / DCH 0.01, 0.03, 0.05, 0.07, 0.10

ItisshownthatthefreeWMresultsinpracticallythesamethrustforce,irrespectiveoftheselected R / DCH .This

force is very near to the experimental measurements, Oosterveld & Oossannen (1975) [15]. On the other hand,
predictions obtained using a frozen WM differ from the experimental measurements by 4% (at the end of the
simulation) with a trend to increase with the passage of time in figure 34. This explains why the traditional
generalized wake model used in various propeller theories (lifting line/surface theory and traditional panel
methods)givesgoodpredictionsaroundthepropellerdesignpoint.

19

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

FIGURE28.3DVIEWOFAFREESHEARLAYERAT t 1.0 s .BLADESCOLOREDWITHPOTENTIALINTENSITY.SHEARLAYERS

COLOREDWITHDIPOLEINTENSITY. J 0.7, R / DCH 0.01 .

FIGURE29.3DVIEWOFAFREESHEARLAYERAT t 1.0 s . J 0.7, R / DCH 0.05 .

20

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

FIGURE30.3DVIEWOFAFREESHEARLAYERAT t 1.0 s . J 0.7, R / DCH 0.10 .

FIGURE31.3DVIEWOFAFROZENWAKEMODELAT t 1.0 s , J 0.7 .

21

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

FIGURE32.SLICEOFTHEFREESHEARLAYERWITHAPLANECONTAININGTHEZAXISAT t 1.0 s .
HAIRLINE: R / DCH 0.01 ,BOLDLINE: R / DCH 0.07

FIGURE33.SLICEOFTHEFREESHEARLAYERAT: t 1.0s, x 3.36m .HAIRLINE: R / DCH 0.01 ,BOLDLINE: R / DCH 0.07

FIGURE34.PROPELLERTHRUSTDIVIDEDBYFLUIDDENSITY ASFUNCTIONOFTIME, J 0.7 .PREDICTIONSWITHDIFFERENT

WMSANDCOMPARISONWITHEXPERIMENTS.

22

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

TheShearLayerGeometryofaNovelFlexibleOscillatingDuctPropulsor
Our last application is the FOD (Flexible Oscillating Duct) propulsor, which conceptually has its origins in the
naturalparadigmofajellyfish.JellyfishpropulsionmechanicshasbeeninvestigatedbyDabiri(2005,2006)[89],
whohasmademanynumericalsimulationsandexperimentalinvestigationsandhasshownthefeasibilityofusing
variablediameternozzlesaspropulsors.Inrecentyears,thedevelopmentofEAP(ElectroactivePolymer)materials
hasopenedthewayfortheconstructionofbulkartificialmuscles,mimickingthemotionofaquaticanimals,Cohen
ed.(2004)[5].Acharacteristicpropertyforsuchtypeoffuturepropulsionsystemsistheiractiveflexibilitythrough
an electrically controlled artificial muscle system. The FOD can be considered as a candidate EAP propulsion
system,sinceitcanproducesignificantthrustwithgoodhydrodynamicefficiencies.TheFODhasbeenintroduced
forthefirsttimeinPolitis&Tsarsitalidis(2011)[21],whereitsfeasibilityasashippropulsionmechanismhasbeen
presentedanddiscussed.
TheFODisanaxisymmetricductwhichchangesitsshapewithtimeinaspecificway,determinedbythemotionof
asectionofitinaradialplanecontainingtheFODaxis.ThesectionoftheFODwiththisplanehastheshapeofa2
Dairfoilanditsmotionissimilartothatofabiomimeticfoil.Morespecifically,itperformsaparalleladvancement
alongtheFODaxissuperimposedtoaheavingmotion(normaltotheaxisoftheFOD),apitchingmotions(with
anglemeasuredfromtheaxisoftheFOD)andanoscillatorychangeincamberdistribution,wherealloscillatory
motionshavethesamefrequency,eachwithitsownphase,Politis&Tsarsitalidis(2011)[21].
Inthefollowingexample,theFODhasadiameter D 2 m ,measuredatthepointoffoilsectionpitchaxis,whichis
situatedat33%chordfromtheleadingedge.Thefoilsectionhasachord c 1.0 m .ANACA6412sectionhasbeen
used. The FOD advances with velocity V 4.0m / s . Each FODs section performs a heave oscillation with
amplitude h0 0.5m ,apitchoscillationwithamplitude 25.0 deg andphase 90.0 deg (withrespecttoheave)anda
camberoscillationwithmaximumcamberamplitude m0 m6412 ( m6412 isthemaximumcamberoftheNACA6412
section) and phase 90.0 deg . Oscillation frequency for all motions is 2.0rps . BEM discretization of the FOD has
beensucceededusing72elementsinthecircumferentialdirectionand16inthechordwisedirection(samenumber
for face and back). Simulation lasts for two periods with 63 time steps per period or 126 time steps total. Total
simulationtimeis 1.0s .TheStrouhalnumberfortheFODis St 0.5 .ApressuretypeKuttahasbeenusedinthe
simulation.
Figure35showsthefreeshearlayergeometryforamollifierradius R / DCH 0.005 where DCH 2m .Extensivefree
vortexinstabilitiesareobserved.Byincreasingthemollifierradiusto R / DCH 0.10 ,theinstabilitiesareeliminated
in figure 36. From the constant dipole lines which are used to color the shear layers in figures 35 and 36, we
concludethattheshearlayersarecomposedfromasetofcircularringvorticesofvariableintensity,coaxialwith
theFOD.Furthermore,irrespectiveofthefilterused,thefreeshearlayersoccupyacertainregionofthe3Dspace
withawellrecognizedpattern.Thispatternresemblestheshapeofarowofflyingsaucers(oneperperiod)with
diametersproportionaltotheirage(i.e.olderflyingsaucershavegreaterdiameters).
Figure37presentscomparisonsoftheFODthrustforceobtainedusingeitherafreeWMorafrozenWM.Forthe
freeWM,thefollowingmollifierradiihavebeenused:

R / DCH 0.005, 0.01, 0.05, 0.10

RegardingpredictedthrustforceusingthefreeWM,weobservethatwiththeexceptionofthetimeneighborhoods
of t 0.26 s and t 0.76 s ,theresultsarealmostindependentofthemollifierradiusasinallpreviousexamples.For
thetimeneighborhoodsaround t 0.26 s and t 0.76 s ,whichcoincidewiththepointswherethrustismaximized
(in absolute value), mollifier independent thrust is obtained only for mollifier radii less than 0.01. Curiously
enough, around the third point t 0.54 s , where thrust is maximized, the calculated thrust is nearly independent
fromthemollifierradiusinthewholerangeofitsvariationinfigure37.Tounderstandthecauseofthisbehavior
weconcentrateatthestructureofthefreewakevorticityandmorespecificallytoitsrelativepositionwiththeFOD,
at points where thrust is maximized. The structure of the free wake vorticity at t 0.54 s and t 0.76 s for
R / DCH 0.10 is shown in figures 38 and 39 respectively. From figure 38, it is shown that at t 0.54 s the thrust
maximumoccursduringtheFODexpansionphase,whileat t 0.76 s thethrustmaximumoccursduringtheFOD
contraction phase. Furthermore, the corresponding free wake positions, with respect to the FOD, are entirely

23

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

differentforthetwocases;itisafactwhichexplainswhythevalueofthemollifierradiusaffectsdifferentlythe
calculatedthrust.
More specifically, during the expansion phase, the stronger wake ring vortices are relatively far from the FOD
section.Asaresult,thecalculatedthrustisindependentfromthemollifierradiusinthewholerangeofitsvariation.
DuringtheFODcontractionphase,thestrongerwakeringvorticesareclosertotheFODsectiontrailingedge.This
hasapronouncedeffectontheKuttaconditionandconsequentlytothecalculatedFODpressuresandforces.The
effectofthemollifierradiustothepositionofthefreeshearlayerat t 0.76 s isshowninfigure40,wherehairlines
denote the free shear layer geometry for R / DCH 0.005 while bold lines denote free shear layer geometry for

R / DCH 0.10 .Fromthisfigure,weobservethatthemollifierradiusaffectsthepositionoftheshearlayerinthe

vicinityoftheFODtrailingedge,explainingthusthedependenceofthrustfromthemollifierradiiat t 0.76 s as
showninfigure37.
FinallynoticethatinthecaseofaFOD,thrustpredictionsobtainedusingafrozenWMdifferconsiderablyfrom
thatwithafreeWMinfigure37.

FIGURE35.FREESHEARLAYEROFAFODAT t 1.0 s .EXTENSIVESHEARLAYERINSTABILITIESOCCUR.FODCOLOREDWITH

POTENTIALINTENSITY.SHEARLAYERCOLOREDWITHDIPOLEINTENSITY. St 0.5, R / DCH 0.005

FIGURE36.FREESHEARLAYEROFAFODAT t 1.0 s .VORTEXINSTABILITIESHAVEBEENSUPPRESSED. St 0.5, R / DCH 0.10

24

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

FIGURE37.THRUST(XAXIS)FORCEOFTHEFODPREDICTIONSWITHDIFFERENTWAKEMODELS.

FIGURE38.FREESHEARLAYEROFAFODAT t 0.54 s . R / DCH 0.10

FIGURE39.FREESHEARLAYEROFAFODAT t 0.76 s . R / DCH 0.10

25

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

FIGURE40.SLICEOFTHEFREESHEARLAYERDURINGFODCONTRACTION( t 0.76 s, z 0.0m )

HAIRLINE: R / DCH 0.005 ,BOLDLINE: R / DCH 0.10

Conclusions
Thedetailsofafilteringtechniquearepresented,whichincombinationwithaboundaryelementformulationand
a time stepping algorithm can be used to simulate the motion of the free shear layers, produced in cases of
incompressible nonviscous flows around systems of unsteadily moving rigid or flexible lifting bodies. This is a
majorinnovationintreatingcomplexunsteadypropulsionproblemssincenosimplifyingassumptions(likethatof
a frozen or generalized wake) are necessary. The method has been applied to a number of flows around lift
producingconfigurations,performingunsteadymotionswithpracticalinterest.Fromthesystematiccalculations,
thefollowingconclusionscanbedrawn:
Inallcases,thereisacertainnondimensionalrangeforthemollifierradius,almostproblemindependent,where
the free wake vorticity is trapped inside a subset of the 3D space and forms a wellrecognized pattern. For the
simulated cases, this attracting configuration varies considerably from case to case. More specifically: (a) for a
steadilymovingwingitisaquadrilateralringvortexwithrollupatitsfreeedges(tipvorticesandstartingvortex),
(b) in the case of a biomimetic wing it is a repeated Sshaped pattern of oblique ring vortices or equivalently a
reverseKarmanVortexStreet,(c)inthecaseofaburststartingpropelleritisamushroomtype3Dvortexpattern,
while(d)inthecaseofaFODithastheshapeofarowofflyingsaucerswithincreasingdiameters,thenumberof
whichisequaltothenumberofthesimulationperiods.
For smaller values of the mollifier, radius instabilities or even chaotic behavior are almost always present in the
distribution of thefree vorticityinside the attractingconfiguration. By increasing the mollifier radii or usingage
dependentmollifierradii,thoseinstabilitiesgraduallydisappear.Itisofgreatimportancethatchaoticdistributions
ofthewakevorticity,obtainedforthelowermollifierradii,resultinpredictionsforforcesveryclosetoeachother,
asfarasthevorticityremainsinsideitsattractingconfigurationanditisnotsoclosetothebodyboundaries.This
result indicates that under the seemingly chaotic picture in the free shear layer vortices, some type of order still
exists.
From the simulations, itis shown thatin the range of theStrouhal numbers where naturalflyersand swimmers
operate,thefrozenwakemodelcanbeanattractivealternativeforpreliminaryforcepredictions.
REFERENCES

[1]

Andro,J.Y&Jacquin,L.(2009).Frequencyeffectsontheaerodynamicmechanismsofaheavingairfoilinaforwardflight
configuration.AerospaceScienceandTechnology,13:7180.

26

DevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016www.daoejournal.org

[2]

Bajer,K.&Moffatt,H.K.(2004).Tubes,SheetsandSingularitiesinFluidDynamics.KluwerAcademic,NewYork.

[3]

Borazjani, I. & Sotiropoulos, F. (2008). Numerical investigation of the hydrodynamics of carangiform swimming in the
transitionalandinertialflowregimes.J.ofExperimentalBiology,211:15411558.

[4]

Bohl,D.G.&Koochesfahani,M.M.(2009).MTVmeasurementsofthevorticalfieldinthewakeofanairfoiloscillatingat
highreducedfrequency.J.FluidMechanics,620:6388

[5]

Cohen,Y.B.ed.(2004).ElectroactivePolymer(EAP)ActuatorsasArtificialMusclesReality,Potential,andChallenges.
SPIEPress.

[6]

Dabiri, J.O. & Gharib, M. (2005a). Starting flow through nozzles with temporally variable exit diameter. J. of Fluid
Mechanics,538:11136.

[7]

Dabiri,J.O.&Gharib,M.(2005b).Theroleofoptimalvortexformationinbiologicalfluidtransport.Proc.R.Soc.,B272:
155760.

[8]

Dabiri,J.O.,Colin,S.P.&Costello,J.H.(2006).Fastswimminghydromedusaeexploitvelarkinematicstoformanoptimal
vortexwake.J.Exp.Biol.,209:202533

[9]

Dong, H., Mittal, R. & Najjar, F.M. (2006). Wake topology and hydrodynamic performance of lowaspectratio flapping
foils.J.ofFluidMechanics,566:30943

[10] Katz,J.&Plotkin,A.(2001).LowspeedaerodynamicsFromwingtheorytopanelmethods.Cambridgeaerospaceseries.
CambridgeUniversityPress.
[11] Krasny,R.,Lindsay,K.&Nitsche,M.(2002).Simulationofvortexsheetrollup:chaos,azimuthalwaves,ringmerger.In:
Bajer,K.&Moffatt,H.K.eds.Tubes,SheetsandSingularitiesinFluidDynamics.KluwerAcademic,NewYork.
[12] Marchioro,C.&Pulvirenti,M.(1994).MathematicalTheoryofIncompressibleNonviscousFluids.SpringerVerlag.
[13] Mikhlin,S.G.(1965),Multidimensionalsingularintegralsandintegralequations.Pergamonpress.
[14] Muijres,F.T.&Lentink,D.(2007).Wakevisualizationofaheavingandpitchingfoilinasoapfilm.ExpFluids43:665673.
[15] Oosterveld, M.W.C. & Oossannen, P. (1975). Further Computer Analyzed Data of the Wagenigen BScrew Series.
InternationalShipbuildingProgress,vol.22.
[16] Parker,K.,vonEllenrieder,K.D.&Soria,J.(2007).Morphologyoftheforcedoscillatoryflowpastafinitespanwingat
lowReynoldsnumber.J.ofFluidMechanics,571:327357.
[17] Parker,K.,vonEllenrieder,K.D.&Soria,J.(2008).Fluidmechanicsofflappingwings.ExperimentalThermalandFluid
Science32:15781589.
[18] Politis G.K. (2004). Simulation of Unsteady motion of a Propeller in a fluid including Free Wake Modeling. Journal of
EngineeringAnalysiswithBoundaryElements,28(6):633653.
[19] PolitisG.K.(2009).ABEMcodeforthecalculationofflowaroundsystemsofindependentlymovingbodiesincludingfree
shearlayerdynamics.AdvancesinBoundaryElementTechniquesX.
[20] Politis, G.K. (2011). Application of a BEM time stepping algorithm in understanding complex unsteady propulsion
hydrodynamicphenomena.OceanEngineering,38:699711.
[21] Politis, G.K. & Tsarsitalidis, V. (2011). Exploring the Potential of an Oscillating Duct as a Marine Propulsor. Second
InternationalSymposiumonMarinePropulsors,smp11,Hamburg,Germany.
[22] Ramsey,W.D.(1996).BoundaryIntegralMethodsforliftingBodieswithVortexWakes.PhDthesis,MITDepartmentof
OceanEngineering.
[23] Riley,A.J.&Lowson,M.V.(1998).Developmentofathreedimensionalfreeshearlayer.J.FluidMech.369:4989.
[24] Rozhdestvensky, K.V. & Ryzhov, V.A. (2003). Aerohydrodynamics of flappingwing propulsors. Progress in aerospace
sciences,39:585633.
[25] Saffman,P.G.(1992).VortexDynamics.CambridgeUniversitypress.
[26] Sarpkaya, T. (1989). Computational methods with vortices The 1988 Freeman Scholar Lecture. Journal of Fluids
Engineering,Vol.111/5.

27

www.daoejournal.orgDevelopmentandApplicationsofOceanicEngineering(DAOE)Volume52016

[27] Shyy, W., Lian, Y., Tang, J., Viieru, D. & Liu, H. (2008). Aerodynamics of Low Reynolds Number Flyers. Cambridge
AerospaceSeries,ISBN139780511478710.
[28] Stettler, J.W. (2004). Steady and Unsteady Dynamics of an Azimuthing Podded Propulsor Related to Vehicle
Maneuvering.PhDthesis,MIT.
[29] Taylor,G.K.,Nudds,R.L.,&Thomas,A.L.R.(2003).FlyingandswimminganimalscruiseataStrouhalnumbertuned
forhighpowerefficiency.Nature(London)425,70711.
[30] Taylor, G.K., Triantafyllou, M.S. & Tropea, C. (eds.). (2010). Animal Locomotion. SpringerVerlag Berlin Heidelberg,
ISBN9783642116322.
[31] Triantafyllou,M.S.,Triantafyllou,G.S.,&Yue,D.K.P.(2000).Hydrodynamicsoffishlikeswimming.AnnualReviewof
FluidMechanics32,3353.
[32] Wang, Z., Chen, P. & Zhang, X. (2008). Wake Vortex Structure Characteristics of a Flexible Oscillating Fin. Journal of
BionicEngineering,5:4954.
[33] Winckelmans, G.S. & Leonard, A. (1993). Contributions to Vortex Particle Methods for the Computation of Three
DimensionalIncompressibleUnsteadyFlows.J.ofComputationalPhysics,109:247273.
[34] Wood, T.H. & Grace S.M. (2000). Free Wake Analysis for Calculating the Aeroacoustics of a WingFlap Configuration.
38thAIAAAerospaceSciencesMeeting,paperAIAA20000607.
[35] Wu,J.Z.,Ma,H.Y.&Zhou,M.D.(2006).VorticityandVortexDynamics.SpringerVerlagBerlinHeidelberg.
[36] Zhang,N.&Zheng,Z.C.(2009).Flow/pressurecharacteristicsforflowovertwotandemswimmingfish.Computers&
Fluids,38:10591064.

28