Numerical analysis of fan noise for the NOVA boundary-layer ingestion configuration

10.1016/j.ast.2019.105532

Publication date

2020

Document Version

Final published version

Published in

Aerospace Science and Technology

Citation (APA)

Romani, G., Ye, Q., Avallone, F., Ragni, D., & Casalino, D. (2020). Numerical analysis of fan noise for the

NOVA boundary-layer ingestion configuration. Aerospace Science and Technology, 96, [105532].

https://doi.org/10.1016/j.ast.2019.105532

Important note

To cite this publication, please use the final published version (if applicable).

Please check the document version above.

Copyright

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy

Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

Contents lists available atScienceDirect

Aerospace

Science

and

Technology

www.elsevier.com/locate/aescte

Numerical

analysis

of

fan

noise

for

the

NOVA

boundary-layer

ingestion

configuration

Gianluca Romani

∗

,

Qingqing Ye,

Francesco Avallone,

Daniele Ragni,

Damiano Casalino

DelftUniversityofTechnology,2629HS,Delft,theNetherlands

a

r

t

i

c

l

e

i

n

f

o

a

b

s

t

r

a

c

t

Articlehistory: Received29July2019

Receivedinrevisedform7October2019 Accepted31October2019

Availableonline7November2019 Keywords: Lattice-Boltzmannmethod Turbofan Boundary-layeringestion Aerodynamics Aeroacoustics

Aim ofthispaperis toinvestigatethe effectsoftheturbulentflowdevelopingoverafuselageonfan noiseforBLIembeddedpropulsionsystems.Suchconfigurationscansufferfrominletflowdistortionsand ingestionofturbulenceatthefanplanewithconsequentimpactonbothbroadbandandtonalfannoise. TheanalysisisperformedonamodifiedversionoftheLow-NoiseNASASDTfan-stageintegratedintothe ONERA NOVAfuselageinordertoreproducetheNOVABLIconfiguration.Thenumericalflowsolution isobtainedbysolvingtheexplicit,transientandcompressiblelattice-Boltzmannequationimplemented inthe high-fidelityCFD/CAAsolverSimulia PowerFLOW®.Theacoustic far-field iscomputed byusing the Ffowcs-Williams& Hawkingsintegral solution appliedtoapermeable surface.Allsimulationsare performedforanoperatingconditionrepresentativeofatake-offwithpowercut-back.Installationeffects duetotheBLIconfigurationarequantifiedbycomparisonwithanisolatedconfigurationofthemodified Low-Noise SDT fan-stageatthe sameoperatingcondition.It isfoundthat theBLIfan-stage, whichis notoptimal,ischaracterizedbystrongazimuthalfanbladeloadingunsteadiness,lessaxisymmetricand coherent rotor wake tangential velocity variations and higher levels of in-plane velocity fluctuations comparedtotheisolatedengine. Thisresulted innodistinct tonalcomponents andhigherbroadband levelsinthefar-fieldnoisespectra,aswellasinanincrementofcumulativenoiselevelsupto18EPNdB. Thisstudy,whichrepresentsthefirsthigh-fidelityCFD/CAAsimulationofafull-scaleaircraftgeometry comprehensiveofaBLIfan/OGV,provideswithaclearunderstandingofthechangeofthenoisesources inBLIintegratedconfigurations.

©2019TheAuthors.PublishedbyElsevierMassonSAS.ThisisanopenaccessarticleundertheCC BY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

1.1.Background

Inordertodealwiththeincreasinglystringentaviation regula-tionsforpollutionandnoiseimpact[1],theuseofUltra-High By-passRatio(UHBR)enginesonnextgenerationaircraftisreceiving moreattentionduetotheirlowerjetcoreflowvelocityandnoise emissions, and enhanced propulsive efficiencycompared to low-andhigh-bypass turbofans. Such engines have a relatively larger fandiameterwithconsequent increaseofthe bladetip speedfor constantcruisevelocity. Asa consequenceofthejetnoise reduc-tionandincreaseofthebladesize,fannoisebecomestheprimary sourceofnoisefortheseconfigurations[2].Moreover,theiractual employment on future aircraft raises new integration challenges,

*

Correspondingauthor.

E-mailaddress:G.Romani@tudelft.nl(G. Romani).

requiring special designs to install such large and heavy engines minimizingtheirimpactonaircraftperformances.

In the last two decades, manyresearchers have put their ef-forts on developingnovel aircraft configurationssuited for UHBR engines integration [3–6]. In this scenario, four different NOVA (Nextgen ONERA VersatileAircraft)aircraft geometries havebeen designed by ONERA in last few years with a particular empha-sis on engine integration: (i) a baseline architecture with wide lifting fuselage, under-wing engines, high wing aspect ratio and downwardorientedwinglets,(ii)a gullwinglayoutcharacterized by an increased dihedral angle in the wing inboard position to limit landing gears length, (iii) a podded configuration with en-ginesmountedontheaftfuselageside,and(iv)aBoundary-Layer Ingestion(BLI)configurationwithenginesinstalledontheaft fuse-lage side and ingesting the boundary-layer convecting over the fuselage[5].Starting fromthesestudies,thepresentwork, which takesplaceintheframeworkoftheEuropeanCommissionproject ARTEM(AircraftnoiseReductionTechnologiesandrelated Environ-mentaliMpact),focusesonthelastconfiguration.

https://doi.org/10.1016/j.ast.2019.105532

1270-9638/©2019TheAuthors.PublishedbyElsevierMassonSAS.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).

e

=

M

=

Machnumber

n

=

Airfoilcontouroutward-pointingnormal

p

=

Staticpressure

r

=

Fanradialcoordinate

R

=

Fanradius

Rs

=

Hemisphereradius

S

=

Ray-hemisphereintersection

˜

S

=

Reflectedray-hemisphereintersection

t

=

Time

T

=

Fluidtemperature

u

=

Wall-parallelfluidvelocity

U

=

Fluidvelocitymagnitude

v

=

Particlevelocityvector

V

=

Fanbladesectionalvelocitymagnitude

˜

V

=

Normalizedfanbladesectionalvelocitymagnitude

w

=

Wall-normalfluidvelocity

x

=

Particlepositionvector

x

,

y

,

z

=

Cartesiancoordinates

y+

=

Non-dimensionalwall-distanceinviscousunits

Greek symbols

α

=

Aircraftangleofattack

γ

=

Aircraftglideangle

δ

=

Boundary-layerthickness

η

f

=

Fanisentropicefficiency

θ

=

Directivityanglealongeachmicrophonesarc

ρ

=

Fluiddensity



=

Airfoilcontour

τ

=

Relaxationtime

T

=

Fanrotationperiod

g

=

Ground-fixedreferenceframe

k

=

k-thgroundmicrophone

j

=

j-thdiscreteparticlevelocitydirection

Superscripts



₌

_{Flowquantityfluctuation}

¯·

=

Time-averagedflowquantity

i

₌

_{i-thflightsub-segment}

Acronyms

BPF

=

Blade-PassingFrequency

BLI

=

Boundary-LayerIngestion

CAA

=

ComputationalAero-Acoustics

CFD

=

ComputationalFluidDynamics

EPNL

=

EffectivePerceivedNoiseLevel

FAR

=

FederalAviationRegulations

FPR

=

FanPressureRatio

FE

=

FineEquivalent

FW-H

=

Ffowcs-Williams&Hawkings

LBM

=

Lattice-BoltzmannMethod

LRF

=

LocalReferenceFrame

NBN

=

Narrow-BandNoise

NOVA

=

NextgenONERAVersatileAircraft

OGV

=

OutletGuideVane

PSD

=

PowerSpectralDensity

PWL

=

PowerLevel

SD

=

StandardDeviation

SDT

=

SourceDiagnosticTest

UHBR

=

Ultra-HighBypassRatio

VLES

=

VeryLargeEddySimulation

VR

=

VariableResolution

1.2. OverviewofBLItechnologyandstate-of-the-art

BLI propulsion systems aim at reducing the required propul-sive power compared to conventional tube-and-wing configura-tions [5–8]. Its theoretical propulsive benefit is based upon the possibilitytoreduce:(i)theoverallaircraftmassanddrag,dueto thenacellepylonremovalandthelower wettedsurfacearea;(ii) the power dissipation in the flow field, by reducing the exhaust jet wasted kinetic energy andfilling-inthe airframe wake veloc-itydefect.Moreover,withBLI,thepropulsionsystemispartiallyor completelyshieldedbytheairframe,dependingontheplacement oftheengine,thus yielding toa potentialnoise reductiondueto acousticshielding.However,manydrawbacksshouldbeaddressed before quantifying the actual benefits associated to BLI, such as theinletflowdistortion onengineefficiency,operability, aerome-chanicsand aeroacoustics.The fuselage boundary-layer ingestion, aswell asthepossiblepresenceofa s-ductinlet,leadtothe par-tialloss ofthe fan inflow axialuniformity, thus causing astrong azimuthalvariationofthefanbladeloadingwithaerodynamicand

aeroacousticdrawbacks. Therefore,thiskindofengineintegration deeply relies upon the possibility toalleviate the flow distortion andnon-uniformityatthefanplane.

Althoughanextensiveresearchhasbeenconductedinorderto investigate BLI propulsionsystems interms of performancesand fuelefficiency[7,9–14],andinletflowcontrolanalysis[15–17],the aeroacousticassessmentofBLIhasnotreceivedmuchattentionin the pastandonlyfewaeroacoustic studiesareavailable in litera-tureforsuchconfigurations.

Defoe et al. [18] have investigated the effects of BLI on the aeroacoustics oftransonic fan rotors. Theyimplemented a body-force formulation for the fan rotor description, extracted from a 3D Reynolds-AveragedNavier-Stokes (RANS)simulation,inan un-steadyEuler calculationandevaluating thefar-fieldnoise via the Ffowcs-Williams&Hawkings(FW-H)integralmethodusinga per-meablesurface. Theyfoundoutthatthedominantmechanismfor changes in far-fieldrotor shocknoise, dueto the boundary-layer ingestion at low free-stream Mach numbers, is the ingestion of

stream-wisevorticitygeneratedbytheinteractionoftheupstream boundary-layervorticitywiththeinletlip.

A noise assessment ataircraft-level for the NASA D8 concept has been carried out by Clark et al. [19] by using the Aircraft NOisePredictionProgram (ANOPP)comprehensivetool topredict thenoise generated by each source component, with the BLI in-fluenceon fan noise empirically modeled based on experimental data.Inthatstudy,boundary-layeringestionwaspredictedtohave adetrimentalimpactoneffectiveperceivednoiselevelsinthe or-derof15EPNdB.

Finally, Murray et al. [20] conducted aeroacoustic measure-mentsforan unshrouded rotor partiallyimmersedina turbulent boundary-layeratlow Mach numberto investigateinflow distor-tioneffectsassociated toairframe-integrated engines. Theyfound outthat,atlowandmoderatethrustconditions,therotorproduces broadbandnoise organized into haystacksgenerated by large ed-diesintheingestedturbulencebeingcutmultipletimesby neigh-boringrotor blades, contrarily to louderand moretonal acoustic signaturesobservedathighthrustconditions.

1.3.Scopeofthepresentwork

In view of the above, the existing research on fan BLI aeroa-cousticsislimitedonlytoexperimental/numericalstudiesat com-ponent level andto the analysis ofBLI full-aircraft configuration bymeansoflow-fidelitycomprehensivecodes.Sincethereareno detailed studies on the physics behind the noise generation for BLI embedded engines at full-aircraft level, the aim of this pa-peristwofold:(i)toperformthefirst,totheauthors’knowledge, high-fidelity CFD/CAA simulationof a full-scale aircraftgeometry comprehensive of a BLI fan/Outlet Guide Vane (OGV) stage;and (ii) to address BLIinstallation effects on fan noise forthe NOVA BLI aircraft configuration by comparison between the same en-gineused ina conventional non-BLI andBLI layout.The analysis iscarriedoutforanoperatingconditionrepresentativeofaflyover withpowercut-backtoreplicateoneoftherequiredconditionsfor noisecertification[21].

Inthispaper,SimuliaPowerFLOW® _{time-explicit,compressible}

andtransient solver basedonthe lattice-BoltzmannMethod/Very LargeEddy Simulation (LBM/VLES) is used to simulate and ana-lyzetheflowandtheacousticnear-fieldaroundtheBLIfan-stage configuration. The aerodynamic noise generated by the fan and its interaction with the ingested turbulence, as well as the fan wake/OGVinteractionisthenestimatedbyusinganacoustic anal-ogybasedonFarassat’s formulation1AoftheFW-Hequation ap-pliedon a permeable surface encompassing the engine andpart of the fuselage. This hybrid CFD/CAA methodology has been ex-tensively used and validatedin the past by several authors in a variety of experimental benchmark, ranging from airfoil trailing-edgenoise[22] andjetnoise[23] torotorcraft[24,25] andaircraft aeroacoustics,bothatcomponent[26,27] andfullaircraftlevel[28,

29]. More specifically, the same computational methodology has beensuccessfully validatedby Casalino et al. [30] and Gonzalez-Martino andCasalino [31] inthe field on turbofanaeroacoustics. Theypredictedtonaland broadbandnoise ofthree fan/OGV con-figurationsofthe22-inNASASourceDiagnosticTest(SDT)fanrig [32] withanaccuracyintheorderoftheexperimentaluncertainty of1dB,atbothsubsonicandtransonictipspeedconditions.

Themanuscriptisorganized asfollows.InSec. 2,an overview oftheLBM/VLESapproach,far-fieldnoiseandon-the-groundnoise computationarepresented.Thegeometriesandthecomputational setupusedinthisstudyaredescribedinSec.3.Aerodynamicand aeroacousticinstallationeffectsassociatedtotheNOVABLI config-urationare discussedin Sec.4.Finally, themain conclusionsand futureoutlookofthisstudyaredrawninSec.5.

2. Numericalmethod

2.1. LBM/VLESflowsolver

TheCFD/CAAsolverSimulia PowerFLOW® isusedinthisstudy tocompute theunsteadyflow. Thesolver isbasedonthe lattice-BoltzmannMethodwithawall-modeledVLESapproachfor turbu-lence [33–36] and it solves the Boltzmann equation for the dis-tribution function g

(

x

,

t

,

v

)

on a hexahedral mesh automatically generated around bodies. The function g represents the proba-bility tofind, inthe elementary volume dx around x and in the infinitesimal time interval

(

t

,

t

+

dt

)

, a number of fluid particles withvelocityintheinterval

(

v

,

v

+

dv

)

.TheBoltzmannequationis solvedbydiscretizingthespacevelocitydomainintoaprescribed numberofvaluesinmagnitudeanddirection.Thesediscrete veloc-ityvectorsaresuchthat,inaprescribedtimestep,oneparticlecan beadvected fromonepointofthemeshtoN neighboringpoints. In thisstudy, an hybrid formulationof thesolver is used, which allows to combine 19 (D3Q19)and 39 (D3Q39) particle velocity states,thelatterbeingusedwheretransonicflowisexpected[37]. ThestandardLBMformulationisbasedonthetime-explicit ad-vectionequation:

gj

(

x

+

vj

t

,

t

+

t

)

−

gj

(

x

,

t

)

=

Cj

(

x

,

t

),

(1) where gj represents the particle distribution function along the

j-thdirection,accordingtothefinitesetofdiscretevelocities,and

vj

t and

t arethespaceandtimeincrements,respectively.The left-hand side ofthe previous equation corresponds to the parti-cleadvection, whilethe right-hand side is thecollision operator, whichrepresentstherateofchangeof gj resultingfromcollision (i.e.theinteractionofparticles).ThecollisiontermCjismodeled withtheBhatnagar-Gross-Krook(BGK)approximation[38,39]:

Cj

(

x

,

t

)

= −

t

/

τ

[

gj

(

x

,

t

)

−

geqj

(

x

,

t

)

],

(2) where

τ

istherelaxationtime parameter,whichisrelatedtothe fluid dimensionlesskinematic viscosityandtemperature, and geq_j

is the equilibriumdistribution function, whichis related to local hydrodynamic properties[37].Forcompressibleflows,theLBM is coupledwithanentropypartialdifferentialequationtosatisfythe conservationofenergy(nonisothermalmodel)[37].Oncethe dis-tributionfunctioniscomputed,hydrodynamicflowquantities,such as flow density, velocity andinternal energy, can be determined through discrete integration of the distribution function gj [37]. Alltheother physicalquantitiescanbe determinedthrough ther-modynamicrelationshipsforanidealgas.

For simulations with a rotating geometry in time around a fixedaxis,thecomputationaldomainisdecomposedintoanouter “ground-fixed” reference frame and an inner “body-fixed” Local ReferenceFrame(LRF).Thelatterdomainischaracterizedbyagrid whichrigidlyrotateswiththerotatinggeometryinsuchawaythat norelative motionbetweenthe LRFgrid andtheenclosed geom-etryoccurs. OutsidetheLRFthefluidflow issolved usingEq.(1), whereasa modifiedversionofEq.(1),whichincludesanexternal body forcetermcorrespondingtotheinertialforce introducedby thenon-inertialrotatingLRF[40],isusedinsideit.Aclosed trans-parentinterface isusedbetweentheinner andtheouter domain inordertoconnectthetwofluidflowregions[41].

The LBM schemeissolved on a gridcomposed ofcubic volu-metric elements (Voxels), the lattice, which is automatically cre-ated by the code. Different Variable Resolution (VR) regions are specifiedbytheusertoincreasethediscretizationeffortinregions ofinterestorinregionswherehighflowgradientsareexpected.A variablegridresolutionbyafactoroftwoisallowedbetween adja-centVRs.Thesolverusesanexplicittime-marchingschemebased

Fig. 1. Sketch of the on-the-ground EPNL footprint calculation procedure.α,γandϕdenote the aircraft angle of attack, glide angle and pitch angle, respectively.

on a unitary Courant-Friedrichs-Lewy(CFL) condition. Hence, the timestep isalsovariedby afactoroftwobetweenadjacentVRs, andthe solutionin coarser VRsis updated at a lower rate com-paredtofinerVRs.Thus, abalanceddomaindecompositionbased ontheequivalentnumberofvoxelsupdatedateverytimestep,i.e. thenumber ofFine EquivalentVoxel (FEV), allowsa speed-upof thetransient flowsimulation.The surfaceofsolid bodiesis auto-maticallyfacetizedwithineachvoxelintersectingthewall geome-tryusingplanarsurfaceelements(Surfels).Fortheno-slipandslip wall boundary conditions ateach ofthese elements, a boundary scheme[42] isimplemented,basedonaparticlebounce-back pro-cessandaspecularreflectionprocess,respectively.Therefore,very complexarbitrarygeometriescanbe treatedautomaticallybythe LBMsolver, simplifyingthetediousmanual worktypically associ-atedwiththevolumemeshingstepusingotherCFDapproaches.

2.2. Far-fieldnoisecomputation

TheCAApropertiesofLBMallowtoanalyzetheacoustic near-field directly extracted from the transient flow solution. Due to the fact that the LBM is compressible andprovides an unsteady flowsolution,alongwithits lowdissipationanddispersion prop-erties[43],itisintrinsicallysuitedforaeroacousticsimulationsand allows to extract the sound pressure field directly in the near-field. In this study, a hybrid direct noise/FW-H acoustic analogy approachisadoptedtocomputethefar-fieldnoisewhileavoiding expensive computationsassociated to the necessity of accurately resolvetheacousticwavespropagationuptothefar-field.

The FW-H codeused in thiswork is partthe post-processing softwareSimuliaPowerACOUSTICS®,whichisalsousedtoperform statisticalandspectralanalysisofanyunsteadysolutiongenerated by the CFD/CAA solver PowerFLOW®. Specifically, the employed FW-H solver [44] is based on a forward-time solution [45] of Farassat’s formulation 1A[46] extended to a permeable (porous) integration surface. Such a formulation involves two surface in-tegrals, referred to as thickness (monopole) and loading (dipole) terms,andneglectsthevolumeintegral(quadrupole term),which accountsfor all the non-lineareffects in the volume outsidethe integration surface. However, by using a permeable surface en-compassing the fan-stage andpart of the fuselage, it is possible to retrieve, by means of the monopole anddipole terms, all the noise sources and non-linearities included inside the integration surfaceitself(i.e.shockwaves,turbulencemixingandpropagation effects).

Itisworthmentioningthatthenecessitytoaccuratelycapture thenear-fieldnoisepropagationfromthesourceregionuptothe FW-H integration surface is a requirement that can take advan-tageoftheintrinsiclow-dissipationandlow-dispersion properties oftheLBM schemecompared topartial differentialequation dis-cretizationschemes.

2.3. EPNLcomputation

Acomputationalprocedurebasedonpre-computedLBM/FW-H acoustic signalsisused inthiswork inorderto compute the Ef-fective Perceived Noise Level (EPNL) on the ground according to theFARPart36procedure[21].MathematicaldetailsontheEPNL toolanditsvalidationcanbefoundinCasalinoetal. [47].Forthe sakeofconciseness,onlyabriefoverviewoftheprocedureis out-linedinthefollowing.First,thepermeableFW-Happroachisused to compute the far-fieldnoise signalsand, inturn, Narrow-Band Noise (NBN) spectra (in the frequency range 50 Hz

−

10 kHz) on 300 microphones distributed over a hemisphere of 60 m radius. Suchahemisphereiscenteredaroundtheaircraftreferenceframe (denoted by xa, ya and za) andisrigidlyconnectedto theframe, assketchedinFig.1.

Then, for each microphonek on a ground carpet, andfor ev-eryflightsub-segmenti of0.5secduration,theemissionposition

(

xg

,

yg

,

zg

)

_ki oftheaircraftisdetermined (withxg, yg andzg be-ing the ground-fixed,i.e.inertial, referenceframe coordinates). A ray istracedbetweenthemicrophonek and thevehicleemission position,andits intersection Si

k withthe rotatedhemisphere de-termines the point wherethe NBNspectra are interpolatedfrom the closest microphones on the hemisphere. Similarly, the inter-section

˜

Si_k betweenthe reflected ray andthe hemisphere is also determined,andtheNBNspectraare interpolatedatthispointto accountforgroundreflection.Finally,theinterpolatedNBNspectra are projectedon thegroundcarpetby applyingspherical spread-ing,atmosphericabsorption,Dopplereffects,amplitudecorrection andgroundreflection.

3. Fan-stageconfigurationsandcomputationalsetup

3.1. Geometriesandoperatingconditions

ThegeometryconsideredinthisstudyistheNOVAlifting fuse-lage,wingandempennage(withoutengineands-duct)- courtesy ofONERA- withatotallengthof44mandasemi-spanofabout 19 m. A modified version of the Low-Noise configuration of the NASASDT,anexistingscaledfan-stageconfigurationpublicly avail-ableintheframeworkoftheAIAAFanBroadbandNoisePrediction Workshop[32], wasintegratedintothefuselage toreproducethe NOVA BLI layout,as sketchedin Fig. 2(a). In particular,the orig-inal Low-Noise SDT configuration, consisting of a 22-bladed fan and 26stator swept vanes, was firstly scaled by a factorof 3.88 to matchtheNOVAfan radius (R

=

1

.

075 m)andequippedwith aredesignednacelle,obtainedbyincreasingtheoriginalinletaxial lengthinordertomatchtheNOVABLIengineintake-fandistance (2.35 m). A sketch of the modified Low-Noise SDT engine is de-picted inFig.2(b).Thisredesignedenginegeometrywas then in-stalledintotheNOVAfuselagebyconsideringa40%buriedintake,

Fig. 2. NOVA aircraft configuration equipped with the redesigned BLI engine nacelle.

asintheoriginalNOVABLIlayout,andtiltandtoeanglesof1◦and 2

.

5◦, respectively. Finally, a s-shaped duct was designedto inte-gratethefan-stageintotheNOVAfuselagegeometry,asshownin Fig.2(c).Thiss-ductconfigurationturnedouttobethebestonein termsofinletflow separationamongthreeother differents-duct geometries,all basedon the abovementioned design parameters andcharacterizedby having tangentsurfaces tothe fuselageand nacellewalls.Sincetheprimarygoalofthisstudyistoaddressthe fannoiseimpactfortheNOVABLIlayout,twodifferent configura-tionsareinvestigated:(i)theisolatedSDTfan/OGVstagewiththe modifiednacelle(Fig.2(b))and(ii)theinstalledSDTfan/OGVstage intotheoriginalNOVAfuselagegeometry(Fig.2(c)).

Theoperatingconditionsconsideredinthisstudyare represen-tative of a take-off with power cut-back, which represents one oftherequired conditionsfornoise certification. The free-stream MachnumberisM_∞

=

0

.

25 andthestaticpressure(p_∞

=

97718 Pa)andtemperature(T_∞

=

286

.

15 K)aretakenfromthe Interna-tionalStandardAtmosphere(ISA)at1000ft.Moreover,theaircraft angleofattackis

α

=

4◦,thepitchangleis

ϕ

=

10◦ andtheglide angleis

γ

=

6◦ (note that the angle of attack is defined by the differencebetweenthe glide andpitchangles, i.e.

α

=

ϕ

−

γ

). It should be pointed out here that such angles, which define the engine incidence with respect to the free-stream together with the tilt and toe angles, are also considered for the isolated en-gine configuration. Finally, the fan angular velocity is



=

2603 RPM,corresponding tothe80%ofSDTfannominalpowerand

re-sultingintoatip MachnumberandaBlade-PassingFrequencyof

Mtip

=

0

.

8680 andBPF

=

954 Hz,respectively.

3.2. Computationalsetup

Fig.3depictssome detailsofthecomputationalsetupandthe gridadoptedinthisstudy.Forthesake ofconciseness,onlythose usedfortheBLIconfigurationarepresentedinthefollowing. Iden-tical setup and computational mesh, except for the presence of the fuselage geometry, are also used for the isolated configura-tion.The rotor andthe spinnerare encompassedbya volume of revolution that definesthe LRF, i.e.the rotatingsliding mesh re-gion used to reproduce the fan rotation. Since no primary jet is consideredinthisstudy,thecenter-bodygeometryisextendedby employingan infinitesolid cylinderdownstreamwithslip bound-ary conditions to avoid flow recirculation behind it. For the BLI configuration,azig-zagstrip of3.5cmheight,5.8cmwavelength and6.4cmamplitudewasplaced3R upstreamtheengineinletin ordertotriggertransitionanddevelopafullyturbulent boundary-layer at the fan-stagelocation, while keeping the computational effort relatively low. The FW-H integration surface (depicted in Fig. 3) usedto compute the acousticfar-field consistsoftwo re-gions:a sphericalsector aroundthe intake,anda conicalsurface intheexhaustregion.ThecenterpartoftheFW-Hsurface (cylin-der) crossesthrough thesolid wallsof thenacelle: therefore,no flowdataisextractedfromthere.Thedownstreamcapofthecone

Fig. 3. Details of the computational setup around the fan-stage: LRF, fuselage transition trip, center-body extension and FW-H surface.

Fig. 4. Sketch of the computational domain: boundary conditions and acoustic sponge.

isnot included,inordertoavoidnoisecontamination duetothe hydrodynamic pressurefluctuationsinthewakeofthe fan. How-ever,the FW-H cone extendsdownstream enough to recoverthe bypassexhaustradiationforthedirectivityanglesofinterest.

A cubic simulation volume of edge length of 690R centered aroundthe engineisused.Free-stream staticpressureand veloc-ityareprescribedon theouter boundary,andanacousticsponge approachisusedtodamptheout-goingacousticwavesand mini-mizethebackwardreflectionfromtheouterboundary(Fig.4).The acousticspongeisdefinedbytwoconcentricspheresofradius40R and150R,respectively,andcenteredaroundtheenginegeometry. Hence,thefluidkinematicviscosityisgradually increasedstarting fromitsrealvaluewithintheinnersphere,uptoanartificialvalue

two ordersofmagnitudehigheroutsidetheouter one.A symme-try plane locatedatthefuselage centerline isusedto reduce the computationalcost.

A total of 16 VR levels are employed to discretize the whole computationaldomain.ThefinestVRregioncoversthevolume be-tween thefan bladetipandthenacellecasing. Thesecond finest VR levelis used to discretize leading- andtrailing-edges of both fan and OGV vanes. The third finest VR level is set as offset of the fan, the OGV and the nacelle bypass lip. The fourth finest VR level covers the whole bypass duct and the fuselage surface upstreamtheintake(fortheBLIcase).ThefifthfinestVRlevel en-closesthepermeablesurfaceusedforFW-Hcomputations.Finally, all the other VRs, characterized by fuselageand/or engine offsets

Table 1

Gridsizeinmillionsofelementsandcomputationalcost.

Case # Voxels # FEVoxels # Surfels # FESurfels kCPUh (10 revs) Isolated engine 541.1 69.6 51.9 17.8 61.5

BLI engine 545.9 70.1 57.5 18.1 62.1

Fig. 5. Iso-surfaces ofλ2 (−75000 1/s2)color-contouredbyvelocitymagnitudedepictingtheturbulentboundary-layerpastthetransitiontrip.(Forinterpretationofthe

colorsinthefigure(s),thereaderisreferredtothewebversionofthisarticle.)

andboxes,areusedtomodeltheremaining partofthe computa-tionaldomainuptoitsboundaries.

FortheBLIconfiguration,anear-wallresolutionwhichensures ay+between100and300isprescribedonthefuselagesurfacein ordertoadequatelycapturetheboundary-layergrowth.Regarding the fan-stage region, the grid resolution employed in this study is based on the “fine” resolution successfully validated against theNASASDTbenchmarkby Gonzalez-MartinoandCasalino[31]. Theydemonstratedthecapabilityofthesolvertopredictabsolute broadband and tonal noise levels of the NASA SDT with an un-certainty of 1 dB, both at high-subsonicand transonic blade tip conditions. Such a resolution level results in a finest voxel size of 0

.

355 mm, roughly 6 voxels along the fan tip gap and 3.88 (i.e.the scalingfactor) timescoarser voxelsforsame VRsforthe presentcomputationalsetup.Duetothislastaspect,theprediction ofabsolutebroadbandandtonallevelsmightbeaffectedbyan un-certainty larger than 1 dB. Nevertheless, the goal ofthe present studyistofocusonBLIinstallationeffects,andhencetohighlight variations relative to the non-BLI configuration, rather than pre-dictingabsolutevalues.

Simulationsareperformedusinga1000coresclusterwithIntel XeonCPUE5-26972.6GHzandrequireapproximately6hoursper fanrevolution forbothisolated andBLIcases.A summary ofthe grid size andcomputational cost forboth BLI and isolated cases arereportedinTable1.

The whole fluid domain is firstly initialized with a uniform stream-wise velocity corresponding to thefree-stream conditions for a coarser simulation (

√

2 times coarser than the finer one), whichisinturnusedtoinitializeafinerresolutioncase.Hence, af-terasettlingtimeof2fanrevolutions,correspondingto0.0461sec ofphysicaltime,samplingisstarted.Acousticdataaresampledat 180kHzalong10fanrevolutions(0.2305sec).Fouriertransformed dataareevaluatedusingabandwidthof25Hz,50%window over-lapcoefficientandHanningweighting.

4. Numericalresults

Inthissection, thenumericalresultsforbothBLIandisolated fan-stageconfigurationsarepresented.First,an assessmentofthe turbulentboundary-layerbeingingestedbytheengineforthe in-stalledcaseispresented.Then,BLIinstallationeffectsfortheNOVA BLI configurationduring a take-off withpower cut-backare out-lined in terms of fan-stage velocity field, fan performances, fan bladesectionalairload,far-fieldnoisedirectivity,noisepowerlevel andon-the-groundnoisefootprint.

4.1. Fuselageboundary-layer

As mentioned, a zig-zag transition trip is employed in this studytotriggertransitionofthefuselageboundary-layerbeing in-gestedby thefan-stage. According to Van der Velden et al. [48], whoperformedLBM-basedDirect NumericalSimulations(DNS) of the flow past transition strips over flat plates, a canonical fully turbulent boundary-layer is experienced for a zig-zag strip after approximately40laminarboundary-layerthicknesses(evaluatedat thelocation ofthetrippingdevice)downstream it.Followingthis study,andconsideringtheactuallaminarboundary-layerthickness evaluatedatthetransitiontriplocation,asettlingtrip-engine dis-tanceof 1R would berequired toensure the ingestionofa fully turbulent boundary-layer into the fan-stage. Starting from these considerations,alargerandmoreconservativetrip-enginedistance of3R wasused.

Aninstantaneousviewofiso-surfacesof

λ

2 criterion[49]

color-contoured byvelocity magnitudeisdepictedin Fig.5.Thisresult qualitatively shows the presence of a turbulent boundary-layer convecting over the fuselage past the transition trip and being ingestedbythefan-stage.Flowstructuresofcoherencelength pro-portionalto thezig-zagstrip areseen mixingandcreatinglarger hairpins downstream it.Moreover, the stretch andre-orientation oftheturbulent structuresintostream-wise orientedfilaments is

Fig. 6. Locations of the fuselage boundary-layer extraction and coordinate system for measuring the wall-parallel (u) and wall-normal (w) velocity components.

Table 2

Boundary-layerthicknessδandshapefactorH atseverallocationsbetweenthezig-zagstripandtheengineintake. xe/R [-] -4.61 -4.36 -4.11 -3.86 -3.61 -3.37 -3.12 -2.87 -2.62 -2.37 δ[m] 0.1338 0.1522 0.1627 0.1702 0.1883 0.1842 0.1834 0.1929 0.1963 0.1993 H [-] 1.94 1.67 1.52 1.43 1.36 1.31 1.25 1.19 1.13 1.09

observedastheflow approachestheengineintake, asa resultof the large flow acceleration occurring in proximity of the engine intake dueto thefavorable pressure gradient induced by thefan downstream. Finally, the generation of horseshoe vortices is fur-therobservedatthejunctionbetweenthefuselageandtheengine nacelle.

Tobetterassessthepresenceofaturbulentboundary-layer be-ingingestedbythefan-stage,boundary-layerprofilesareextracted onthe fuselage at10different equispacedlocationsbetweenthe zig-zagstripandtheengineintake,asdepictedinFig.6.Such loca-tionsarereferredtotheenginecoordinatesystem(denotedbyxe,

ye andze coordinates),whoseorigincoincideswiththefancenter location, x-axis alignedwith the engine axisand positive down-stream, z-axis normalto the fuselageand directedoutwards and

y-axisdefinedby theright-hand rule.At each oftheselocations, time-averagedstream-wisevelocityprofilesu (Fig.7(a)),and time-averaged Reynolds stress profiles uu, ww and uw (Fig. 7(b), Fig. 7(c) and Fig. 7(d), respectively) are extracted from the un-steady flow solution, where u and w represent the wall-parallel and wall-normal velocity components, respectively. In Figs. 7(a) to7(d),thevelocitystatisticsandthewall-normaldistanceare re-spectivelynormalizedbythefree-streamvelocity U_∞

=

88

.

78 m/s andthelocalboundary-layerthickness

δ

(whichissummarizedfor eachlocationinTable2alongwiththeshapefactorH ).

Movingfromthe zig-zagstrip tothe engineintake, the mean stream-wise velocity (Fig. 7(a)) shows an increment of the ve-locity gradient at the wall, with the transition from laminar-like toturbulent-like profiles already occurringfor xe

/

R

≤ −

3

.

61 (i.e. within roughly 1R distance from the tripping device). This is further confirmed by the shape factor H which goes below the threshold value of 1

.

3

−

1

.

4 starting from xe

/

R

≤ −

3

.

61, as expected for a turbulent boundary-layer [50]. Moreover, the outer parts of the different mean stream-wise velocity profiles tend to collapse on top of each other,with edge-velocity values slightlyexceedingthefree-streamone,forxe

/

R

≤ −

3

.

61. Contrar-ily,themeanstream-wise velocityprofiles show increasing edge-velocities at more downstream locations (up to approximately 1.75U_∞ for xe

/

R

= −

2

.

37), as a consequence of the increas-ing favorable pressure gradient inthe stream-wise direction. The

stream-wise(Fig.7(b)),wall-normal(Fig.7(c))andshear(Fig.7(d)) Reynoldsstressesshowarapidreductionofthepeakvaluewithin thefirst5stream-wisestations(

−

4

.

61

≤

xe

/

R

≤ −

3

.

61),i.e.within 1R from the zig-zag strip, where the settling of the turbulence enforced by the zig-zag strip is expected [48]. For more down-stream positions, a weaker reduction of the turbulent levels is furtherobservedfortheuuanduwcomponents.Contrarily,the wall-normalcomponentshowsnoreductionoftheturbulentlevels with convergingprofiles forlocationsclose to the engine intake. This might be relatedto the stretching andre-orientation of the boundary-layervortices inthe stream-wisedirection, witha con-sequent re-distribution of the turbulent kinetic energy from the

uuto the ww componentsintheoverall Reynoldsstresses en-ergybudget.Finally,itisinterestingtopointoutthattheReynolds stress levels, extractedinproximity ofthe engine intake(xe

/

R

=

−

2

.

37),areconsistentwiththoseofacanonicaldeveloped turbu-lentboundarylayer,whosevaluesatawall-normaldistanceof0.2

δ

areapproximatelyuu

/

U_∞2

=

4

.

5

·

10−3,ww

/

U2_∞

=

1

.

6

·

10−3and

−

uw

/

U2

∞

=

1

.

3

·

10−3 accordingtoKlebanoff[51]. 4.2. Fan-stagevelocityfield

Fig.8depictsaschematic descriptionofaplanenormaltothe fuselageandpassingthroughtheengineaxis,anddifferent cross-flowplanesalongtheinlet,interstageandbypassexhaustsections, whichareusedinthefollowingtoextractthefan-stageaxial(i.e. aligned withengine axis) andin-plane (i.e. perpendicular to the engine axis) velocity fields along the engine. In this sketch, the

xe, ye andze coordinatesrepresentthoseoftheenginecoordinate systempreviouslydefined.

Fig. 9 showsthe instantaneous axial velocity field forthe BLI caseextractedonaplanenormaltothefuselagesurfaceand pass-ing through theengine axis at differenttime instants. The same quantity is depicted on the same plane for the isolated case in Fig.10.TheBLIconfigurationshowsaflowaccelerationatthe in-take section, dueto thereducedinletthroatarea oftheinstalled configuration compared to the isolated one, and the presence of turbulence impinging thefan rotor. Such turbulent structuresare connectedtothefuselageturbulentboundary-layerbeingingested

Fig. 7. Time-average stream-wise velocity (a) and Reynolds stress (b, c and d) profiles at several locations between the zig-zag strip and the engine intake.

Fig. 8. Sketch of the planes used to extract the fan-stage velocity fields across the engine.

bythefan-stage,aswellastotheflowseparationoccurringat ap-proximately 60% of the s-duct length, the latter induced by the adversepressure gradientdueto therapidincrease ofthe intake cross-sectional area. Conversely, the isolated configuration shows

a ratheruniform velocityfield upstream the fan rotor, exceptfor the first25% ofits extension, inwhichthe flowtends to recover fromitsinitialmisalignmentwiththeaxisengineduetothe pres-enceofnon-zeroangleofattack,andtiltandtoeangles.Upstream

Fig. 9. Instantaneous axial velocity field on a plane normal to the fuselage surface and passing through the engine axis at different time instants, BLI engine.

Fig. 10. Instantaneous axial velocity field on a plane normal to the fuselage surface and passing through the engine axis at different time instants, isolated engine.

traveling wavesare observedalong theintake fortheBLIengine, unlike the isolated one. The BLI configuration shows a thicker boundary-layeron the intake wall opposite to the fuselage com-paredto theisolated case, which alsotends to separate in

prox-imityofthewallbeneaththefan.Theinstantaneousaxialvelocity field furthershowstheoccurrenceoftwodifferentfanwake/OGV interactionmechanismsbetweenthefuselageandthenacellesides fortheBLIcase,whereastheisolatedcaseshowsthetypical

rotor-Fig. 11. Time-averaged velocity field at different sections upstream the fan, isolated engine.

statorinteractionmechanismwiththe fanviscous wakes imping-ing oneach statorvanes at theblade-passing frequency. The BLI configurationshowsanon-axisymmetricvelocityfieldbetweenthe rotorandthe stator,withmuchhighervelocities observedinthe region opposite to the fuselage side with respect to the isolated one.Inthissameregion,theBLIcasefurthershowsstronger turbu-lentstructuresthanthoseconvectingonthefuselageside.Similar considerationscanbe madeforthefluidregions downstreamthe statoraswell.

4.2.1. Inletflowfield

Figs.11and12show thetime-averagedaxialandin-plane ve-locitycomponentsoncross-flowdisksupstream thefanplanefor

the BLIand isolated configurations,respectively. For thein-plane velocity plots, the in-plane velocity vectors are also shown. In these upstream-looking-downstream views the fan blades rotate counter-clockwise asindicated by the blackcircular arrow.Three different stream-wise locations (i.e. xe

/

R

= −

1

.

95, xe

/

R

= −

1

.

05 and xe

/

R

= −

0

.

37), respectively corresponding to 15%, 55% and 85% of the inlet axial length, are considered. Moreover, the con-vention adopted to describe the fan blade azimuth angle is also reported.According to thisconvention, the fan bladesweeps the BLIareaforazimuthalanglesbetween45◦and180◦ (Fig.12).

The isolated case showsa moderate level ofdistortion across the engine intake, where the free-stream misalignment with the engineaxis(dueto thenon-zeroangleofattack,andtiltandtoe

Fig. 12. Time-averaged velocity field at different sections upstream the fan, BLI engine.

angles) is responsible of the non-symmetric acceleration of the flowaroundtheintakelip(Fig.11(a))andtheariseofanin-plane velocitycomponentdirectedinboard(Fig.11(b)).Thevelocityfield uniformityisthencompletelyre-establishedwithinthefirsthalfof theintake extension,where theflowapproaching thefan section showsaquiteconstant meanaxialvelocity(Figs.11(c)and11(e)) andamoderatein-planevelocity directedoutwardsalongthe ra-dialdirection, asa consequenceofthe slightlydivergent inlet ge-ometry (Fig. 11(d)) andthe presenceof the spinnerdownstream (Fig.11(f)).

Regarding the BLI configuration, the reduced engine intake frontal area yields to an increase of the axial (Fig. 12(a)) and in-plane inward directed (Fig. 12(b)) velocity components

com-pared to the isolated case, the latter primarily showing a radial pattern withthe highest magnitudevalues occurring around the junction between the fuselage and the engine nacelle. For fur-ther downstream sections, the meanflow exhibitshigh levels of non-uniformity anddistortionintermsofbothaxialandin-plane velocities.Morespecifically,theaxialcomponentshowstwonearly symmetric recirculation/low-velocity regionson thefuselage side. Such flow separation areas are already visible at 55% of the in-let length(Fig. 12(c))andextendfurtherdownstream (Fig. 12(e)) accordingto a nearly symmetricalpatternaroundthe inletplane of symmetry (around 120◦ in the fan blade azimuth). A nearly symmetricalpatternaround120◦ ofthefanbladeazimuthisalso observable forthe in-plane velocity atxe

/

R

= −

1

.

05 (Fig. 12(d)),

Fig. 13. Phase-locked averageofaxial(a)andtangential(b)velocitycomponentsandstandarddeviation(SD)ofaxial(c)andin-plane(d)velocitycomponentsonaninterstage diskatxe/R =0.36,BLIengine.

which shows the presence of a strong secondary flow directed fromtheengine nacelleside towards thefuselage one, asa con-sequence of the centrifugal pressure gradient due to the s-duct geometric curvature. This transverse secondary flow is also ob-servedfortheinletsectionatxe

/

R

= −

0

.

37,whichfurthershows anentrainmentmotionimpartedby therotor bladeson the low-velocityflowapproachingthefanface(Fig.12(f)),i.e.forr

/

R

>

0

.

9 andazimuthalangleswithin45◦-90◦and135◦-210◦,respectively.

Itshould be recalled that the currentNOVABLI configuration representsa ratheridealized BLI-layout,dueto the relativeshort s-ductlength(2.35m) andthelarge portionofintake embedded intothefuselage (40%).Amoreconservative sizing,e.g. basedon a longer s-duct length and a lower percentage of buried intake, wouldhelp in reducing the amountof flow separation occurring onthes-duct,aswellasofflowdistortionatthefanplane.Thisat thecostofadditionalmass,dragandinletfrictionlosses.

4.2.2. Interstageflowfield

Figs. 13 and14 show upstream-looking-downstream viewsof thevelocity field foraninterstage sectionbetweenthe rotor and thestator(xe

/

R

=

0

.

36)fortheBLIandisolatedconfigurations, re-spectively.Morespecifically,thevelocity fieldisdecomposedinto phase-lockedaverage ofaxial (Figs.13(a)and 14(a)) and tangen-tial(Figs.13(b)and14(b))velocitycomponents,andstandard de-viation of axial (Figs. 13(c) and 14(c)) and in-plane (Figs. 13(d) and14(d)) velocity components. It is worth mentioning that the phase-lockedaveragecontourplotshighlightthepresenceof peri-odicnon-uniformitiesinthemeanflowassociated totherotating

fan blades (i.e.viscous bladewakes andtip-vortices),which gen-erate tonal noise at BPF and its harmonics when they interact with the stator vanes. Instead, the phase-locked standard devia-tioncontourplotscanbeusedtoexaminethepresenceofrandom fluctuations inthe flow (i.e. turbulence), which represent poten-tialsources ofbroadband noisewhen theyimpinge onthe stator surfaces.

Contrarilyto theisolated configuration,wherethe twotypical expected flow regions downstream the fan, i.e. the “viscous” re-gionassociatedtobothrotorbladewakesandtip-vorticesandthe “potential” flow region outside the viscous ones [52], are clearly defined throughout the radial coordinate (Figs. 14(a) and 14(b)), the same regions are visible only up to r

/

R

=

0

.

8 of the span-wise coordinate forthe BLI case(Figs. 13(a) and 13(b)). For the isolated configuration,the axial velocity inthe potential flow re-gions isratheruniform(Fig.14(a))along thespan-wisedirection, whereasthetangential(Fig.14(b))onestendtodecreaseforlarger

r

/

R. Moreover, for a given radial position, the axial velocity is lower in theviscous regions of theflow compared tothe poten-tialones, whereasthetangential onesresulttobehigher.Similar trendscanalsobefoundfortheBLIconfiguration,althoughsome additional considerations need to be outlined. First, the phase-locked average axial velocity shows a non-uniform distribution alongthe azimuthalcoordinate,withhighervalueswithin 0◦-45◦ and135◦-360◦andlowervaluesintheexplementarycircular sec-torcomparedtotheisolatedcase. Thesametrendalsooccursfor thetangential componentwithin the first55% oftheradial coor-dinate,whereasanoppositesituation isobservedfor0

.

55

<

r

/

R

<

Fig. 14. Phase-locked averageofaxial(a)andtangential(b)velocitycomponentsandstandarddeviation(SD)ofaxial(c)andin-plane(d)velocitycomponentsonaninterstage diskatxe/R =0.36,isolatedengine.

highertangentialvelocities abover

/

R

=

0

.

9 comparedto the iso-latedone. Regarding the turbulent fluctuations, the isolated case showsquiteuniformstandarddeviationcontours,withthelargest velocity perturbations occurring within the viscous wake and at the tip for both axial (Fig. 14(c)) and in-plane velocity compo-nents (Fig. 14(d)). Contrarily, the BLI configuration shows larger levels of fluctuations and flow non-uniformity for both compo-nents(Figs.13(c)and13(d),respectively).Morespecifically, veloc-ityfluctuationsroughlythreetimeshigherareobservedwithinthe outer30%oftheradialcoordinateandforazimuthalposition com-prised between 0◦-60◦ and 180◦-360◦, respectively. A secondary diffusedregion ofhighturbulence levels isalso presentbetween 50◦-180◦,i.e.incorrespondenceofthefuselageBLIarea.Moreover, thickerviscouswakesare observedfortheBLIconfiguration com-paredtotheisolatedone,especiallyforazimuthalanglesbetween 180◦and360◦.

Overall,theBLIinterstageflowfieldischaracterizedbyless ax-isymmetricandcoherenttangentialvelocity variationsandhigher levels ofin-plane velocity fluctuations. Suchvelocity components are thought to be more important in the generation of the ro-tor/stator interaction noise [52]. This type of noise is associated totheunsteadyloadingonthestatorvanes,whichisgeneratedby fluctuationsof theflow velocity component normalto the stator surface.Inview ofthis, broadbandnoise isexpectedtodominate more thefar-field noise spectrumfor the BLIconfiguration com-paredtotheisolated case. Thedetailedaeroacousticanalysiswill beperformedinSecs.4.5and4.6toprooftheassumption.

4.2.3. Bypassexhaustflowfield

Toconcludetheanalysisofthefan-stagevelocityfield,Figs.15

and 16 show upstream-looking-downstream views of the veloc-ityfield ona bypass exhausttransversesection (xe

/

R

=

1

.

59)for the BLIandisolatedconfigurations,respectively. Again,the veloc-ity field is presented in terms of phase-locked average of axial (Figs. 15(a) and16(a)) and tangential (Figs. 15(b) and 16(b)) ve-locity components, and standard deviation of axial (Figs. 15(c) and 16(c)) and in-plane (Figs. 15(d) and 16(d)) velocity compo-nents.

Asalreadypointedout,theBLIcaseshowsanon-axisymmetric phase-locked average axial velocity field along the azimuthal co-ordinate downstreamthe stator(Fig.16(a)),withthe highestand lowest values respectivelytaking placeon the nacelle side up to

r

/

R

=

0

.

7 and above r

/

R

=

0

.

9. Contrarily, the isolated configu-ration presents the expected axisymmetric axial velocity pattern characterized by the viscous wakes being convected from each stator vane (Fig. 16(a)). The BLI configuration further shows a non-uniform andlower flow swirlrecoverydownstream the sta-tor(Fig.15(b))comparedtotheisolatedone(Fig.16(b)),withhigh valuesofthe tangentialvelocity still persistingalong mostofthe outerpartofthebypassexhaust(r

/

R

>

0

.

9)andfortheazimuthal sector between225◦ and315◦. Concerningthe turbulence levels in the bypass exhaust, the isolated case shows again quite ax-isymmetric standard deviation contours,withthe largest velocity fluctuationsoccurringabove 80%oftheradialcoordinateforboth theaxial(Fig.16(c))andin-plane(Fig.16(d))velocitycomponents. Finally,similarlytowhatobservedfortheinterstagevelocityfield,

Fig. 15. Phase-locked averageofaxial(a)andtangential(b)velocitycomponentsandstandarddeviation(SD)ofaxial(c)andin-plane(d)velocitycomponentsonadiskin thebypassexhaustatxe/R =1.59,BLIengine.

Table 3

Fanpressureratio,fanisentropicefficiencyandrelativedifference be-tweenisolatedandBLIengines.

Parameter Isolated engine BLI engine Relative difference FPR [-] 1.27 1.24 -2.5%

ηf [-] 0.917 0.836 -9.2%

theBLIconfigurationmanifestslargevaluesofvelocityfluctuations forboththeaxialandin-planecomponents(Figs.15(c)and15(d), respectively), with levels approximately three times higher than thoseoftheisolatedcaseforr

/

R

>

0

.

7 andfortheazimuthal sec-torswithin0◦-60◦ and180◦-360◦.

4.3.Fanperformances

Table3showsBLIinstallationeffects onthe fanperformances intermsoffanpressureratioFPRandisentropicefficiency

η

f.As a consequence of the highly distorted flow and the ingestion of turbulence,areductiontheFPRby2

.

5% andoftheisentropic effi-ciencyby9

.

2% isobservedwhenthesamefanisoperatedatsame RPM in the BLI layout with respect to the conventional non-BLI case.

4.4.Fanbladesectionairload

Fig. 17 depicts the sectional thrust coefficient time-history

cTV

˜

2 at6 different span-wise sections uniformlydistributed be-tween 45% and 95% ofthe blade span. For each section and fan

bladeazimuthalposition,thecTV

˜

2coefficientiscomputedby inte-grationoftheairfoilpressuredistributionp overtheairfoilcontour



usingthefollowingformula:

cTV

˜

2

= −



 pn

·

ixed



1 2

ρ

∞V 2

_

c

˜

V2

= −

2

ρ

_∞a2 ∞



c



 pn

·

ixed



(3)

where n is the outward-pointing normal to the airfoil contour,

ixe is the engine axis unit vector (positivewhen directed down-stream)andd



istheinfinitesimalairfoilcontourelement. More-over, V

˜

=

V

/

a_∞ is the fan blade sectional velocity (V

= 

r) at the radial coordinate r normalized by the free-stream speed of sounda_∞,

ρ

∞isthefree-streamdensityand



c istheairfoilchord. Notethat the lineintegral inEq. (3) is computedclockwisewith respecttotheairfoilcontour



.Asexpected,theisolated configu-rationshowsalmostconstanttime-historiesofthesectionalthrust for each span-wise location, witha meanvalue increment mov-ing frominboard (Fig. 17(a)) to outboard (Fig. 17(f)) sections of the fan blade. Onlythe outer section at r

/

R

=

0

.

95 shows some weak unsteadiness due to the interaction between the blade tip and theboundary-layer developing along the inlet wall. Contrar-ily, theBLI configurationshows a low-frequencythrust unsteadi-ness (predominantly1/rev) forinboardblade sections(Figs.17(a) to17(c)),asaconsequenceofthestrongmeanflowdistortion.For thesesections,anincrementofthesectionalthrustisobserved ap-proximately between60◦ and180◦ (i.e.in correspondenceofthe BLIarea) comparedto theisolated case, whereas lowervalues of

Fig. 16. Phase-locked averageofaxial(a)andtangential(b)velocitycomponentsandstandarddeviation(SD)ofaxial(c)andin-plane(d)velocitycomponentsonadiskin thebypassexhaustatxe/R =1.59,isolatedengine.

thecTV

˜

2 coefficientareobservedelsewhere.Astheradial coordi-nateincreases,moreintenseandimpulsiveunsteadiness,aswellas lowermeanvaluesofthesectionalthrustareobserved(Figs.17(d) to17(f))comparedtotheisolatedcase,duetotheingestionof tur-bulenceassociatedtotheflowseparationoccurringonthenacelle andfuselagesides,respectively.

All the aforementioned thrust fluctuations represent an addi-tionalsourceofnoise.Amongthem,thoseoccurringontheouter partoftheblade areexpectedtocontribute moreto thefar-field noise,duetothehigherlevelsofunsteadinessandMachnumbers comparedtothoseinbladeinboardregions.

4.5. Far-fieldnoisedirectivity

Fig. 18 depicts the microphone array used in this study for far-field noise computations. It is composed by 7 even meridian arcsof10mradius,centered aroundthefan centerandcovering a semi-spherical surface. Each arc is characterized by 15 micro-phonesdistributedevery10◦,withdirectivityanglesrangingfrom 20◦upstreamtheengineto160◦downstreamit.

As mentioned in Sec. 3.2, the far-field noise is computed by integrationoftheFW-Hequationon apermeablesurface encom-passing the engine. Since the FW-H formulationadopted in this workdoesnotincludethevolumeintegral,spurioussignalsmight arisewhenthepermeablesurfaceisintersectedbyturbulence(i.e. thefuselageturbulent boundary-layer)[53].The presenceofsuch spuriouseffectshasbeenassessed(fortheBLIcase)bycomparing far-fieldnoisepredictionsfromthewholepermeableFW-H surface

to those obtained by removing that portion of the surface inter-sected bythefuselage boundary-layer.Itturned out thatthetwo different approachesprovided almost identical results within the directivity angles of interest, thus allowing the useof the whole FW-H permeable surface (even forthe BLI configuration) for the far-fieldnoisecomputations.

A spectral representation(PowerSpectral Density, PSD) ofthe far-fieldnoisedirectivitynormalizedbytheBPFisshowninFig.19

and Fig. 20 forthe BLI and isolated cases, respectively. In addi-tion, far-field noise differences between such configurations are depicted in Fig. 21. For the sake of conciseness,only the results forthearcsat

φ

=

0◦ (groundarc)and

φ

=

90◦ (sidelinearc) are shown infollowing.For both theBLI andisolated configurations, thenoiseisradiatedmostefficientlydownstreamtheengine. How-ever, besides such a similarity, the two examined configurations show quite different results. The isolated engine presents both broadband andtonalnoise contributions,withthelattershowing distinct peaks atmultiples ofthe blade-passingfrequency down-streamtheengine(uptoBPF-3).Conversely,tonesatharmonicsof theBPFdonotemergewithrespecttobroadbandlevelsfortheBLI case. Haystacked peaks are found around BPF-1 for downstream directivity angles, asbetter highlighted in Fig. 22for theground arcat140◦ and150◦ observerangles.Such peaksarebelievedto be generated by the correlated unsteady airloadson neighboring blades, andassociatedtoboththe stream-wiseelongatedvortices coming fromthefuselageturbulent boundary-layerandthe large eddiesbeingshedfromthes-ductwall,whichmightbe cut mul-tiple timesbysuccessiveblades (asimilarphenomenon hasbeen alreadyexperimentallyobservedbyMurrayetal. [20] and

Alexan-Fig. 17. Phase-locked fan blade sectional thrust coefficient time-histories at different span-wise locations.

deretal. [54] forarotorcaseingesting aplanarboundary-layer). FortheBLIcase,broadbandlevelsturnouttobefrom10to20dB overallhigherthanthoserelatedto theisolatedconfiguration,for mostof the frequencies and directivity angles considered.

More-over, the BLI layout appears to be as noisy as the isolated one onthesidelinearcfordirectivityanglesaround 90◦ and frequen-cies higherthan BPF-2(Fig. 21(b)),orquieter by 5-10dB on the groundarc (Fig. 21(a)). This last point might be relatedto some

Fig. 18. Sketch of the microphone array used for far-field noise computations. The angleφdenotes the position of each meridian arc.

Fig. 19. Far-field noise directivity on ground (a) and sideline (b) arcs of 10 m radius and centered around the fan, BLI engine.

Fig. 20. Far-field noise directivity on ground (a) and sideline (b) arcs of 10 m radius and centered around the fan, isolated engine.

noiseshielding effectsintroducedby thepartialplacementofthe engineintothefuselage.

4.6. Noisepowerlevel

In Fig.23 one-side narrow-band source Power Level (PWL) is presentedforboth BLIandisolated configuration(Fig. 23(a)). For convenience, thedifference betweenthe powerlevels of thetwo casesis alsoshown inFig. 23(b).Inthis study,the PWLis com-puted by integrationof the PSDsover the semi-spherical surface portion corresponding to the aforementioned microphone array,

usingthefollowingformula:

PWL

(

f

)

=

θmax



θmin φ



max φmin R2_ssin

(θ )

[

1

+

M∞cos

(θ )

]

2_PSD

₍

_f

_{, θ )}

2

ρ

_∞a_∞ d

φ

d

θ

(4) where f is the frequency, Rs is the hemisphere radius,

φ

is the meridian arcangularposition (defined asinFig. 18) andvarying from

φ

min

=

0◦ to

φ

max

=

180◦,whereas

θ

isthe directivityangle varyingfrom

θ

min

=

20◦to

θ

max

=

160◦.Moreover,M∞isthe

free-Fig. 21. Far-field noise directivity on ground (a) and sideline (b) arcs of 10 m radius and centered around the fan, difference between BLI and isolated engines.

Fig. 22. Far-field noise spectra on ground arc at downstream directivity angles: 140◦(a) and 150◦(b), BLI engine.

Fig. 23. Source Power Level: BLI and isolated engines (a) and their difference (b).

stream Mach number, and

ρ

∞ and a_∞ are the ambient density andspeedofsound,respectively.

As already observed, broadband component dominates the powerlevelspectrum forthe BLIconfiguration,asaconsequence ofthe lessaxisymmetric rotor wakes andhigherlevels of turbu-lence impinging on the stator vanes. No tones at harmonics of

the BPF are observed, and only weak haystacked peaks around BPF-1 are found. As mentioned before, such broadened peaks might be connected to blade-to-blade unsteady airloads correla-tionassociatedtothesimultaneous impingementoflargevortical structures withmultipleneighboringfanblades.Finally,regarding theisolated engine,thePWLpresentsdistincttonesatthefirst 3

Fig. 24. EPNL on a plane 1.2 m above the ground during a take-off flight path with power cut-back: BLI engine (a), isolated engine (b) and their difference (c).

blade-passingfrequencies,andlowerbroadbandsourcepower lev-els,from4-5dB athighfrequencies upto 15-18dBatlow ones, withrespecttotheBLIconfiguration.

4.7. Effectiveperceivednoiselevel

TheanalysisofBLIinstallationeffectsonfannoiseisconcluded by investigating the Effective Perceived Noise Level (EPNL) foot-print during a take-off flight path with power cut-back. In the presentstudy,theEPNL iscomputed (asbriefly described in2.3) accordingto the FARprocedure [21] byconsidering a rectilinear flightpathof2kmlength(from xg

= −

1000 mtoxg

=

1000 m), 210m altitudedifference (from zg

=

200 mto zg

=

410 m) and 23.7 sec duration. The free-stream conditions applied are those presentedinSec.3.1andanatmosphericrelativehumidityof70% isconsidered.Figs.24(a)and24(b) showtheEPNLonaplane 1.2 mabove thegroundoveran area of2.5x2.5km2 forthe BLIand isolated cases,respectively,whereas theEPNL difference between thetwo configurations isshownin Fig.24(c). Sincein thisstudy simulationsareperformedbyemployingasymmetryplanelocated at the centerline of the aircraft, the EPNL footprint is computed by mirroringthe sourceofnoise (i.e.thenoise hemisphere)with respecttothesameplane(locatedat yg

=

0 m) priortothe pro-jectionoftheNBNspectraonthegroundmicrophonecarpet.The twocontributionsofthemirroredhemispheresareadded incoher-ently.

Asexpected,duetothetake-offflightpathconsidered,boththe EPNLcontoursshowthatthehighestnoiselevelstakeplaceatthe beginning ofthe flight trajectory, i.e.whenthe distancebetween thesourceofnoiseandthegroundistheminimum.Then,forboth cases,theon-the-groundnoiselevelsgraduallydecreasesalongthe flightdirectionasthenoisesourcealtitudeincreases.Nevertheless, thetwoconfigurationsshowquitedifferentEPNLlevelsand direc-tivitypatterns.First,theEPNLmapischaracterizedbyapeakvalue of108EPNdBforthe BLIcase,which isroughly 10EPNdBlarger thanthatoftheisolatedone(98EPNdB).Secondarily,theBLI con-figurationshowsthatthenoiseontheground,duringthetake-off maneuver, ismainly radiatedalong the sidelineand downstream directions,contrarilytotheisolatedcasewhosetake-offnoise ra-diation is predominantlydirected upstream. Overall, the BLIcase turns outto befrom4EPNdB(frontside)to 18EPNdB(aftside) nosier than the isolated one for the operating condition hereof considered.

5. Conclusionsandfutureoutlook

Forthefirsttime,ahigh-fidelity CFD/CAAsimulationofa full-scale aircraft geometry comprehensive of a BLI fan/OGV stage was performed. A modified version of the Low-Noise configura-tion of the NASA SDT fan-stage was embedded into the ONERA NOVA fuselage in orderto reproduce the NOVA BLI aircraft con-figuration. The numerical flow solution was obtained by solving

the explicit, transient and compressible lattice-Boltzmann equa-tion implemented in the high-fidelity CFD/CAA solver Simulia PowerFLOW®. The acoustic far-field was computed by using the Ffowcs-Williams & Hawkings integral solution applied to a per-meablesurface encompassing thefan-stage. Installation effectsof theBLI configuration,which isnot optimal, were investigated by comparisonwithanisolatedsetupofthemodifiedLow-NoiseSDT fan-stage geometry in terms of fan-stage velocity field, fan per-formances,fan bladeunsteady airload,far-fieldnoiseand on-the-groundnoisefootprint.Allsimulationswereperformedforan op-eratingconditionrepresentativeofatake-offwithpowercut-back. Theanalysisofthefan-stagevelocityfieldshowedthat,forthe considered operating condition and geometry, the embedded BLI fan-stagecauseshighlevelsofmeanflowdistortionandflow sep-arationatapproximately60%ofthes-ductlengthandinproximity ofthefanplaneonthenacelleside.Themeanflowdistortionwas found to be responsible of a low-frequency periodic variation of thefan blade sectional thrust forinboardblade sections.The in-gestion of turbulence, associated to the flow separation on the s-ductandinletwalls, led to highlevelsof unsteadiness,aswell asto a deficitin thrust generation in outboard blade regions. A reductionof theFPR by 2

.

5% andofthe fan isentropicefficiency by 9

.

2% was observed forthe BLI configurationcompared to the isolatedone. Inaddition,theBLI engineinterstageflow field was characterizedbylessaxisymmetricandcoherenttangential veloc-ity variations associated to the rotor wake, andhigher levels of in-plane velocity fluctuations, compared to the isolated one. Fi-nally,theBLIconfigurationshowedanon-uniformandlowerswirl recoverydownstream thestatorcompared tothe isolatedengine, aswellashigherlevelofturbulentfluctuations.

Far-fieldnoisedirectivity predictionsrevealedthatthenoiseis radiatedmostefficientlydownstreamtheenginefortheBLIlayout, asalsoobservedfortheisolatedengine.However,whilethelatter manifestedbothbroadbandandtonalnoisecontributions,notones clearlyemergedwithrespecttobroadbandlevelsfortheBLIcase. WeakhaystackedpeakswerefoundaroundBPF-1fordownstream observerangles.Suchpeaksmightbeconnectedtoblade-to-blade unsteady airloadscorrelation associated to the simultaneous im-pingementoflarge vorticalstructuresseparatingfrom thes-duct wall andfuselage turbulent boundary-layer withmultiple neigh-boringfan blades. Overall, theBLI configuration showedfrom10 to20dB higherbroadbandlevels inthe far-fieldcompared those relatedto the isolated configuration, for mostof the frequencies anddirectivityanglesconsidered.TheBLIlayoutresultedtobeby 5-10dBquieterthantheisolatedoneonthegroundarc,for direc-tions nearly perpendicularto the engineaxis andforfrequencies higherthan BPF-2,asa resultofsome airframe shielding related tothepartialplacementoftheengineinsidethefuselage.Finally, theBLIconfigurationisfoundtohaveadetrimentalimpacton cu-mulativenoiselevelsduring atake-offflightwithpowercut-back upto18EPNdB.

Itshouldberecalledthatthefanstageconsideredinthisstudy isdesignedforisolatedengines,thusitisnotsuitableforaoperate inhighly distortedflows such asin BLIconfigurations. Moreover, thecurrent NOVA BLIconfiguration represents a ratheridealized BLI-layout, due to the relative short s-duct length and the large portionof intake area embedded into the fuselage. A more con-servativesizing, e.g.basedon alonger s-ductlengthanda lower percentageofburiedintake,wouldreducetheamountofflow sep-aration and distortion atthe fan section, with expected benefits onfan-stageoperabilityandnoiseemissions,howeveratthe addi-tionalcostofhighernacellemass,dragandinletfrictionlosses.

Asfutureoutlook,simulationsbyemploying flow-control tech-niques will be performed for the BLI configuration in order to mitigatetheflow separationoverthe s-ductwallandtherelated benefitsintermsofnoiseemissionwillbeassessed.

Declarationofcompetinginterest

Theauthorsdeclarenoconflictofinterestregardingthisarticle.

Acknowledgements

Thisprojectissupported bytheprojectARTEM(Aircraftnoise ReductionTechnologies andrelatedEnvironmental iMpact)which hasreceivedfundingfromtheEuropeanUnion’sHorizon2020 re-search and innovation programmeundergrant No. 769 350.The authorswouldfurtherliketothanktheDepartmentof Aerodynam-ics,AeroelasticityandAeroacousticsofONERAfortheprovisionof theNOVAfuselagegeometry,withwingandempennage,thatwas usedinthisstudy.

References

[1]S.Kallas,M.Geoghegan-Quinn,M.Darecki,C.Edelstenne,T.Enders,E. Fernan-dez,P.Hartman,Flightpath2050Europe’s VisionforAviation,Reportofthe HighLevelGrouponAviationResearch,ReportNo.EUR98,European Commis-sion,Brussels,Belgium,2011.

[2]C.E.Hughes, NASA CollaborativeResearchon the UltraHighBypass Engine CycleandPotentialBenefitsforNoise,Performance,andEmissions,Technical MemorandumTM-2013-216345,NASA,2013.

[3]E.M.Greitzer,P.Bonnefoy,E.DelaRosaBlanco,C.Dorbian,M.Drela,D.Hall, R. Hansman,J. Hileman,R.Liebeck, J.Levegren,etal.,N+3AircraftConcept DesignsandTradeStudies.Volume1,NasaCR-2010-216794/vol2,NASAGleen ResearchCenter,Cleveland,Ohio,2010.

[4]C.Heykena,L.Savoni,J.Friedrichs,R.Rudnik,Engineairframeintegration sen-sitivitiesforaSTOLcommercialaircraftconceptwithover-the-wingmounted UHBR-turbofans,in:ProceedingsofMontreal2018GlobalPowerand Propul-sionForum,2018.

[5]L.Wiart,O.Atinault,R.Grenon,B.Paluch,D.Hue,DevelopmentofNOVA air-craftconfigurationsforlargeengineintegrationstudies,in:33rdAIAAApplied AerodynamicsConference,2015,p. 2254.

[6]L.Wiart,O.Atinault,J.-C.Boniface,R.Barrier,Aeropropulsiveperformance anal-ysisoftheNOVAconfigurations,in:30thCongressoftheInternationalCouncil oftheAerodynamicalSciences,2016.

[7]A.Plas,D.Crichton,M.Sargeant,T.Hynes,E.Greitzer,C.Hall,V.Madani, Per-formanceofaboundarylayeringesting(BLI)propulsionsystem,in:45thAIAA AerospaceSciencesMeetingandExhibit,2007,p. 450.

[8]M. Drela, Power balance in aerodynamic flows, AIAA J. 47 (7) (2009) 1761–1771.

[9]L. Hardin,G. Tillman,O. Sharma,J.Berton,D.Arend,Aircraft systemstudy ofboundary layeringesting propulsion,in:48th AIAA/ASME/SAE/ASEEJoint PropulsionConference&Exhibit,2012,p. 3993.

[10]G. Carrier,O.Atinault,R. Grenon,C.Verbecke,Numericalandexperimental aerodynamicinvestigationsofboundarylayeringestionforimproving propul-sion efficiency offutureairtransport, in:31stAIAA AppliedAerodynamics Conference,2013,p. 2406.

[11]K.M.Sabo,M.Drela,Benefitsofboundarylayeringestionpropulsion,in:53rd AIAAAerospaceSciencesMeeting,2015,p. 1667.

[12]P.Lv,D.Ragni,T.Hartuc,L.Veldhuis,A.Rao,Experimentalinvestigationofthe flowmechanismsassociatedwithawake-ingestingpropulsor,AIAAJ.(2016) 1332–1342.

[13]P. Lv, A.G. Rao, D. Ragni, L. Veldhuis, Performance analysis of wake and boundary-layeringestionforaircraftdesign,J.Aircr.53 (5)(2016)1517–1526.

[14]A.Uranga,M.Drela,E.M.Greitzer,D.K.Hall,N.A.Titchener,M.K.Lieu,N.M.Siu, C.Casses,A.C.Huang,G.M.Gatlin,etal.,Boundarylayeringestionbenefitof theD8transportaircraft,AIAAJ.(2017)3693–3708.

[15]A.Anabtawi,R.Blackwelder,P.Lissaman,R.Liebeck,Anexperimental investiga-tionofboundarylayeringestioninadiffusingS-ductwithandwithoutpassive flowcontrol,in:37thAerospaceSciencesMeetingandExhibit,1999,p. 739.

[16]S.Gorton,L.Owens,L.Jenkins,B.Allan,E.Schuster,Activeflowcontrolona boundary-layer-ingestinginlet,in:42ndAIAAAerospaceSciencesMeetingand Exhibit,2004,p. 1203.

[17]L.R.Owens,B.G.Allan,S.A.Gorton,Boundary-layer-ingestinginletflowcontrol, J.Aircr.45 (4)(2008)1431–1440.

[18]J.J. Defoe,Z.S. Spakovszky, Effectsofboundary-layeringestion on the aero-acousticsoftransonicfanrotors,J.Turbomach.135 (5)(2013)051013.

[19]I.Clark,R.H. Thomas,Y.Guo,Aircraftsystemnoiseassessmentofthe NASA D8subsonictransportconcept,in:2018AIAA/CEASAeroacousticsConference, 2018,p. 3124.

[20]H.H.Murray,W.J.Devenport,W.N.Alexander,S.A.Glegg,D.Wisda, Aeroacous-ticsofarotoringestingaplanarboundarylayerathighthrust,J.FluidMech. 850(2018)212–245.