Experimental
study
of
a
helicopter
model
in
shipboard
operations
Neda Taymourtash
a,
b,
∗
, Daniele Zagaglia
b, Alex Zanotti
a, Vincenzo Muscarello
a,
Giuseppe Gibertini
a, Giuseppe Quaranta
aaDepartmentofAerospaceScienceandTechnology,PolitecnicodiMilano,Milan,Italy
bAerospaceSciencesDivision,JamesWattSchoolofEngineering,UniversityofGlasgow,Glasgow,UK
a
r
t
i
c
l
e
i
n
f
o
a
b
s
t
r
a
c
t
Articlehistory:
Received12March2020
Receivedinrevisedform14January2021 Accepted26April2021
Availableonline3May2021 CommunicatedbyVassilisTheofilis
Keywords:
Helicopter
Windtunnelexperiment Shipairwake
Aerodynamics Multibodysimulation
Thepaperpresentstheexperimentalinvestigationoftheaerodynamicinteractionbetweenahelicopter modeland ashipmodel withasimplifiedgeometry.Inthe firstphaseoftheexperiment,aseriesof wind tunneltests werecarriedoutin ordertostudy the flowfeaturesonthe flight deckfor several wind conditions, without the presence of the helicopter. Pressure measurements and Particle Image Velocimetry surveys were performedto assess the effectof wind velocity and directionon the flow fieldinthelandingregionovertheshipdeck.Moreover,theeffectoftheAtmosphericBoundaryLayer wasinvestigated.Inthesecondphaseoftheexperimentalcampaign,ahelicoptermodelwaspositioned inaseriesofpointsrepresentativeofatypicalsternlandingtrajectoryandaverticaldescentabovethe landingspot.Thelandingmaneuverwasperformedinthreedifferentwindconditions,includingno-wind, headwindandwindblowingfromportsideofthedeck.Therotorloadsandmomentsweremeasuredby meansofasix-axisbalanceforalltestpoints.Theuseofdifferentmeasurementtechniquesinthepresent experiments providesacomprehensive databasesuitable forthe study of therotor-ship aerodynamic interaction.Additionally,theexperimentalresultsareused todevelopanidentificationalgorithmtobe incorporatedintothe flightsimulatorenvironment tocapture the effectofshipairwake onthe rotor loadsduringshipboardoperations.
©2021TheAuthors.PublishedbyElsevierMassonSAS.ThisisanopenaccessarticleundertheCC BY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction
Helicoptersare regularly requiredtoperform challenging mis-sions inconfined areasandcloseto obstacles. Search andrescue missionsoverlandandwater,urbantransport,interventionin nat-uraldisasters,suchasflooding,orearthquake,aresomeexamples in which helicopters interact with the surrounding environment. Inthesesituations,performanceandhandlingqualitiesofthe heli-copterarehighlyaffectedbythepresenceoftheobstaclesinclose proximity. Theaerodynamic interactioncould beevenmore com-plicated when the environment is non-stationary. A well-known
example is offshore operations which are among the most
de-manding tasks forpilots. In thiscase, dueto the combinationof moving flight deck, flying close to the ship hangar wall, chang-ing speed anddirection ofthe wind andturbulent ship airwake, the pilot’s workloadis significantly increased which may
endan-*
Correspondingauthorat:DepartmentofAerospaceScienceandTechnology, Po-litecnicodiMilano,Milan,Italy.E-mailaddresses:Neda.Taymourtash@polimi.it(N. Taymourtash),
Daniele.Zagaglia@glasgow.ac.uk(D. Zagaglia),Alex.Zanotti@polimi.it(A. Zanotti),
Vincenzo.Muscarello@polimi.it(V. Muscarello),Giuseppe.Gibertini@polimi.it
(G. Gibertini),Giuseppe.Quaranta@polimi.it(G. Quaranta).
gerflightsafety.Ithasbeenshownthatthemostofthefrequency content of the unsteady airwake is concentrated in the rangeof 0.2-2Hz[1].Thisbandwidthcoversthewidely acceptedrangeof pilotclosed-loopcontrolfrequencies,characterizedbyacross-over frequencylowerthan1.6Hz[2].
Safetyanalysisforsuchdemanding missions needs aseriesof at-seatrialswhichareinherentlyhazardousandextremely expen-sive.Furthermore, each combinationofship-helicopter should be testedforarangeofwindspeedanddirectioninordertofindthe Ship-HelicopterOperational Limitations (SHOL)[3]. Consequently, developmentofthehelicopter-shipDynamicInterface(DI) simula-tionisconsideredasaviablesolutionwhichreducesthecostand hazardsoftime-consumingat-seatestcampaigns[4].Sucha sim-ulationtool could be used to findthe optimaltrajectory forsafe landing and to design and test of new flight control systems. A better understanding of the environmental conditions could lead tothedevelopmentofhigh-fidelitysimulationenvironmentto im-provepilottraining. Allthoseelements willcontributeto the im-provementofsafetyofrotorcraftoperations,whichistheobjective oftheNITROSproject[5].
Regarding the complexity of the flow field generated by the
helicopter-shipinteraction,developmentofanappropriateairwake modelwhich can capture the induced airloads ofthe rotor is of
https://doi.org/10.1016/j.ast.2021.106774
1270-9638/©2021TheAuthors.PublishedbyElsevierMassonSAS.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).
R Rotordiscradius . . . m
|
U|
In-planevelocitymagnitude . . . m/s U∞ Free-streamwindvelocitymagnitude . . . m/s(
u,
v,
w)
Velocitycomponentsinrotorreferenceframe m/s(
u,
v,
w)
Rootmeansquareofthevelocitycomponents m/s vsh Flowdistortionvelocity . . . m/sIGE InGroundEffectcondition
OGE OutofGroundEffectcondition
PIV ParticleImageVelocimetry
RMS RootMeanSquare
SF SmoothFlow
SHOL ShipHelicopterOperationalLimitations
great importance. Various numerical or experimental approaches canbetakenforairwakemodelling whichresultindifferentlevels of simulation fidelity. In both numerical and experimental anal-ysis of the shipboard operation, the effect of ship airwake and
rotor wake coupling is worth considering. The most simplified
approach is uncoupled simulation which means there is no
in-teraction betweenrotorcraft and ship airwake.One-waycoupling approach,which hasbeenextensivelyimplementedinflight sim-ulation environments so far, accounts only for the effect ofship airwakeon therotor inflow. Inthisapproach, theairwake ofthe ship is pre-calculated, using either steadyor unsteady Computa-tional FluidDynamics (CFD) withoutconsidering the presence of thehelicopter.Thismethodrepresentsoneofthefewviableways to perform pilot-in-the-loop real-time simulations. The ship air-wakevelocitiesareincorporatedintotheflightdynamicscodevia look-up tables, assuming superposition of the ship airwake and rotor induced flow. Based on subjective pilot workload ratings, this approach could reasonably capture the increased workload due to influenceof the ship airwake, includingthe effect of un-steadiness [1,6]. However, the superposition method, has shown a low accuracy for cases of close proximity between the heli-copter and structure of the ship. To assess the coupling effects on the helicopter flight characteristics, Crozon et al. [7] studied the ship-helicopter aerodynamic interaction simulating four dif-ferentcases: isolatedship, isolatedrotor, shipborne rotorand su-perpositionofisolatedrotorandship.Bothsteady-state
Reynolds-Averaged Navier-Stokes (RANS) with actuator disk method, and
unsteady RANS using blade-resolving representation of the rotor wereinvestigated.Theresultsshowedsignificantdifferencesin ro-torloadingbetweenisolatedrotorinforwardflightandnear-deck operationswhichhighlighttheimportanceofcouplingeffects.The validity of this conclusion has been further investigated to
de-termine theminimumdistance betweenhelicopteranda ground
obstacle where the interaction can be considered negligible [8], that was quantified in an approximate distance of 5 main rotor radiiawayfromtheobstacle,independentlyofwindspeed.
Consequently, for the interactive aerodynamic environment in shipboard operations, the most representative approach must be
considered the two-way coupling, or fully-coupled simulation
which includesmutualinteraction ofthehelicopterandship air-wake.Inthisapproach,theaerodynamicsolverandflightdynamics simulation of the helicopter should be run simultaneously with
communicationbetweentwo codes.Theflightdynamicscode cal-culatestheloadingoftherotor,aswellastheattitudeandposition of the helicopter, which are passed to the aerodynamic solver. Then, the localvelocity datais computed by aerodynamic solver andfedback totheflightdynamicssimulation[9–11].Depending onthe computational cost ofthe numericalalgorithm tocapture theaerodynamic interaction,thisapproach mightbe usedinreal time flight simulation. Towards this aim, Zhao et al. [12] devel-opedareal-timerotorwakemodelbyimplementingtheresultsof arotorcraft/shiphybridsolver,whichcouplesaviscousVortex Par-ticleMethod(VPM)withanunstructuredCFD solver,toaugment thewidelyusedPeters-Hefinitestatedynamicwakemodel.In an-other effortto supportreal-time pilotedsimulation,a free-vortex
wakemodelwas developedandintegratedwitha USNavyflight
dynamicssimulation[13].Thismodelincludesavortex-based so-lutionstrategytoaccountforinteractionsbetweentherotorwake andanexternalspatialandtime-varyingdisturbancefield.The re-sultsof the coupled free wake aerodynamic andflight dynamics model were compared with flight test data for vertical descent andshipboardoperations.VPMmayrepresentagoodcompromise betweenaccuracyandcomputational cost,howeverfull RANS of-ferthemostcompleteapproachfora fullycoupledsimulation.In [14],Orucetal.presentedtherecentadvancementsofanongoing projecttowardsthisaim.Theyachievedthereal-timeexecutionof aDIsimulation,withafully-coupledNavier-StokesCFDsolverand a helicopterflight dynamic code (GENHEL-PSU)for thecase ofa simplified shedding wake. However, the final goal requires more substantialcomputingimprovements,asindicatedin[14].
A complementary approach to numerical simulation could be
followedtryingtounderstandandquantifytheeffectsofthe cou-pledshiphelicopterwakesthroughscaledexperimentsperformed inthewindtunnel.Thisapproachhastheadditionaladvantageof providingadatabaseofteststhatcouldbeexploitedtobetter vali-datenumericalsimulation.Asamatteroffact,theapproachtaken inthisworkreliesonwindtunnelexperimentstoimprovethe ca-pability of estimating the flow-field and consequently fidelity of flight simulation.This isthe first steptowards development ofa fully-coupled flightdynamicssimulation withwindtunnel inthe loop,werethemeasuresonawindtunnelmodelcouldbecoupled toa flightmechanics numericalmodelofthe helicopterto simu-latethe approachtrajectory andestimate potential problemsthat
dueto ship airwake,cansignificantly decreaserotor thrust up to 15%.Theresultsalsoconfirmedthestronginfluenceofwindspeed anddirectionontheairwake,andconsequentlyontherotorthrust and operational envelope of the helicopter. Another experiment
was conducted usingthe same ship model,to measureunsteady
loadsactingontherotor-lesshelicopterfuselage[16].Areasonable
correlation was found betweenRoot Mean Square (RMS) loading
levelsandpilotworkloadobtainedfromflighttest[16].The addi-tionalinfluenceoftherotordownwashontheunsteadyloadingof the fuselage was also studiedand comparedwith rotor-lesscase [17].
Kääriäetal.proposedadifferenttestsetuptocharacterizethe aerodynamicloadingofa1:54model-scalehelicopterimmersedin the airwakeof a generic frigateship [18]. The experimentswere conducted in a water tunnel using a speciallydesigned Airwake Dynamometer(AirDyn)identifyingspecifictime-averagedand un-steady loading characteristics caused by the severe spatial and temporal velocity gradients in the airwake. The setup has been usedalsotoinvestigatethepotentialbenefitsofaerodynamic mod-ifications to the ship geometry [19]. Various modifications were proposed andmanywere found effectiveinreducing the RMSof forcesandmoments. Inparticular,thepromising designconcepts
were aside-flapandnotchmodification whichbothshowed
con-sistentimprovementsof25-50%inloadsRMS.
Toidentifytherotordownwashandshipairwakecoupling,flow visualizationtechniqueshavebeenutilizedaswell whichprovide informationondevelopmentoftheinteractingflowfieldduringthe landing maneuver.Landman conductedan experiment, exploiting Particle ImageVelocimetry (PIV)withtheshipandrotorin isola-tionandthenthecombinedcasetoinvestigatethesuperimposed velocityfieldandrecirculationregion[20,21].Furthermore,an ex-tensive rotorthrust survey wasconductedhighlightingsignificant variationsclosetothelandingdeck.
Inadditiontoshipboardoperations, therearesome studies re-garding theinteraction of rotorcraft withsimpleshape obstacles, like buildings.Forinstance,Quinliven andLong evaluated the ef-fectsoftheaerodynamicwakefromalargeupstreamobject,such as a building, on a rotorcraft operating in proximity to that ob-ject [22]. In thisexperiment, smoke visualization was used with highintensitylightinordertovisualizetheflowfieldand,induced velocities were measured using the pressure probes distributed spanwisealong each blade.Theresults showedthatthepresence ofthe upwindbuildingtended to decreasetheinduced velocities attheleadingedgeoftherotordisc,whichisconsistentwiththe recirculationregionbetweentherotorandbuilding.Aseriesof ex-perimentswerecarriedoutatPolitecnicodiMilanobyZagagliaet al. to study helicopterinteractions witha groundobstacle.Initial testswere aimedto replicatedifferenthoveringconditionsin ab-senceofwind[23].Then,theperformanceoftherotorinmoderate windy conditionswas assessed forseveral positionswith respect to the obstacle [24]. In particular, the helicopter model, includ-ing a four-bladed rotor andfuselage, was positioned in different pointsrelativetoasimplifiedvolumewithaboxshape.Theforces andmomentsactingontherotorweremeasuredwithasix com-ponentstraingaugebalancenestedinsidethefuselage.PIVsurvey wasusedtoinvestigatethedetailsoftheinteractingflowfield.The
litecnicodiMilano(GVPM).Thankstothedimensionsofthewind tunnel test section a larger geometric scale with respect to the literature was used in the presentexperimental investigation for both the helicopterandship models.The first campaign focused ona shipairwakecharacterization study withoutthepresence of the helicopter. Withthis aim PIV surveyson longitudinal planes andpressuremeasurements wereperformedinthe shipdeck re-gion. This campaign was devoted to assess the influence of the windvelocity,directionandAtmosphericBoundaryLayer(ABL)on topologyoftheflow-fieldoverthedeck,andalsotohavea charac-terizationofthebaselineflowfieldtobeusedtoidentifytheeffect oftheship-helicopterwakeinteractioninthefollowingtest cam-paigns. Then, a second campaign was carried out to simulate a typical landing trajectoryandanalyse the rotor loadsindifferent positionswithrespecttothedeck.Duringthiscampaign,therotor
forces and moments were measured by a six-components strain
gauge balance, the rotor inflow was surveyed by PIV and
pres-surefield was measured on both theflight deck andthe hangar
oftheship.Theresultspresentedwillshowtheeffectoftheship airwakeonthemeanrotor loads.Othereffectswere identifiedin thefrequencyspectrumofloads,however,duetolimitationinthe testinghardwareused–inparticularinthehelicoptermotor ther-malmanagement–thefrequencyresolutionobtainedwastoolow toprovidereliabledatafortherangeofinterest.Consequently,the analysisoftheeffectsofunsteadiness,andtheconsequentincrease ofpilotworkload,hasbeenpostponedtofuturetestcampaigns.
Then,the experimental resultswere used todevelop an iden-tification algorithm aimed to reconstruct the average and linear variationoftheinflowabovetherotordiskbyaddinganairwake modelrepresentativeoftheinteractionsoftherotorwiththeship airwakein thesurrounding environment.This instrumentcan be consideredan effectivetooltoobtain amorerealistic representa-tionoftheshipboardoperationsinthesimulationenvironment.
2. Experimentalsetup
The tests were carried out in the large test chamber of the
GVPMwhich is 13.84 m wide, 3.84 m highand35 m long. The
testrig, asshowninFig.1,consistsofa 4-bladed helicopterand a simplified ship model [25]. The helicopter model was held by ahorizontal strut fixedto a systemoftwo motorised orthogonal slidingguideswhichwasabletochangetherelativepositionofthe helicopterwithrespecttotheshipinbothverticalandlongitudinal directions.Afixedcoordinatesystemwasdefinedtointroducethe testpoints, whichrepresentsthepositionoftherotor hubcenter withrespect to the ship. The X – Z plane ofthis referenceframe was aligned with the longitudinal symmetry plane of the flight deck,withtheoriginonthefloorofthewindtunnelandtheend ofthestern,seeFig.2(a).
2.1. Shipmodel
Theshipmodelisa1:12.5-scalemodelofSimpleFrigateShape 1 which is a highly simplified but representative ship geometry, developed as a part of an international collaboration in which Canada,Australia,NewZealand,UK andUSAevaluatedtheability
Fig. 1. The test rig mounted inside the GVPM large test chamber.
Fig. 2. Sketch of the 1:12.5-scale model of SFS1 equipped with pressure taps.
ofCFDcodes tosimulatecomplexairwakes[26].Thewindtunnel dimensions enabled selection of a geometric scale of 12.5, quite largerthanmodelsusedinsimilarworksintheliterature[15,18]. Thisallowed foran experimentReynolds numbermuchcloser to fullscale values.Theforepartoftheship wasnotreproduced,as the effect ofthis part onthe airwake in the landing region was considered negligible. This choice allowed mounting of both the models and the traversing system on the large turn-table of the test section(13mofdiameter),andthisinturnallowedeasy in-vestigationoftheeffectofwinddirection.Theshipmodelconsists ofa rectangularprism withastep onits rear andanother prism ontopwhichisactinglikeashipsuperstructure,asrepresentedin Fig. 2(a),reportingthe main dimensionsofthe model.The land-ingpointofthehelicoptermodelwasplacedonthecenterofthe turn-table, thus the ship can be rotated toboth sides, while the landing pointremainsfixedwithrespecttotheboundariesofthe test section.The flightdeckandhangarwall wereequippedwith 77and35pressuretapsrespectively. Thepressuremeasurements were performedusing fourlow-range 32-ports pressurescanners
embedded inside the ship model. The declared accuracy of the
pressurescannersledtoanestimateduncertaintyforthepressure coefficientofapproximately0.15.
2.2. Helicoptermodel
The helicopter model was the same used for the
investiga-tion of the rotor-building aerodynamic interaction described in Refs. [23,24]. The model consisted of a fuselage and a rotor whichhasfouruntwistedanduntaperedrectangularblades,with
NACA0012airfoil,achordofc
=
0.032m,andradiusofR=
0.375 m.A rigid hub withouthinges was adoptedthroughout thistest campaignandthe bladeswere rigidlyconnectedto thehub with averystiffconnectorandaconstantpitchangleof10◦.Sincethe swashplatewasnotincluded,therotorcouldnotbetrimmedtoa targetsetofforcesandmoments.However,thankstotherigidhub alongwithrelativelystiffblades,theaeroelasticdeformationswere negligible.Consequently,thevariation ofloads whileapproaching thedeckwereexpectedtobedominatedbytheinteractions with the obstacle and the flowfield generated by it, andso could be takenas an indexto highlight andquantify the mean effects of theinteractionsontherotoraerodynamics.Therotor rotationalspeedof2580RPM wasmaintained inall testsbymeansofabrushless,low-voltage,electricalmotorwithan electriccontroller.AHall-effect sensorgivingtheone per revolu-tionsignalwas usedasfeedbacksignalforRPMcontrol.Therotor rotational speed drift (less than 20 RPM) was considered in the evaluationoftheforce/momentcoefficientsastherotorRPMwere measuredsimultaneouslywiththeforceandmomentsduringthe testruns.Theloadsactingontherotorweremeasuredusinga six-componentsstrain gauge balance nestedinside the fuselage. The nominalaccuracyoftheloadmeasurementswas0.25% oftheload cell full rangeon each axis, whichresults in approximately0.4% and 0.48% of the Out-of-Ground Effect (OGE) thrust and torque values,respectively.Inordertoreduce thebalancethermaldrifts, eachtestpointcorrespondedtoasinglerunwherethemotorwas startedfromrestandthenstoppedagainattheendofthe acqui-sition.Thebalancesignalsacquisition took placeoveraperiodof 5 seconds. Each test point measurement was repeated for three
Fig. 3. Position of the PIV investigation planes.
different runsandthe resultswere averaged. A repeatabilitytest over30measurementsintheOGEcondition( Z
/
R=
4)exhibited astandarddeviationofthemeasuredthrustandtorquecoefficients ofabout0.3% [24].2.3. PIVsetup
The PIV system comprised a Litron NANO-L-200-15 Nd:Yag
double-pulse laser withan output energy of 200 mJ and
wave-lengthof532nm,andtwoImperxICL-B1921MCCDcameraswith
a 12-bit, 1952
×
1112 pixel array in tandem configuration. Eachcamera was equippedwithaNIKKOR 50-mmlens. The laserwas
positioned on a suitable strut downstream ofthe ship model,so that the laser sheet was aligned with the X – Z plane. The laser wasmountedonalongitudinaltraversingsysteminordertocover the whole area ofinvestigation along theshipdeck providing the same lightpower. The cameraslineofsight was aligned perpen-diculartothelasersheet.
The area of investigation of the first test campaign covered the whole shipdeck(more than2 m
×
0.5 m), thus, inorder to achieve a better resolution ofthe image pairs, the measurementarea included 10 multiple adjacent windows of 480mm
×
270mm witha small overlapping among them. In the second
cam-paign, two adjacent windows with the same dimensions of the
previous tests, and with a small overlapping among them were
surveyed tocover the longitudinal plane ofthe whole rotor disk for inflow investigation.In particular, the fore region ofthe disk was prioritised in the surveys in order to obtain more informa-tion ontherotor inflow. Toavoidlaser reflectionovertheblades during thosetests, theclosest lineofthePIVareaofinvestigation totherotorwas48mmabovethediskplane.The datameasured overthislinewereconsideredtoextractthenormalvelocityalong the longitudinal symmetry axis of the rotor. The position of the
PIV measurement area for the two test campaigns is shown in
Fig.3.
The synchronisation of the two laser pulses with the image pairexposurewascontrolledbyasix-channelQuantumComposer QC9618pulsegenerator.APIVpart30particlegeneratorbyPIVTEC withLaskinatomizernozzleswasusedfortheseeding,which con-sisted ofsmalloil droplets withdiametersof 1-2μm.The entire test section was filled with seeding during PIV tests. The image pair analysis was carried out using PIVview 2C software [27]. A single-pass interrogation was used for thecorrelation ofthe im-agepairswithaninterrogationwindowof64
×
64pixelswithan overlap factorof 50%. Thisled to a resolutionof 8mm betweentwoadjacentmeasurementpoints.Theresultsthatwillbeshown inthefollowingsectionsaretheensemble-averageofthe measure-mentsover 800image pairsacquiredforeach window.Forthose imagestheoutlierregionsaremaskedwithablankbox.
The accuracy of the PIV measurement can be estimated
con-sidering a maximum displacement error of 0.1 px when using
sub-pixel interpolation,asfound in [28]. Taking intoaccount the opticalmagnificationandthepulse-separationtime, whichvaried between300and600 μsaccordingtothelocalmaximumvelocity, themaximumin-plane velocity errorisestimatedto be between 0.04and0.08m/s.
2.4.Atmosphericboundarylayer
The GVPM Large test chamber is equipped with several
de-vices allowing the production of different velocity profiles with differentturbulenceintensities.Forthepurposeofthisstudy,two differentfreestreamvelocityprofileswereconsidered.TheSmooth Flow(SF), inwhichtherewere noupstream turbulatorsandsoa uniform free stream velocity profile was generated, and the At-mosphericBoundaryLayer(ABL),whereturbulatorswereinserted upstreamoftheshiptoobtaina velocityprofilecorrespondingto
thewell-knownpowerlawmodel:
U Uref
=
Z Zref α (1)whereU isthemeanvelocityatheight Z ,Uref isthereference
ve-locitymeasuredatheight Zref and
α
isanexponentthatdependsontheroughnessoftheterrain.TheABLvelocityprofilegenerated inthis experimentcorresponds to the wind profilewith
α
=
0.1 whichissuggestedbySimiuetal.[29] forcoastalareas.Thewind tunnelfreestreamvelocityforeachtestrun wassetusingaPitot probeplacedabovetheshipsuperstructure.Thischoicewasdriven bythefactthatanemometersareusuallyplacedonship’smastto belessaffectedbytheshipairwake.Inthisexperiment,theprobewas positioned 180mm above and90 mm upstream of the top
superstructure,asshowninFig.4.Forthetestswithred-wind,i.e. windcoming fromtheport side oftheship, thePitotprobe was adjustedin orderto have thestaticport in thesame positionof thehead-windtest.
Fig.4showstwoadoptedboundarylayerprofiles withrespect totheshipsize.ComparisonoftheABLwiththeprofilecalculated
by power law model shows a good agreement up to the
Fig. 4. Comparisonofthemeanvelocityprofile(topleft)andturbulenceintensity(topright)forABLandSF.Shipdimensions(bottom)forreference.(Forinterpretationof thecoloursinthefigure(s),thereaderisreferredtothewebversionofthisarticle.)
showsthatABLwillsignificantlyincrease theturbulenceintensity compared with SF. For instance,at height of the flight deck the turbulenceintensityisincreasedfrom2% inSFto10% inABL.For further informationaboutthe features oftheflow insidethetest section,refertoZassoetal.[30].
2.5. Scalingparameters
The main scaling objective in this experiment is to correctly replicate the advanceratio ofthe full scale model, ensuring also the similarityofthe reducedfrequency.Advance ratioscaling re-latesthreeparametersofthetest:angularspeedoftherotor, geo-metricscaleandfreestreamvelocity.Thehelicoptermodelisnota scaledmodelofanexactaircraft,howevercouldbetakenas repre-sentativeofamediumsizehelicopterwiththegeometricscalein theorderof1:13.So,thegeometricscalewasfixedbythescaleof the existinghelicoptermodel.The rotationalfrequencyofthe ro-torwasselectedhighenoughtoincreasetheReynoldsnumberand Machnumber,whilereachingthedesiredadvanceratiowithinthe limitsoffreestreamvelocityofthewindtunneltestsection. More-over,toavoidissueswithvibratoryloadscausedbyhighrotational frequencyitwasdecidedtolimittheangularspeed.Consequently, a velocity scale of 1:2.16 was set, leading to frequency scale of 6.08:1thatwerebothmaintainedduringalltests.Theparameters ofthescaledhelicoptermodelcomparedwithBo105,takenasan exampleofa realmedium sizehelicopter,are summarizedin Ta-ble1.
It isworth mentioning thatsharp-edged bodies,such asSFS1, arenotsensitivetoReynoldsnumber,duetoturbulentseparation.
Table 1
ParametersofthewindtunnelmodelandBo105.
Characteristic Scaled model Bo105
Number of Blades 4 4
Rotor Radius (m) 0.375 4.9 Angular Velocity (rad/s) 270 44.4 Blade Chord (m) 0.032 0.27 Free Stream Velocity (m/s) 4.76 10.29 Advance Ratio 0.047 0.047 Tip Mach Number 0.3 0.63 Tip Reynolds Number 2.2e5 3.9e6
Inanycase,Healy[31] suggeststhattheship-basedReynolds num-berforwindtunnelmodelsshouldbehigherthan11,000toshow thisindependence. Consideringthelower freestreamvelocity se-lected for the tests with the helicopter model (U∞
=
4.8 m/s), the Reynolds number of the ship, computed using the ship ref-erencelength,isabout350,000whichiswellabovetheminimum requiredReynoldsnumber.3. Experimentalresults
The experiments were conducted in two different test
cam-paigns.Thefirstcampaignwasfocused onthecharacterizationof theflowfieldontheflightdeckwithoutpresenceofthehelicopter
model, while the second campaign was planned to identify the
modification to the flowfield, due to the interaction of the heli-copterrotorwiththeshipairwake,andtotheloadingoftherotor whileperformingalandingmaneuverontheflightdeck.
Fig. 5. Effect of the free stream velocity magnitude on ship-deck airwake, contours of normalized in-plane velocity magnitude and streamlines in the symmetry plane.
Table 2
Parametersofthefirsttestcampaign. Test number Advance ratio Wind
direction No.ofPIV planes Boundary layertype TN1/TN10 0.0473 H 2/1 SF/ABL TN2/TN11 0.0828 H 0 SF/ABL TN3/TN12 0.1182 H 1/1 SF/ABL TN4/TN13 0.0473 R15 0 SF/ABL TN5/TN14 0.0828 R15 0 SF/ABL TN6/TN15 0.1180 R15 0 SF/ABL TN7/TN16 0.0473 R30 3/1 SF/ABL TN8/TN17 0.0828 R30 0 SF/ABL TN9/TN18 0.1180 R30 1/1 SF/ABL
ThreewindspeedswereselectedbasedonapossibleSHOL en-velope withamaximumspeedof50knotsatfullscale.However, sincetherotor cannotbetrimmedathighspeedsduetothelack of swashplatein thehelicopter model,the teststhat include the helicopterwereperformedonlyatthelowestadvanceratio.
A selection of theexperimental resultsof each test campaign is presented separately in the following sections. The complete databaseispublicly availableonrequest totheauthors.Itshould
be mentioned that load measurements in the second campaign
will be presented in the rotor reference frame whose x axis is noseto tail,vertical axisisbottom to topandlateral axispoints starboard, asrepresentedin Fig.3(b).Theoriginof therotor ref-erencesystem,representingthereduction pointforthemeasured momentsisthecenteroftherotor.
3.1. Shipairwakeinvestigation
Themainobjectiveofthiscampaignwastoassessthetopology oftheflowfieldovertheflightdeckwithrespecttothree param-eters:freestreamvelocitymagnitude,winddirectionandeffectof the presence ofthe ABL.Todo so,nine tests (TN1to TN9)were conductedwiththreedifferentwindvelocitiesblowingfromthree different directions, includinghead-wind (H) and red wind from 15◦(R15)and30◦(R30).Thelattertwocasesrefertothecondition inwhichthewindisblowingfromportside oftheship.Pressure
measurements were made for all the nine tests with and
with-out presenceoftheABL.PIV measurementswere performedover threelongitudinalplanesatdifferentspanwiselocations,asshown in Fig. 3(a), and only for few significant cases to limit the test-ingtime.Testparametersofthefirstcampaignaresummarizedin Table 2, wheretheadvanceratios reproducedin theexperiments correspondto20,35and50knotsforthefullscalehelicopter.
Fig.5compares thePIVmeasurements ofthesymmetryplane ofthedeckfortwofreestream velocities.In bothconditions,the topologyoftheflowfieldoverthedeckistheonecommonlyseen downstreamofathreedimensionalbackwardfacingstep[32].The flowfieldcanbedecomposedintothreemainzones:recirculation, reattachment and redeveloping regions as observed in [33]. Re-gardingthesharpedgesoftheship model,therecirculationzone
occursimmediatelybehind the hangarwall andreattaches about half-wayalongtheflight deck.Theairwakeisdirecteddownward over the recirculation zone, reflecting the fact that streamlines are bendingtowards the region oflow pressure. The comparison between Fig. 5(a) and Fig. 5(b) shows that the structure of the
flow field does not change by increasing the Reynolds number,
asshown in other studies [34,35]. As a matter offact, the
min-imum ship-based Reynolds number of 3.5
×
105 is high enoughtocreate aturbulent flow field andhave aflow topology mainly drivenbythepresenceofthesharpedgesoftheship.Theresults showninFig.5comparewellwiththeinvestigationperformedin [35],where caseswithdifferentlength ofthesuperstructure and deckarecomputed.Inparticular,consideringthegeometryofSFS1, wheretheratiobetweenthelengthofthesuperstructure(Ls)and
theheightofthehangarwall(Hw)isLs
/
Hw=
7.5,theestimationofthelengthoftherecirculationzoneobtainedbytheexperiments isLdr
/
Hw=
1.96,slightlylowerthatthevalueestimatedusingtheImprovedDetached-EddySimulationturbulencemodelin[35],that isinsteadequaltoLdr
/
Hw=
2.0.Furthermore, in-plane velocity magnitudes at four sections along the deck, with position X normalized with respect to the length of the deck (Ld
=
2186 mm), are compared in Fig. 6. Itshould be mentioned that the height is normalized withrespect to the height of the reference Pitot probe, Zref
=
1517 mm. Tobe moreclear, thelimitsof thehangar wall, fromflight deckup to Hw, is specified within a black box. From these diagrams it
is possible to estimate the position of the vortex corethat is at X
/
Ld=
0.82 and Z/
Zref=
0.46, andthat can be compared withthe results of Ref. [35] — shown for Ls
/
Hw=
7.0 — equal to X/
Ld=
0.85 andZ/
Zref=
0.47Fig.7 showsthe effectofthe wind directionon theship air-wake at U∞
=
4.8 m/s. On the symmetry plane, the size ofthe recirculationzonebecomeslargerwhenthewindisblowingfrom R30,extendingtheareainterestedbythevortexintheupperpart ofthedeck.However,thepositionofthereattachmentpointisnot affected significantly. In this case, almost half of the flight deck iscoveredbyasignificantly lower-than-freestream-velocityregion which may particularly affect the loading of the rotor and pilot activities duringshipboard operations.In theport side plane,the recirculationregionstarts todevelop closetoedge ofthehangar, whilemostofthedeckremainsunaffected,sinceinthisregionthe airdoesnotflowpastthehangarstepbeforearrivingtothedeck. Moving starboard, the recirculation region increases in size and thustheportionoftheshipdeckinterestedby low-speedflow in-creasesaswell.Consequently,itispossibletosaythatthe symme-tryofthehorseshoevortexcreatedbehindthehangarwall,whose topologyisshownin[36,35],isclearly lostwhenthewindblows fromthe side, withthe vortex core position that seems to drop onlyonthestarboardside.Sectionalvelocityprofilesarecompared forthiscaseinFig.8alongthreePIVplanes:starboard,symmetry andport planes.These plots clearly show that starting fromthe sectionatx/
Ld=
0.2,thelossofsymmetrybetweenstarboardandFig. 6. Time-averaged normalized in-plane velocity magnitude for different stations along the deck on the symmetry plane.
Fig. 7. Effect of the wind direction on shipdeck airwake, contours of normalized in-plane velocity magnitude and streamlines, U∞=4.8 m/s.
Fig. 9. Effect of the wind direction on pressure coefficient contours of the isolated ship, U∞=4.8 m/s.
Fig. 10. Effect of the ABL on shipdeck airwake, contours of normalized in-plane velocity magnitude and streamlines in the symmetry plane, U∞=4.8 m/s.
Fig. 11. Effect of the ABL on pressure coefficient contours of the isolated ship, U∞=4.8 m/s,β=0◦.
portsideisquitelarge.Fig.8showsthattherotorbladeswill ex-perienceasignificantlydifferentfluidflowwhilerotating,andthat thosedifferenceswillbecomesteeperwhilethehelicopterwill fol-low alateralrepositiontrajectory toenterintothelanding region starting fromtheport side.Those speeddifferenceswillbe more significantwhentheaircraftisclosertothehangarwall.
This flow behaviour is confirmed by the comparison of the
pressure coefficient distribution measured on the deckfor head-wind and R30 case, shown in Fig. 9. In particular, looking at Fig.9(b)fortheR30case,thehighpressurerecoveryregionmoves downstream andstarboard onthe flightdeckwithrespectto the headwindcaseundertheinfluenceofthe wind blowingfromthe port side of the ship. A low-pressure region can be observed as wellonthestarboardsideonthehangarwall,differentlyfromthe
headwindcasewherethepressureremains homogeneousonthis
partofthedeck.
To quantify the effects of the ABL on the flow field a
spe-cific test with turbulators upstream of the ship was performed. Fig.10comparestheairwakevelocitymeasuredonthesymmetry plane with SF and ABL in headwind condition. Looking at time-averaged velocities, the airwake structure is very similar. This is
further confirmed by the comparison of the pressure coefficient
distribution on the deck, shown in Fig. 11. These results are in linewiththe outcomesofthe workby Thedin etal. [37],where thesame topology,intermsof time-averagedflowfield, resulted fromtime-accurateCFDsimulations of theuniformflow in com-parisonwithsteadyandunsteadyABLcases.Theresultspresented in[37] demonstratethattheunsteadyfeaturesduetopresenceof
Fig. 12. Effect of the ABL on shipdeck airwake, contours of root mean square of the in-plane velocity components, comparison of TN1 and TN10, U∞=4.8 m/s.
turbulence inboundary layerare evident only comparingthe in-stantaneous snapshots of the velocity profiles. The effect of ABL was also investigatednumerically by Forrest andOwen [38] that comparedtheresultsofDetached-EddySimulationwithflighttest data measured at hanger height of Type23Frigate atGreen 10◦. Thecomparisons,madeforlimitednumberofdatapointsavailable from full scale at-sea measurements, showedthat CFD computa-tions withABLandSF predictessentially thesamewakepattern. Only,aslightlylowerintensitywasfoundfortheABLcaseinmost locationsdueto thelower incident velocities.However, it should be noted that they did not include the effect of increased free stream turbulence, andABL model was limited to a logarithmic velocity profile. Moreover, Polsky [39] investigated ABL effectby time-accurateCFDsolutionoftheflowfieldoverthedeckof Land-ing Helicopter Assault(LHA)experiencingbeamwinds(Red 90◦). The outcomesofthiswork showedthat theinclusion oftheABL improvedthecomparisonwithexperimental dataoftheaveraged flow-fieldintheplaneofonelandingpointwhichwascompletely exposedtothebeamwind,whileintheareaswherethe obstruc-tion of the superstructure dominated the flow field the results werenotmodifiedsignificantly.
As a matter of fact, a definite conclusion about the impor-tanceofincludingan ABL inexperimentscould beprovided only
by time-resolved PIV measurements showing possible different
topologies of the flow field due to the higherunsteadiness level introduced by turbulent ABL profile. However, time-resolved PIV wasnotavailableduringthistestcampaign.
Thecontoursofrootmeansquareofthein-planevelocity com-ponentsevaluated byPIVarepresentedinFig.12togivean indi-cationoftheturbulencelevelinvolvedinthetwocasesconsidered. Thiscomparisonalsopresentsa similarbehaviour,with compara-blevaluesbetweentheSFandtheABLtestcases.So,itispossible
to assert that in head wind condition, the mean flow over the
flightdeckisnotoverlyinfluencedbythepresenceoftheABL. Regarding the effectof ABL on rotor trim it should be noted that the flowfield outside the deck is quite different in the two cases. The reason is clearlooking at the two velocity profiles in the test chamber forSF and ABL cases shown in Fig. 4. In fact, when theairspeedisthe sameatthealtitudeof thePitotprobe, itshowsadifferenceofabout15–20%attheflightdeckaltitudes.
Consequently, it may be expected that the variation of dynamic pressure,betweenthedeckarea andoutsidethedeck, willlikely affecttheflightofhelicopterperformingtheshiplanding maneu-ver.
3.2.Helicopter–shipinteraction
Theobjectiveofthesecondtest campaignwasinvestigationof therotor loadsduring a landing maneuver on the ship deck, af-fected by the presence of the ship and its airwake. To simulate the landing maneuver, the helicopter was positioned in a series ofpoints representative ofa typical stern landing trajectory. The trajectory, asshown in Fig.13, consistsof five points (P1 to P5) thatcanbedividedintotwodistinctivesegments:theinitialphase inwhichthehelicopterapproachestheflightdeckfromthestern side along the center line of the flight deck, followed by a de-scentphase,i.e.anobliquepathtowardsthelandingpoint,which isconsidered close tothe centerof theflight deck. Furthermore, three additionalpoints above the landing point were selected to simulateaverticaldescent(P5toP8).
This campaign was conducted first in a condition similar to hover,i.e.withnowindblowinginthewindtunnel,andthenfor onlyone freestreamvelocity (U∞
=
4.8 m/s,correspondingtoan advanceratioof0.0473)blowingfromtwodifferentdirections,i.e. headwindandR30.Indeed,higherfreestreamvelocitycouldlead tomoments beyondthebalance fullrangevalues duetothe im-possibilityoftrimming therotor. Giventhelimiteddifferencesin meaneffectofanABLontheflowfield,helicopter-shipinteraction testswereperformedwithout presenceofan ABL. Pressuremea-surements on the flight deck were performedand inflow of the
rotorwas surveyedby PIV forsome selectedtest points. The po-sitionoftherotor centerforalltest points (P1toP8) islistedin Table3,in whichthe heightrefers to the verticaloffsetwith re-specttothelanding spoton theflightdeck( Z
−
Zdeck), and X isthedistancefromthesternoftheship.
Fig.14comparestheloadsmeasuredinhovertestforboth tra-jectorieswith standard IGEmodels taken fromthe literature, i.e. Cheeseman&Bennett[40] andFradenburgh[41].Asshownin[42], theyarecomparablewiththeexperimentalmeasuresatfullscale. Thesamehelicoptermodelusedinpreviouswindtunneltest
cam-Fig. 13. Side-view of the landing trajectory. Circles and crosses represent the position of the rotor center, respectively in the stern landing and vertical descent.
Table 3
Positionoftherotorcenteralongthe landingtrajectory.Theheightismeasured with respecttothe shipdeck,exceptforOGEtest whichismeasuredfrom the ground.
Test point Height [mm] Height/R [-] X [mm]
P1 600 1.60 500 P2 600 1.60 200 P3 500 1.33 −200 P4 400 1.07 −600 P5 300 0.80 −1000 P6 400 1.07 −1000 P7 500 1.33 −1000 P8 600 1.60 −1000 OGE 1135 3.03 700
Fig. 14. Groundeffect:Comparisonofthrustmeasuredduringtheexperimentwith datafromCheeseman&Bennett[40] andFradenburgh[41].ForP1andP2theZ/R
distanceismeasuredfromthewindtunnelgroundandnottheshipdeck.
paign[23] showedaperfectmatchwithFradenburghdatainideal IGEconditions.So,alldifferencesshowninFig.14maybe associ-atedwiththeeffectoftheadditionalpresenceoftheshipmodel. It is interesting to see that, at the same altitude from the ship deck,thecasesfartherfromthehangarwall(P3,P4)showalways a higherthrust than those closer(P7,P6). Anexplanation ofthis differenceisgiveninthefollowing,usingmomentsmeasurements too.Inaddition,thepartialocclusionofthewakeduetotheship deckinP2is notsufficient toleadto significantlydifferentloads thanthoseobtainedinP1.
Fig. 15 compares the thrust and moments measurements in
hover andwind-on testsforbothlanding trajectories.In this
fig-Table 4
LoadcoefficientsatOGEposition.
Rotor position CT CQ
OGE 7.28×10−3 7.74×10−4
ure,theresultsareexpressedinnon-dimensionalform,normalized withrespecttothethrustandtorquecoefficientsmeasuredatOGE positionreportedinTable4.Asareminder,theloadmeasurements arecastintherotorreferenceframe(i.e.x axisnosetotail,z axis bottom to top and y axis starboard), as represented in Fig. 3(b) implyingthat a positive pitch moment corresponds toa nose-up moment,anda positiverollmomenttoport rotationoftherotor thrust.
As shown in Fig. 15(a), thrust is increasing as the helicopter approachesthelanding point,owingtothegroundeffectinduced bytheshipdeck.Thethrust variationismoresignificantinhover (wind-off)condition, wherean increase ofabout15% oftheOGE valueisobtainedatthelandingpointforthesternlanding.
Thesamebehaviour canbeobservedinverticallandingaswell (see Fig. 15(b)). However, the ground effect is less intense with respecttothesternlandingatthesameheight,whenconsidering the wind-off case. Forinstance, in hover when the helicopter is placed in P6 ( X
= −
1000 mm, Z−
ZDECK=
400 mm), the rotorexperiencesonly8% increaseinthrust withrespect to10% in P4 ( X
= −
600 mm,Z−
ZDECK=
400 mm),whichisatthesameheightbutfartherfrom thehangar wall. Thissmall difference iscaused bythedevelopment ofa recirculationregionbetweenthehangar wallandthehelicopter[43],whichconsistsoftheforepartofthe rotorwakethatisdeflectedbytheshipdeckandthehangarwall andthenre-ingestedintotherotor.Thiscausesaslightlyincreased induced velocity on the forepart ofthe rotor, together withthe consequentthrustlossandasmallnose-down(negative)variation ofpitchmoment particularlyapparent forthevertical descent as thehelicopterismoved downwards(seeFig. 15(d)). Forthestern landing atpoint P3, theslight peak ofthe pitching moment can bedueto thenon-symmetrical groundeffectastherotor diskis onlypartiallyinsidetheflightdeck(seeFig.15(c)).Concerningthe roll moment,variations are more limited but a general trend to decrease its value moving toward the flight deck isrecognizable forbothlandingtrajectories.
Forheadwindtests,itcanbeappreciatedthatthebeneficial ef-fect dueto ground is mitigated withrespect to hover as can be seeninFigs. 15(a)and15(b) forboth sternlanding andthe low-ermostpositionofthe verticaldescent. Thisiscoherentwiththe verticalinduced velocity measured by PIV above thelongitudinal symmetryaxisoftherotor.Fig.16showsthecontoursofthe ver-tical velocity component measured by PIV at landing point (P5) inhoverandheadwindcondition. Inparticular,thecomparisonof the velocity profiles on the lower edge of the PIV measurement
Fig. 15. Loadsactingontherotor:comparisonofsternlanding(left)andverticaldescent(right),forthreedifferenttestconditions:hover(TN21),headwind(TN19)andR30 (TN20).Ontheleft,goingfromrighttoleft,thehelicopterismovingfromP1toP5.Ontheright,goingfromrighttoleftthehelicopterismovingfromP8toP5.
area is showninFig.17(a) asan indication ofthe downwash on therotor disk.Thisfigureshowsthat thereductionofthe down-washdue togroundeffect isclearly mitigated whenthe wind is blowing.
Itshould benoted thatinthe firstpoint (P1) wherethe heli-copteristotallyoutsideoftheshipairwake,alargepitchmoment isgeneratedbytherotorduetotheinfluenceoftheexternalwind onthefixedrotor.Asexpected,thisbehaviourismainlyrelatedto
Fig. 16. Contours of the vertical velocity component measured by PIV above the longitudinal symmetry axis of the rotor in hover and headwind with rotor positioned at P5.
Fig. 17. ProfilesoftheverticalvelocitycomponentmeasuredabovethelongitudinalsymmetryaxisoftherotorontheloweredgeofthePIVmeasurementareainhoverand headwindwithrotorpositionedatP5andP8.
thenon-symmetricalinducedvelocitydistributionthatislargerin therear partoftherotor diskastherotorwake ismoving back-ward. This effect is obviously reduced when the helicopter is in the ship airwake andthe horizontalvelocity resultsto be lower, implying significant negative variations ofthe pitch moment can beappreciatedwhilethehelicopterenterstheairwakeoftheship, about60% and45% instern landingandvertical descent, respec-tively.Anotherpossiblecontributiontothisisthedownwash pro-ducedbytherecirculationbehindthehangarwallontheforepart ofthediskthatislargerwhenthehelicopterisclosertotheflight deck.The comparisonofthe verticalvelocity profilesbetweenP5 andP8showninFig.17presentsthedifferenceofdownwash dis-tribution coherentwiththetrendofthe pitchmomentalong the vertical descent. On the other hand, the roll moments for both landing trajectories present positive values as expected in head-windcondition,whilethevariationsarerathersmall,althoughnot negligible.
Forthe R30 tests, the beneficial effectdueto groundis more mitigated than inheadwind. Inparticular, about7% of thrust re-ductionisobservedwithrespecttohovertestwhentherotorcraft is placedatlanding spot(see Figs. 15(a) and15(b)).Thisis even moreaccentuatedduringtheverticaldescent,particularlyatP8in whichthethrustreachesavaluelowerthaninOGEcondition(see Fig.15(b)).
Looking at the moment coefficients the most apparent effect
withrespecttotheheadwindcondition isrelatedtotheroll mo-ment,whereasthetrendofthepitchmomentisquitesimilar.This couldbemainlyexplainedconsideringtheeffectofthefreestream oninducedvelocitydistributionontherotordiskasalready
men-tionedforthe headwind condition. Indeed,in thiscasethe non-symmetricaldistribution from windward to lee side of the rotor diskproducesanaerodynamicmomentcontributionthatprojected alongthelongitudinalaxisofthehelicopterresultsinalarge neg-ative roll moment, ascan be seen in Figs. 15(e) and 15(f). This effectisreducedapproaching theflightdeckduetothereduction ofthehorizontalvelocity,explainingthepositivevariations ofthe rollmomentmeasuredalongboththelandingtrajectories.
To have a better understanding of the interacting flow field, pressure contours over the flight deck are presented in Fig. 18
comparing the headwind and R30 for three different positions
ofthe helicopter:P5, which corresponds to the landing spot, P4, whichisapointontheobliquepathandP8, whichisthehighest pointoftheverticalpath.Cpvalueshigherthanoneareconsistent
withthepressure beingnon-dimensionalisedusingthe wind dy-namicpressure,sinceinthe consideredtest conditionsthe rotor-inducedvelocitywashigherthantheasymptoticwind.Looking at thelandingspot(P5),Fig.18(b),astronghigh-pressureregion cor-respondingto the rotorwake impingement area isclearly visible inboth cases,buta deeperlow-pressure region onthe port side of the deck can be appreciated in R30 condition, owing to the wind blowing fromthe port side of the ship. The pressure con-tours remain qualitatively similar for the other two test points, howeverwithlower pressurepeaksin correspondencetothe ro-torwakeimpingement area, relatedtothe higherpositionofthe rotor withrespect to the deck.The higher pressuremeasured in headwindwithrespectto R30iscoherentwiththelarger ground effect.
Fig. 18. Comparison of the pressure coefficient contours for headwind and R30 with rotor positioned at different height over the deck.
4. Numericalmodelling
A multibodymodelof theexperimental rotor hasbeen
devel-oped using MBDyn, a free general-purpose multibody dynamics
analysissoftwaredevelopedatPolitecnico diMilano [44].MBDyn features the integratedmultidisciplinary simulation ofmultibody systems,including nonlinearmechanics ofrigidandflexible bod-iessubjectedtokinematicconstraints,alongwithsmartmaterials, electric andhydraulic networks, active control andessential ele-ments of rotorcraft aerodynamics [45]. The multibody model de-veloped forthisstudy consistsofa hingeless, stiff-in-plane rotor withfour elasticblades connectedto thehub through arevolute hinge, which allowsthe rotation aboutthefeatheringaxisof the
blade.This degree offreedom, along witha rigidpitch link con-nected to theswashplate, allows pitch control. The eigenanalysis ofthemodelinvacuumshowsthatthenon-dimensionalflapping frequencyatnominalrotor speed is
ν
β=
1.22/rev. Therotor pa-rametersarethesameasthescaledmodelpresentedinTable1.TheBladeElement/MomentumTheoryaerodynamic modelhas
beenconsideredaswell,usingaNACA0012airfoil.Thecalculation oftheloadingatthebladetiphasbeencorrectedforthree dimen-sional effects considering 2% of aerodynamic tip loss. Moreover, groundeffecthasbeenincorporatedintothesimulationbasedon themodelpresentedbyFradenburghin [41].
As mentionedin Section 3.2the currentexperiment doesnot
κ
x=
fx 43
(
1−
cosχ
−
1.
8μ
2
)/
sinχ
(3)κ
y= −
2 fyμ
Here, fxand fyareempiricalfactorsusedtomodifytheinflow
distributioninbothlateralandlongitudinaldirections( fx
=
fy=
1in Drees model) [47]. These factors have been set in order to generate the same load coefficients, including thrust, torque and in-planemoments,whenthehelicopterispositionedintheinitial point ofthelanding trajectory(P1). Thismodelhasbeenusedas thebaselineforthedevelopmentoftheflowdistortion identifica-tionalgorithmexplainedinthefollowingsection.
5. Flowdistortionidentificationalgorithm
Inordertoimprovethefidelityofthesimulationenvironment, the flow field of the rotor should be representative of the
aero-dynamic interaction between the airwake of the ship and the
rotor-inducedwake.Withthisaim,anidentificationalgorithmwas developedtoreconstructtheeffectofthevariationoftheflowfield due to interaction of the rotor wake with the surrounding flow caused by the presence of the ship model. The starting point of thisidentificationwere theforcesandmomentsmeasured onthe scaled rotor model. Thisidentified flow distortion is represented through anadditionaldownwash term, thatisincluded asan ex-ternalinputtothesimulationenvironment.Thistermisexpected toreproducethelow-frequencycontentoftherotorloadsresponse whilerotorcraftismovingthroughtheairwakeoftheship.
Inthepresentstudy,theflow distortionvelocityisconsidered tohavealineardistributioninradialandazimuthaldirection (sim-ilartoinflow),definedasthefollowingequation:
vsh
=
vsh0+
vshcr cos(ψ )
+
vshsr sin(ψ )
(4)Thissimple distributionwas chosen becauseit istheonly model
who can be identified through the measurement of global rotor
thrustandmoments.Consideringinfact,therelationshipbetween inflowspeedandrotorloadsdescribedbyPeters-Heinflowmodel [48], iftheliftdistributionontherotordiskislimitedtoalinear trend, sothat only thrust, pitch androllmoments are the resul-tants, thenonlyan inflowdistributionlikeEq. (4) isactivated.To identifythecoefficients vsh0,vshc andvshs,asimpleseriesof op-timizationproblemsissolved. First,theaveragevalueoftheflow distortion vsh0 is identified bylooking forthe value ableto
gen-eratethesamethrust coefficientmeasuredintheexperiment.So, thefollowingminimizationobjectivefunction( J )hasbeendefined
J
=
(
cT−
cTwt)
2 (5)where CT is the computedthrust coefficient with themultibody
modelandCTwt isthemeasuredone.
Then,thecoefficientsofthelineardistributionoftheflow dis-tortion velocity (Eq. (4)), vshs and vshc, are identifiedto improve thematchingofthein-planemoments.So,inthiscasethefigure ofmeritfortheoptimizationproblemisrelatedtothematchingof
momentcoefficientsnamely:
loadcoefficients,advanceratioandposition(radialandazimuthal), asindicatedinthefollowingequation:
U
=
U∞+
v=
U∞+
vsh(
ci,
r, ψ,
μ
)
+
vi(
ci,
r, ψ,
μ
)
(7)whereU∞ isfreestreamvelocityand
v istheadditional veloc-ityduetothepresenceofwakesoftheobstacleandoftherotor, andci are theforce and momentcoefficients. The simulation
re-sults,implementingthesolutionoftheoptimizationalgorithm as anexternalflow distortion,arecomparedwiththeload measure-ments in headwind condition (TN19) for all test points. In this case the hypothesis of flow distortion perpendicular to the ro-tordiskhasbeen made, consideringthe in-planespeed variation negligible withrespect to thein-plane velocity generated by the rotorangularspeed(r).Fig.19comparestherotorloadsforboth trajectories,includingsternlandingandverticaldescent.The hor-izontalaxisrefers to thesame coordinatesystemusedin Fig.15
andalltheresultsarepresentedintherotorreferenceframe, de-finedinSection3.
Thesimulationresultsarehighly consistentwiththe
measure-ments of thrust and in-plane moments, demonstrating that the
external flow distortion field is representative of the effects of the ship airwake on therotor performance. However, the torque
curve computed by MBDyn shows an offset with respect to the
experiment.Thisoffsetcouldberelatedtotheadditionaldrag con-tributionprovidedbythe innerpartoftherotormeasuredinthe experimentandnotmodelled.
Thefirstharmonicsoftheflowdistortionarecomparedforall test points in Fig. 20.The resultsshow that the initial points of thelanding(P1andP2)donotneedadditionalmodificationofthe velocity field since theeffect ofthe ship airwakeon therotor is negligible.However,movingtowardsthelandingpoint,the ampli-tudeoftheflow distortionis larger,whichis consistentwiththe variationofthein-planemoments(Fig.19).
Theverticalvelocityabove thelongitudinalaxisoftherotoris comparedwiththePIVmeasurementsextrapolateduptotherotor disk,sincethe closestlineof thePIV windowwas 48mm above therotordisk.Fig.21showstheresultsobtainedforthecasesof thelandingpoint,P5,andpointP6whichhasaverticaloffsetwith respectto P5. Thisfigureshowswitha blueline thedistribution ofthe velocity normalto the rotor diskfor the original unmod-ifiedmodel of the rotor. This distribution is clearly uncorrelated with the distribution detected by the PIV. The red line instead, representsthe normalvelocity distributionobtainedby including the flow distortion term, Eq.(4), in the simulation.The slope of theredlinesseeminbetteragreementwithPIVdistributionsthat present,inbothcasesP5andP6,highervelocitypeaksontheleft handside.Tofurtherclarifythisaspect,Fig.22comparestheslope ofthevelocity fielddistribution calculatedjoiningthetwo veloc-ity peaks of PIV distributions, with the slopes used in the two multibodymodels.Forbothcases,P5andP6,theinclusionofflow distortion terms allows to recover a positive slope. Furthermore, themagnitudeof theslope ishigher onP5than on P6, both for PIVmeasuresandforthemodelsthatincludedtheflowdistortion terms.Therefore, it is possible tosay that the usage of the flow distortioncorrectiontermsseems tomodifythevelocity distribu-tioninadirectionthatismorecompatiblewiththetrendsshown bythePIVmeasures.
Fig. 19. Comparison of rotor loads from experiment and MBDyn simulation, TN19: U∞=4.8 m/s,β=0◦.
Fig. 20. First harmonic of flow distortion in different test points, TN19: U∞=4.8 m/s,β=0◦.
6. Conclusions
Thispaperpresentsan experimentalstudy oftheaerodynamic interactionbetweenascaled-downhelicoptermodelandthe Sim-ple Frigate Shape 1 model. First, a series of wind tunnel exper-iments was carried out to study the structure of the flow field over the deck without the presence of the helicopter. This cam-paign was performedfor a range of wind velocity anddirection using larger scale modelswith respect tothe literature. Pressure
measurementsandPIVsurveyontheflightdeckdemonstratedthe significanteffectofwinddirectiononthesizeofrecirculationzone anditsexpansion overtheflightdeck.Theatmosphericboundary layer consistent with the coastal area was also simulated in the wind tunnel.The experimental resultsshowed that the presence ofboundarylayerdoesnotsignificantlyaffectthemeanflow fea-turesovertheflightdeck,butthatcansignificantlyaffecttheflight dynamicsoftheaircraftperformingthemaneuver.
Fig. 21. Normalized vertical velocity along the longitudinal axis of the rotor, comparison of PIV measurement with normal velocity in MBDyn, TN19: U∞=4.8 m/s,β=0◦.
Fig. 22. Comparisonofthecosineharmonicsofthenormalvelocity,TN19:U∞=4.8 m/s,β=0◦.
In thesecond partofthe testcampaign, the helicoptermodel was positioned in a series of points representative of a typical stern landing trajectory and a vertical descent above the flight deck,tomeasuretherotorforcesandmoments.Thegroundeffect ofthedeckwasnoticeableasthethrustincreaseswhilehelicopter movestowardsthedeck.Thisbeneficialeffectismitigatedin pres-enceofthewind,especiallywhenthewind isblowingfromport sideoftheship.Regardingthein-planemomentsinwind-ontests, thereisasignificant variationofpitchmomentalongthelanding trajectory which remains almost unchangedin caseof headwind and R30. However, the variation of roll moment showsa strong dependencytothewinddirection.
The experimental results enabled to achieve a deeper insight abouttheinteractionalaerodynamicsbetweenshipandhelicopter inshipboardoperations,particularlyusefulforthedevelopmentof
high-fidelitysimulationenvironmenttoimprovepilottraining.The experimentalresultswereusedtodevelopapreliminarymodelof theflowdistortioneffectabletorecoverarealisticrepresentation oftherotor loadsduringthesimulationofthelandingmaneuver. However, the identification of unsteady interactional effects will requireamodificationtotheexperimentalsetuptocollectthe nec-essarydata.
Declarationofcompetinginterest
Theauthorsdeclarethattheyhavenoknowncompeting finan-cialinterestsorpersonalrelationshipsthatcouldhaveappearedto influencetheworkreportedinthispaper.
Acknowledgements
ThisresearchwassupportedbyNITROS(NetworkforInnovative Trainingon ROtorcraftSafety)projectwhichhasreceivedfunding fromtheEuropean Union’sHorizon2020research andinnovation
program underthe Marie Skłodowska-Curie grant agreement No.
721920.
References
[1]S.J.Hodge,S.J.Zan,D.M.Roper,G.D.Padfield,I.Owen,Time-accurateship air-wakeandunsteadyaerodynamicloadsmodelingformaritimehelicopter sim-ulation,J.Am.HelicopterSoc.54 (2)(2009)022005.
[2] D.T. McRuer,Interdisciplinaryinteractions and dynamicsystems integration, Int.J.Control59 (1)(1994)3–12,https://doi.org/10.1080/00207179408923067. [3]L.A.Hoencamp,M.D.Pavel,Helicopter-shipqualificationtestingforthenh90 nfhhelicopter,in:38th EuropeanRotorcraftForum,Amsterdam,the Nether-lands,2012.
[4] I.Owen,M.D.White,G.D.Padfield,S.J.Hodge,Avirtualengineeringapproach totheship-helicopterdynamicinterface-adecadeofmodelling and simula-tionresearchat the UniversityofLiverpool,Aeronaut. J.121 (1198)(2013) 1233–1248,https://doi.org/10.1017/aer.2017.102.
[5]G.Quaranta,G.Barakos,M.White,M.Pavel,M.Mulder,Nitros:aninnovative trainingprogramtoenhancerotorcraftsafety,in:AmericanHelicopterSociety International74thAnnualForum,Phoenix,USA,2018.
[6] D.M.Roper,I.Owen,G.D.Padfield,S.J.Hodge,Integratingcfdandpiloted sim-ulation toquantifyship-helicopteroperatinglimits, Aeronaut. J. 110 (1109) (2006)419–428,https://doi.org/10.1017/S0001924000001329.
[13]J.D.Kelle,D.A.Wachspress,J.C.Hoffler,Realtimefreewakeandshipairwake modelforrotorcraftflighttrainingapplications,in:71stInternationalAnnual ForumAmericanHelicopterSociety(AHS),2015.
[14] I.Oruc,J.F.Horn,J.Shipman,S.Polsky,Towardsreal-timepilot-in-the-loopcfd simulationsofhelicopter/shipdynamicinterface,Int.J.Model.Simul.Sci. Com-put.8 (4)(2017),https://doi.org/10.1142/S179396231743005X.
[15] S.J.Zan,Experimentaldeterminationofrotorthrustinashipairwake,J.Am. HelicopterSoc.47 (2)(2002)100–108,https://doi.org/10.4050/JAHS.47.100. [16] R.G.Lee,S.J.Zan,Windtunneltestingofahelicopterfuselageandrotorina
shipairwake,J.Am.HelicopterSoc.49 (2)(2004)149–159,https://doi.org/10. 4050/1.3092869.
[17] R.G.Lee,S.J.Zan,Windtunneltestingofahelicopterfuselageandrotorina shipairwake,J.Am.HelicopterSoc.50 (4)(2005)326–337,https://doi.org/10. 4050/1.3092869.
[18] C.H.Kääriä,Y.Wang,G.D.Padfield,J.S.Forrest,I.Owen,Aerodynamicloading characteristicsofamodel-scalehelicopterinaship’sairwake,J.Aircr.49 (5) (2012)1271–1278,https://doi.org/10.2514/1.C031535.
[19] C.H. Kääriä, Y. Wang,M.D. White, I. Owen,An experimentaltechnique for evaluatingtheaerodynamicimpactofshipsuperstructuresonhelicopter oper-ations,OceanEng.61(2013)97–108,https://doi.org/10.1016/j.oceaneng.2012. 12.052.
[20]Y.Nacakli,D.Landman,Helicopterdownwash/frigateairwakeinteraction flow-fieldpivsurveysinalowspeedwindtunnel,in:AmericanHelicopterSociety 67thAnnualForum,VirginiaBeach,VA,2011.
[21]D. Stargel, D. Landman, A wind tunnel investigation ofship airwake/rotor downwashcouplingusingdesignofexperimentsmethodologies,in:50thAIAA AerospaceSciencesMeeting,Nashville,Tennessee,2012.
[22]T.Quinliven,K.Long,Rotorperformanceinthewakeofalargestructure,in: AmericanHelicopterSociety65thAnnualForum,Grapewine,TX,2009.
[23] G.Gibertini,D.Grassi,C.Parolini,D.Zagaglia,A.Zanotti,Experimental investi-gationontheaerodynamicinteractionbetweenahelicopterandground obsta-cles,Proc.Inst.Mech.Eng.,PartG,J.Aerosp.Eng.229 (8)(2015)1395–1406,
https://doi.org/10.1177/0954410014550501.
[24] D.Zagaglia,A.Zanotti,G.Gibertini,Analysisoftheloadsactingontherotorof ahelicoptermodelclosetoanobstacleinmoderatewindyconditions,Aerosp. Sci.Technol.78 (8)(2018)580–592,https://doi.org/10.1016/j.ast.2018.05.019. [25]N.Taymourtash,D.Zagaglia,A.Zanotti,G.Gibertini,G.Quaranta,Windtunnel
investigationofahelicoptermodelinshipboardoperations,in:45thEuropean RotorcraftForum,Warsaw,Poland,2019.
[26]C.Wilkinson,S.J.Zan,N.E.Gilbert,J.D.Funk,Modellingandsimulationofship airwakesforhelicopteroperations:acollaborativeventure,in:AGARD Sympo-siumonFluidDynamicsProblemsofVehiclesOperatingNearorintheAir-Sea Interface,Amsterdam,theNetherlands,1999.
10.1016/j.jweia.2007.06.040.
[34]S.Zan,Surfaceflowtopologyforasimplefrigateshape,Can.Aeronaut.Space J.47 (1)(2001)33–43.
[35] T.Li,Y.Wang,N.Zhao,Numericalstudyoftheflowoverthemodifiedsimple frigateshape,Proc.Inst.Mech.Eng.,PartG,J.Aerosp.Eng.(2020),https://doi. org/10.1177/0954410020977752.
[36] R.Bardera-Mora,A.Conesa,I.Lozano,Simplefrigateshapeplasmaflow con-trol,Proc.Inst.Mech.Eng.,PartG,J.Aerosp.Eng.230 (14)(2016)2693–2699,
https://doi.org/10.1177/0954410016630333.
[37] R.Thedin,S.M.Murman,J.Horn,S.Schmitz,Effectsofatmosphericturbulence unsteadinessonshipairwakesandhelicopterdynamics,J.Aircr.57 (3)(2020),
https://doi.org/10.2514/1.C035643.
[38] J.S.Forrest,I.Owen,Aninvestigationofshipairwakesusingdetached-eddy simulation, Comput.Fluids 39 (4)(2010)656–673,https://doi.org/10.1016/j. compfluid.2009.11.002.
[39]S. Polsky,Cfdprediction ofairwakeflowfields for shipsexperiencingbeam winds,in:21stAIAAAppliedAerodynamicsConference,Orlando,Florida,2003.
[40]I.Cheeseman,W.Bennett,Theeffectofgroundonahelicopterrotorinforward flight,R&M3021,ARC,1955.
[41] E.A.Fradenburgh,Thehelicopterandthegroundeffectmachine,J.Am. Heli-copterSoc.5 (4)(1960)24–33,https://doi.org/10.4050/JAHS.5.4.24.
[42]J.G. Leishman, Principle of Helicopter Aerodynamics, Cambridge University Press,Cambridge,MA,2000.
[43] D.Zagaglia,M.Giuni,R.B.Green,Investigationoftherotor–obstacle aerody-namicinteractioninhoveringflight,J.Am.HelicopterSoc.63 (3)(2018)1–12,
https://doi.org/10.4050/JAHS.63.032007.
[44] P.Masarati,M.Morandini,P.Mantegazza,Anefficientformulationfor general-purpose multibody/multiphysics analysis, ASME J. Comput. Nonlinear Dyn. 9 (4)(2014)041001,https://doi.org/10.1115/1.4025628.
[45] P.Masarati,D.J.Piatak,G.Quaranta,J.D.Singleton,J.Shen,Softinplane tiltro-toraeromechanicsinvestigationusingtwocomprehensivemultibodysolvers, J.Am.HelicopterSoc.53 (2)(2008)179–192,https://doi.org/10.4050/JAHS.53. 179.
[46]W.Johnson,RotorcraftAeromechanics,1stedition,CambridgeUniversityPress, 2013,p. 136,Ch.5.
[47]W.Johnson,Developmentofacomprehensiveanalysisforrotorcraft.1- Rotor modelandwakeanalysis,Vertica5 (2)(1981)99–129.
[48] D.A.Peters,D.D.Boyd,C.J.He,Finite-stateinduced-flowmodelfor rotorsin hoverandforwardflight,J.Am.HelicopterSoc.34 (4)(1989)5–17,https:// doi.org/10.4050/JAHS.34.5.
