AEROACOUSTIC SIMULATION OF A COMPLETE H145 HELICOPTER INDESCENT FLIGHT

Ulrich Kowarsch, Constantin Ohrle, Manuel Keßler and Ewald Kramer

Institute of Aerodynamics and Gas Dynamics (IAG), University of StuttgartPfaffenwaldring 21, Stuttgart, 70569, Germany

kowarsch@iag.uni-stuttgart.de

Abstract

In the past years, the aeroacoustic noise emission of a helicopter became one of the most important, but alsochallenging issues in helicopter development. The blade vortex interaction phenomenon is one of the dominantphenomena characterizing the helicopter’s aeroacoustic footprint, which is insufficiently predicted by low fidelitycomputational methods. For a high fidelity noise prediction of a helicopter configuration, a multidisciplinary CFD-CSD-CAA tool chain has been established at the Institute of Aerodynamics and Gas Dynamics of the Universityof Stuttgart. With higher order CFD computed noise generation at the near field and the noise convection usinga Ffowcs-Williams Hawkings based CAA code, very good agreement to measured aeroacoustic noise in windtunnel as well as free flight experiments of helicopters is achieved. However, the simulations had been limited tothe main rotor’s geometry up to now, where some residual deviations to the experiment still exist. In this paper, wepresent a high fidelity aeroacoustic simulation of a complete helicopter configuration and the benefit comparedto an isolated rotor simulation in predicting its aeroacoustic noise emission. Shading and reflection effects areclearly resolved, influencing the behaviour of the helicopter’s noise radiation. The simulated aeroacoustic noiseemission of the helicopter lies within the experimental variation and shows therefore highly promising results forthe next generation of aeroacoustic noise prediction.

1 NOTATION

BVI blade vortex interactionBPF blade passing frequencycl sectional lift coefficientCAA Computational Aero-AcousticsCFD Computational Fluid DynamicsCSD Computational Structure DynamicsDOF degrees of freedomEPNL Effective Perceived Noise LevelGP ground plateMR main rotorPNLT tone corrected Perceived Noise LevelR rotor radiusr/R relative blade radial stationRMS root mean squareL Sound Pressure LevelTR tail rotorWENO Weighted Essentially

Non-Oscillatory SchemeWVL Wavelet Transformation

2 INTRODUCTION

Helicopters often show the most annoying noise char-acteristics in decent flight. Strong impulsive noisecaused by interactions of vortices with the rotorblades, known as blade vortex interaction (BVI) phe-nomenon, accounts for increased noise levels in thisflight state. Current research is focused on a reli-able computational prediction of the aeroacoustic be-haviour of the rotor in this flight state. In recent yearsa growing environmental awareness in society can beobserved and therefore more and more attention hasbeen drawn to low noise concepts for instance. Toachieve an increased acceptance of helicopters in so-ciety, reduced noise levels and therefore an improvedaeroacoustic predictive capability in the helicopter de-sign phase are necessary. Limited computationalresources and too high numerical dissipation of thevortex structures previously prevented the successfulprediction of this phenomenon using CFD methods.Recently, besides the availability of higher computa-tional resources, numerics have made progress in thefield of higher order methods to improve the vortexconservation. Investigations of Boyd [1] and Ahmed [2]

already presented the advantage using such numer-ical methods for helicopter simulations in this flightstate. Boyd showed that CFD codes using higher or-der methods are in principle capable of resolving BVI

measured in a wind tunnel experiment. A previousjoint work between the Institute of Aerodynamics andGas Dynamics (IAG) of the University of Stuttgart andAirbus Helicopters Deutschland GmbH [3] showed theadvantage of using higher order methods to resolvethe aeroacoustic noise emission for a full scale free-flight CFD simulation. Comparisons with experimen-tal data confirmed the significant improvement resolv-ing BVI phenomena more reliably than former low or-der simulations. However, the usual simulation ap-proach to consider only the isolated main rotor geom-etry for the CFD investigation still showed slight devi-ations to flight test measurements. Within the presentinvestigation the concept is pushed forward and thefull helicopter geometry, including fuselage, skids andtail rotor, is considered. The aim is to resolve the influ-ence of these components to the aeroacoustic noiseemission in terms of shading, diffraction and reflec-tion. As in the previous isolated rotor investigation, aflight test according to the ICAO approach test certifi-cation is simulated, from which noise emission mea-surements are available. The aeroacoustic results arecompared to microphone measurements as well as tothe results of the isolated rotor simulation.

3 METHODS

3.1 CFD Solver: FLOWer

For the current investigation, the block structuredfinite volume Reynolds-averaged Navier-Stokes(RANS) CFD code FLOWer of the GermanAerospace Center (DLR) [4] is used. The RANSequations are closed using the Wilcox k-ω turbulencemodel [5] with a fully turbulent flow. The time dis-cretization is achieved by integrating the governingdifferential equation in space with the implicit dualtime-stepping approach according to Jameson [6].The consideration of relative grid movements usingan Arbitrary Lagrangian Eulerian (ALE) approachenables the code for helicopter flow simulation. Inaddition, the Chimera technique for overset gridssimplifies the meshing of complex helicopter ge-ometries like rotor-fuselage configurations includingrelative grid movements. To consider the effectsof fluid-structure interaction on the rotor blade, themesh is deformed to a given structural deformationof the blade for each time step. The efficiency of thecomputation is achieved by a multi-block structure ofthe grid to enable parallel computing.Within the past years, several important helicopterrelated features have been implemented into FLOWerby IAG and Airbus Helicopters Deutschland. Withinthe so-called HELI-version, an automated strongand loose coupling exchange with CSD Tools, a griddeformation algorithm and extensive rotor related

post-processing output have been included [7].In the course of this code improvement cooperation,the CFD solver was extended with different methodsof fifth order spatial WENO schemes by IAG [8] toguarantee a detailed conservation of the flow fieldand especially vortices. First applications of thehigher order methods in FLOWer confirmed signif-icant improvements in wake conservation and loadprediction in case of wake-structure interactions [9;10].For the presented simulation, a fifth order WENO-Zscheme according to Borges [11] is used for the con-vective fluid state reconstruction at cell boundaries.The resulting Riemann problem is solved using theupwind HLLC scheme according to Toro [12]. Theviscous fluxes are solved with conventional centraldifferences of second order accuracy.

3.2 CSD: CAMRAD II

Besides a high-fidelity aerodynamic investigations us-ing CFD, an accurate reproduction of the helicoptersflight state as well as major structural dynamic char-acteristics are required. Therefore, the Comprehen-sive Analytical Model of Rotorcraft Aerodynamics andDynamics (CAMRAD II) code [13] is integrated into thenumerical tool chain. The rotorblade’s structural dy-namics are modelled by a collection of finite beamsegments whose loads can be prescribed by pro-gram input. This allows a forwarding of loads fromFLOWer to CAMRAD II and the feedback of resultingblade deformations. Coupling CFD and CSD createsa closed loop by exchanging loads and deformationinformation. For this investigation, the cycle is per-formed after each rotor revolution by the exchangeof periodic loads and deformations. This exchangeof periodic rotor characteristics is known as the weakfluid-structure coupling. To achieve a physical flightstate, the flight mechanics of the helicopter is takeninto account to achieve a force and moment free flightstate of the helicopter. Therefore, a free-flight heli-copter trim is performed including 6 degrees of free-dom (DOF). Beside the 3 main rotor controls, collec-tive and cyclic pitch, the fuselage pitch and roll orien-tation as well as the tail rotor collective are taken intoaccount. The yaw angle of the helicopter is fixed atzero degree. Analogue to the structural dynamic trim,forwarded periodic CFD loads allows the CSD code todetermine the load residuals to be eliminated by DOFangle adaptation.

3.3 CAA: Aeroacoustics

For calculation of noise emission towards the far field,the CAA code ACCO [14], developed by IAG, is ap-plied. The basis for the computation is given by theconverged solution of the CFD computation in termsof a data extraction over the time.ACCO uses the Ffowcs Williams-Hawkings equation

(1)

∂2ρ ′

∂ t2 − c2∇2ρ ′ =∂ 2

∂xi∂x j[Ti jH( fS)]

−∂

∂xi

[

(p′ni +ρui(un − vn))δ ( fS)]

+∂∂ t

[(ρ0vn +ρ(un− vn))δ ( fS)]

for acoustic modelling. ρ ′ is the density fluctuation, p′

the pressure fluctuation, ni the normal vector, un thenormal component of the fluid velocity, vn the normalcomponent of the surface velocity, δ the Dirac deltafunction, Ti j the Lighthill tensor, and H( fS) the Heavi-side function of the surface function of the integrationsurfaces fS = 0. With the use of the wave equation onthe left hand side, the flow outside of the integrationsurface is assumed to be an undisturbed freestream.The right hand side of equation (1) represents thesource terms, which can be interpreted as volume dis-placement (monopoles), load fluctuations (dipoles), ifthe integration surface fS coincides with the physicalsurface, and turbulence, shear layers, and compress-ibility effects (quadrupoles).

4 FLIGHT TEST DESCRIPTION

Main focus of this paper is on the descent flight, asthis flight condition often shows the most annoyingnoise characteristics due to strong impulsive noise.To assess the aeroacoustic predictive capabilityof the numerical tool chain described before, thesimulation results will be compared to flight test data.The considered flight tests were performed in June2012 within the scope of the noise certification cam-paign of the H145 helicopter. Figure 1 schematicallyillustrates the noise certification approach procedureincluding trajectory constraints. In accordance to thehelicopter noise certification rules and regulations [15],the helicopter follows a flight path with a descentangle of 6 at a flight speed of Vy (best rate of climbspeed). Besides the center microphone, which isoverflown in an altitude of 120 m above ground, twomicrophones were installed perpendicular to the flighttrack at a lateral distance of ±150 m.By way of derogation from noise certification forthis test case, microphones are considered whichwere placed above ground plates (cf. Figure 2).In this case, the signal experiences a defined total

Fig. 1. Flight boundaries for approach test condi-tion [15] .

reflection on the ground plate (GP), which can beconsidered analogue in the simulation by a doublingof the free-space computed pressure-time signal.Aeroacoustic key values are computed according tothe ICAO evaluation procedure. For the evaluation,the microphones signals are considered as measuredin the form of pressure-time signals. A more detaileddescription and evaluation of the noise measurementcampaign of the H145 gives Reference 15 and 16.

Fig. 2. Installation of considered microphonesover a ground plate (GP) for the approach flight.

Table 1 gives an overview of the averaged flighttest conditions as well as major figures of the H145helicopter. The numerical results also considermeasured wind data (direction and speed) andatmospheric sound absorption.

Characteristics

Main rotor blades 4Main rotor radius 5.5 mMain rotor freq. ΩMR

Fenestron® blades(unevenly distributed, symmetric) 10Fenestron® freq. ΩT R=8.2 ΩMR

Fenestron® modulation freq. 2 ΩT R

TAS 70 knFlight path angle 6.0

Height over mic. 120 mNo. of mic. 3

Table 1. Figures of H145 and average flight testcondition.

5 SIMULATION

5.1 CFD / CSD

The helicopter geometry considered in the CFDsimulation covers all aerodynamically relevant com-ponents, including the main rotor with a detailedmodel of the rotor head, the fuselage including skidsand the full Fenestron® anti-torque system with statorand rotor blades. For a more precise simulation ofthe aft body wake interacting with the Fenestron®, theturbine exhaust is simulated in terms of a boundarycondition prescribing the mass flow through the en-gine and the exhaust temperature. Conservativenessis ensured with a mass coupling between the inletand outlet surface of each engine. Table 2 shows grid

Component No. of blocks No. of cells (M)

Background 6563 106.3Main rotor blade 4× 230 4× 6.9Blade root 4× 56 4× 1.0Rotor hub system 610 9.2Fuselage 1505 26.7Skid system 375 7.1Fenestron® stator+ rotor 670 11.9

Total 10867 191.7

Table 2. Grid components of complete helicopterCFD setup.

components of the complete helicopter configurationincluding 59 separate meshes. The body meshesare extruded from y+ ≈ 1.0 (on main rotor bladesy+ ≈ 0.5) with a cell height ratio of 1.15 until theresolution of the Cartesian background mesh isreached. On body mesh boundaries, the exchangeof the fluid state is performed using the Chimeratechnique. The near-field in the off-body mesh hasa resolution of 6% of the blade chord length. Thecoarsening of the off-body mesh towards the far fieldis enabled by the use of hanging grid nodes. Thebackground mesh is created automatically with a

Fig. 3. Merging of grid components to a full heli-copter CFD setup using the Chimera technique.

mesh refinement in specified areas of interest. Withthe focus on aeroacoustic simulation, the area withthe finest grid resolution is defined to enable a sur-face definition around the complete helicopter, whichdoes not have any inflection points. This guaranteesan extraction of explicit pressure disturbances usedas a CAA integration surface. In addition, the finestgrid resolution ensures a wave propagation of thepressure disturbances onto the integration surfacewith little numerical dissipation. Figure 3 shows anoverview of the merged grid components.Previous simulations showed sufficient mapping offorces and moments acting on the surface using lowerorder methods in combination with reasonable gridresolution. Therefore, the simulation is started usingthe low computational effort 2nd order JST method forthe free-flight trim process. The free-flight CFD-CSDtrim coupling is performed after 2 computed mainrotor revolutions to overcome start-up effects. Thestop criterion for the free-flight trim process is setto a residual less than 0.1 m/s2 translational acceler-ation and 1 /s2 angular acceleration. This processrequired 12 iterations between the CFD and CSDcode to converge to the stop criterion, with up to1.5 CFD computed main rotor revolutions betweeneach iteration step. The long trim process is mostlya consequence of the fact that the investigatedslow descent flight state is highly unsteady due tolow power requirement and therefore low requiredanti-torque thrust. This leads to a high sensitivityand low damping against small disturbances tothe tail rotors collective angle adjustment. Table 3shows the remaining residuals of the accelerationsas well as the deviation of the trim objectives fromthe average values measured in the experiment.Comparing the deviations to the RMS value of theexperiment shows that most values lies within thetypical fluctuation range of the flight test. The mostsignificant discrepancy is present in the tail rotorcollective angle, representing the compensation ofthe torque moment. Due to the already mentionedlow thrust required and the highly detached flow, a

deviating estimation in the numerical tool chain is notunexpected.After convergence of the trim process, the simulation

H/C acceleration residual [m/s2]

avertical 0.04alongitudinal -0.08alateral -0.01

Trim objective ∆CFD−exp[] RMS of exp.

Control angelΘ0,MR 0.52 0.67Θc/lateral 0.36 0.20Θs/longitudinal 0.90 0.37Θ0,T R 3.19 0.93

Fuselage orientationΘpitch 0.68 0.67Φroll 0.17 2.47

Table 3. Trim objective of 6 DOF free flight trim.

is switched to 0.25 azimuthal resolution and 5th or-der WENO-Z discretization, and continued for 3 moremain rotor revolutions to achieve a full higher orderspatial flow field with a fine temporal discretization.The switch showed no significant change in forcesand moments giving reason for a further CFD-CSDtrim iteration.Mandatory for the consideration of the long timeaeroacoustic noise emission is a database with atleast one period of a periodically assumed flow field.For a helicopter, this is aimed for the primary noisesources, the main and the tail rotor. In case of theH145 helicopter, the Fenestron® tail rotor system hasa 8.2 times higher rotational frequency as the mainrotor, leading to a periodicity each 5th main rotorrevolution. Thus, the CFD simulation is continued forthese 5 main rotor revolutions with comprehensiveoutput for the evaluation and CAA post-processing.An extensive block splitting of the structured gridsenables an efficient parallelization of the simulation.The CFD simulation was performed on the HighPerformance Computing Center (HLRS) in Stuttgarton the Cray XC40 Hornet system using 6000 coreswith ≈700 WENO time steps per 24 hours.

5.2 Aeroacoustic post-processing

As already mentioned, the data used for the aeroa-coustic noise emission is extracted from the CFD fieldsolution in each time-step. In this case, 2D integrationsurfaces for fast pre-processing are chosen. Due tomissing volume data provided to CAA with this re-striction, a fully developed flow field has to be presentto include quadrupole noise. Using the RANS CFDmethod, quadrupole noise is only produced at shocksor in the shear layer due to the usage of turbulence

models in the blade boundary layer. This leads toan integration surface located as far as necessaryand as close as possible to the body surface. Withthe usage of the higher order method it is ensuredthat only minor numerical dissipation between thenoise sources and integration surface is introduced.Figure 4 shows the integration surfaces investigated

Fig. 4. Location of fluid state extractions used asintegration surfaces for the CAA computation.

in this paper. The reference integration surface ispresented by the permeable surface around thecomplete helicopter (green) in a small distance fromthe body surface. With the location within the finestgrid resolution a detailed transport of the pressuredisturbances is ensured.The primary noise drivers in the considered flightstate are lift and load fluctuations (impulsive noise),which are dominated by the main rotor and producedon the physical surface of the rotor. Therefore, another integration surface is evaluated, which coin-cides with the physical rotor surface to approximatelyextract the main rotor proportions of the total signal.With the neglected fuselage pressure signature andnegligible reflection feedback from the fuselage, thedifference between the integration surface around thecomplete helicopter (green) and the main rotor blades(blue) reveals the influence in terms of reflection,shading and diffraction effects of the fuselage.The tail rotor noise is approximately separated byan integration surface closing the shroud of theFenestron® (red). On each integration surface, thepressure signature is extracted for each CFD timestep leading to a sampling frequency of ≈8900 Hzwith overall 7200 samples.With the solution of the Ffowcs Williams-Hawkingsequation, time domain acoustic pressure is computedat locations of interest in the far field, which arecalled observers. Further post-processing of thetime domain pressure history leads to aeroacousticspecific values such as narrow-band spectra, SoundPressure Level (L) and tone corrected PerceivedNoise Level (PNLT).

In case of near-field acoustic evaluations, pressuretime history at discrete points is directly extractedfrom the CFD solution.

Wavelet transformationResolving the characteristic disturbances of BVIin a pressure time signal with the usually usedFourier transformation lacks a detailed mapping ofthe temporally restricted events. The character ofthe non-trigonometrical shaped BVI events leadsto undesired leakage effects applying a Fouriertransformation. Only a finite frequency range, usuallybetween the 6th-40th BPF, can be assigned to BVInoise. This leads to noise associated to BVI duringtime spots where no BVI events occur and may becaused by other components in the same frequencyrange. Therefore, a wavelet transformation is per-formed to isolate the BVI events of the pressure timesignal in order to quantify their contribution to theoverall signal. Inspired by the BVI identification pro-posed by Stephenson and Greenwood [17], the Morletwavelet is chosen, which adequately characterizeshelicopter acoustic signals, especially the impulsive,wavelet shaped BVI events. Applying the wavelettransformation to the pressure time signal p(t) givesthe scales of mother wavelets p( f , t) assigned to theinvestigated frequencies f over the considered timewindow t. A filter criterion is designed to isolate BVIevents which satisfy the condition

(2) pBVI( fi, ti) =

p( fi, ti) if EBVI(ti)>

αcut EMR(ti ± trev)

0 otherwise

with EBVI(ti) as the spectral wavelet intensity indecibel of p( fi, ti) over a frequency band between4th-16th BPF, EMR(ti ± trev) the spectral intensity indecibel between 1st-4th BPF representing the mainrotor’s base noise level averaged over one mainrotor revolution trev around the considered time ti.The scaling factor αcut gives the threshold level forwhich proportion of the main rotor’s base intensityhas to be reached to define the occurring waveletas a BVI event, defined for this investigation to 0.85.In addition, a blending function requires the filtercriterion to be satisfied for a time range greater than5 main rotor azimuth, in order to suppress undesireddetected disturbances. After the application of thefilter criterion, pBVI( f , t) is transformed back intothe time domain and post-processed for instanceto identify the BVI contribution to the overall noiseemission.Figure 5 shows the application of the designed filterto a pressure time signal sequence. The BVI eventsare successfully extracted from the raw pressuretime signal and the time windows between the eventsremain untouched.

time [-]

pres

sure

[-]

(a) Raw pressure time signal

time [-]

pres

sure

[-]

(b) Extracted BVI events

time [-]

pres

sure

[-]

(c) Residual signal without BVI events

Fig. 5. Adaption of the designed wavelet filter todetect BVI events in a pressure time signal.

Fig. 6. λ2-visualization of the flow field around the H145 in a BVI releva nt descent flight with the pres-sure contour on the surface.

6 NUMERICAL RESULTS

6.1 Aerodynamic flow field

The vortex structure of the flow field around thehelicopter is shown in Figure 6 using the λ2-criterionfor the vortex visualization. The high grid resolutionin the near field area of the helicopter in combinationwith the higher order WENO-Z method yields ahighly detailed resolution of the vortex structure.The blade vortices are preserved compact duringtheir convection downstream, which is a necessarycondition for a detailed representation of the bladevortex interactions, clearly notable in the flow field.Moreover, tip vortices sweeping over the subsequentblade cause shear layer roll up effects, which causesecondary vortex structures interacting with the bladeas well. This can be seen at the advancing bladeside where several non blade tip vortices are presentgenerated by the described phenomenon.At the engine inlet, the mass flow into the engines isnotable with a lower pressure increase compared tothe stagnation point at the fuselage front side. In thewake of the skids a typical von Karman-like vortexstreet is visible produced by the cylindrically shapedskids.The highly resolved grid is kept up to the tail fin, fromwhere on a coarsening to the far field is deployed.This is also present in the vortex visualization with a

thickening of the vortices as well as strong dissipationof small scale disturbances, and notable in the areabehind the fin.The influence of the engine exhaust is visualized in

Fig. 7. Engine exhaust visualization by tempera-ture contour.

Figure 7 by slices marking temperature hot spotsbehind the fuselage. The cooling rate of the hightemperature exhaust, prescribed by the boundarycondition, is shown within the exhaust jet immediatelyafter the exhaust. Convecting downstream, the jetsdisintegrate into a turbulent structure due to thehigh exhaust speed and its shear layer. Moreover,the exhaust interacts with the rotor downwash andthe highly turbulent rotor hub wake, increasing the

turbulence. After the mixing taking place in the areaof the tail boom, the jet still contains high temperatureareas influencing the inflow of the Fenestron® andinteracting with the stabilizers.Figure 8 shows the time averaged normalized densitydistribution of the Fenestron® inlet. Due to the free-stream condition, density reduction is mainly drivenby temperature superelevation in the field causedby the engine exhaust wakes. An elliptical areawith distortion downstream with reduced density ispresent at the Fenestron® inlet disc. The interactionwith the rotor hub wake and especially the enginewake leads to an efficiency reduction of around 6.1 %due to the reduced inlet density.Figure 9 shows the different wake structures caused

Fig. 8. Time averaged normalized density distri-bution on Fenestron ® inlet.

Fig. 9. Delimitations of the wakes caused by dif-ferent structural component of the helicopter.

by the helicopter components. The wake areas areidentified by the computation of mass-free parti-cles seeded directly downstream to the respectivestructure component. The particles are injectedinto the flow field every 5 main rotor azimuth andconvected over 3 main rotor revolutions. The finalparticle situation is taken and compared with a λ2

vortex isosurface. By an inverse distance weighting,the allocation of the vortex to the respective com-ponent wake is determined. The expected verticaldistribution is found, whereby blending effects are

visible especially between the rotor hub wake andthe engine exhaust. A highly turbulent wake structureis present for the fuselage, which indicates strongflow detachment at the fuselage rear. The interactionwith the tail rotor is mainly driven by the fuselage,rotor hub and engine wake. The influence of thedownwash is mostly seen at the rotor hub wakefading significantly towards the fuselage and skidswake.

Main Rotor ForcesPrior to the investigation of the main rotor forces,

Fig. 10. Percentage portion of the root meansquared sectional lift coefficient cl computed over5 rotor revolutions with respect to the mean liftcoefficient.

the periodicity of the main rotor is to be ascertained.Due to the consideration of the rotor hub geometry,which not necessarily induces rotor harmonic flowdisturbances, no intermittency with the main rotorsBPF frequency can be expected. The influenceof other components like the tail rotor system isexpected to be negligible. To resolve the influenceof non-MR periodic flow phenomena to the occurringBVI events, a quantification in terms of the RMSlift coefficient over 5 rotor revolutions is depicted inFigure 10. The RMS values show that only non mainrotor harmonic fluctuations are found in the rotor hubwake. However, the area of concern investigating theBVI on the rotor disc (30 -330 azimuth) shows aflawless periodicity.One primary driver for aeroacoustic impulsive noiseemission by the rotor is the time rate of load changeon the blades. Therefore, the azimuth angle wastaken as the independent variable and the azimuthalderivative of the blade load is formed. In Figure 11,the azimuthal derivative of the sectional lift coefficientcl in polar coordinates is shown. As the underlyingdata for these calculation, the averaged sectional

Fig. 11. Azimuthal (time) derivative of the sec-tional lift coefficient ( cl/dΨ)

lift coefficient over 5 rotor revolutions is taken. BVIevents appear as strong short time load fluctuationson the retreating blade side in the range betweenΨ=270 -320 as well as on the advancing side be-tween Ψ=30 -90 . Comparing the advancing and theretreating blade side, the superposition of the forwardflight speed and the rotation speed characterize thetime occurrence of the BVI events. On the advancingblade side, short time, high frequency BVI driven loadfluctuations are found, whereas the BVI events onthe retreating blade side occur with lower frequency.Behind the rotor head in the vicinity of Ψ=0 high loadgradients are encountered as well. Since the RMSvalue shows high non-rotor-periodic load fluctuations,non-rotor-harmonic noise is expected to arise in thisarea.

6.2 Aeroacoustic emission

Aeroacoustic noise of the full helicopterThe impulsive noise emitted by the main rotor is indirect dependency on the load fluctuations shownin Figure 11. However, the most important char-acteristic for the overall noise emission of the rotoris the spatial distribution and temporal occurrenceof the noise sources in addition to their individualstrength. The superposition of the emitted acousticwaves shows strong directional dependency dueto phase offsets and varying radiation. Figure 12shows a hemisphere under the helicopter colouredwith the computed sound pressure level, based onthe integration surface surrounding the completehelicopter. High noise levels are found on the star-board side of the helicopter with a maximum directionupstream-downward. This coincides with the typicalradiation direction of the impulsive dipole-shaped

noise caused by BVI on the advancing blade side. Atthe port side a considerably lower noise emission ispresent with a spot in the downstream area. Beyondthe BVI events taking place at the retreating bladeside, the high noise level in this area is caused by theFenestron® tail rotor.

Fig. 12. Aeroacoustic noise footprint of the heli-copter on a 3-rotor-radii-hemisphere (integrationsurface complete helicopter).

BVI radiationBesides the fundamental noise produced by themain and tail rotor thrust, the BVI events have ahigh contribution to the helicopter’s noise emission.After the discussion of the complete aeroacousticnoise emission, Figure 13 shows the radiation ofnoise in terms of the Sound Pressure Level, whichis assignable to BVI. The isosurface shows thedirectivity of the high BVI noise areas and the planeshows the BVI related noise footprint 4.5 radii underthe helicopter. Comparing the BVI radiation with theoverall noise carpet in Figure 12 shows that the localhigh noise areas coincide with the BVI radiation ar-eas. However, at the starboard side of the helicopter,no BVI assignable noise could be identified in thepressure time signal despite a high sound pressurelevel found in the overall noise footprint in Figure 12.The occurring noise in this area is mainly driven bythe fundamental rotor noise superimposed with tailrotor noise. The strongest BVI noise emission with itsradiation marked by the isosurface is found upstream.Tracing back its radiation to the helicopter, the sourcearea is found at the advancing blade side of the mainrotor. The slightly non-linear shape of the isosurface

towards the far field refers to an influence of thefuselage by reflection and diffraction effects. A furtherBVI associated noise emission downstream is foundwith lower strength and expansion. The source forthis area is found in the retreating blade side area ofthe main rotor. The previously discussed coherenceof the noise sources is made clear in this visual-ization as well. The advancing blade side shows asignificantly stronger BVI noise radiation comparedto the retreating blade side, despite comparablestrength and amount of occurring load fluctuationevents taking place in the respective quadrant.Minor wrongly detected BVI areas in the vicinityof the tail rotor are found which are caused by theassumed distance weighting of the computed noiselevel, whose error increases towards the helicoptersproximity.

Fig. 13. Radiation of BVI assignable noise emis-sion of the helicopter with a slice 4.5 rotor radiiunder the helicopter.

Aeroacoustic influence of helicopter componentsThe different approximated noise emissions of themain helicopter components in terms of main rotorand Fenestron® emission are shown in Figure 14.The carpets are computed on the basis of the integra-tion surfaces close to the respective rotor illustratedin Figure 4. For comparison purposes, the noiseemission using the integration surface around thecomplete helicopter is shown as well. Full helicopterand main rotor blades carpets have the same colorscaling, whereas the Fenestron® carpet has an offset(Lre f marks the same sound pressure level).Comparing the main rotor (cf. Figure 14(b)) with theoverall noise emission (cf. Figure 14(a)), a reductionin the complete helicopter noise footprint can befound. This is mainly caused by the shading of thefundamental rotor noise by the fuselage componentsincluding the tailboom. Besides the reduction of themaximum, the spread of the high noise area (cf.Figure 14(a) marker (1)) is reduced in the vicinityof the helicopter and mostly downstream, whereas

(a) Full helicopter

(b) Main rotor blades

(c) Fenestron®

Fig. 14. Noise emission on a carpet 4.5 rotor radiiunder the rotor disc coloured with the sound pres-sure level for different CAA post-processed inte-gration surfaces.

the shape directly in front of the rotor only showsslight differences (2). Notable is the shading by thetailboom in the constriction at the rear with a slightlateral displacement to the starboard (3). Upstreamat the port carpet boundary (4), a lower noise levelis found compared to the main rotor noise emission,whereas at the upstream starboard carpet boundary(5) the noise increases. This asymmetric influence,approximately unaffected by the tail rotor, indicatesexplicit influence of the fuselage to directed localnoise emission. The influence of the tail rotor noise(cf. Figure 14(c)) causes the overall footprint toremain at higher levels towards the carpet’s starboardboundary. At the port side the influence of the tailrotor leads to a local maximum compared to the mainrotor noise emission (6).As the most important influence, the reflection effectof the fuselage is determined on the basis of theCFD extracted pressure time signals on the fuselagesurface (cf. Figure 15). A Fourier transformation isperformed and the BVI related frequency range be-tween the 6th-16th BPF is considered. To determinethe orientation of the reflected pressure waves, thecorrelation between the spatial phase distributionon the surface for each frequency is considered. Aspatial differential operator on each element of thesurface is formed, which allows the computation ofthe incoming pressure wave angle. A sum over theconsidered frequency range with the wave anglesscaled with the respective sound pressure level ofthe frequency gives the BVI assigned wave angleof incidence on the fuselage surface (blue vectors,averaged over several surface points for illustrationpurpose). With the surface patch orientation, thereflected wave (orange vectors) is completed. Theincoming wave vectors are extended backwards totheir intersection point with the rotor disc, at whichthe temporal load gradient introduced in Figure 11 isdepicted. The intersection points with the rotor discallow a clear assignment of the considered waves tothe BVI events of the main rotor blades in the areaof 60 azimuth. The reflection shows no significantscattering of the waves due to the non-curvaturegeometry, but rather a constant reflection orientationslightly upstream. At the port side of the fuselage,only highly scattered reflections can be indicatedowing to hardly occurring noise reflection.

The influence of the fuselage components to theBVI radiation is shown in Figure 16 by the differencebetween the wavelet computed BVI signature of theaeroacoustic noise based on full helicopter and mainrotor blade integration surface. The carpet is located4.5 radii under the helicopter. Areas with no relevantBVI assignable noise (cf. Figure 13) are excludedfrom the comparison and coloured in grey. A highlyincreased area of the BVI related noise emission

Fig. 15. Computation of the aeroacoustic wavereflection on the fuselage surface (6 th -16th BPF).Contour on surface shows the sound pressurelevel of the considered frequency range.

is found at the starboard side in front of the rotor.This area coincides with the previously computedreflection on the fuselage surface and shows asignificant increase in the BVI signature compared toan aeroacoustic computation of the main rotor bladesonly. The opposite impact of the fuselage is foundat the port side of the helicopter, where shadingeffects of the advancing blade side BVI reduce theBVI footprint in similar quantity.A further BVI noise reduction area is found far down-stream on the advancing blade side which resultsfrom tailboom shading of the retreating blade sideBVI events.

Fig. 16. Difference of BVI assignable noise 4.5radii under the helicopter between full helicopterand blade integration surface (Grey areas ex-cluded due to low overall noise level).

6.3 Aeroacoustic approach flight

To quantify the predictive capability of the numericaltool chain the computed aeroacoustic noise emissionis compared to microphone data measured duringthe introduced approach test flight campaign.

Even if the high fidelity URANS computation includesthe dominating aeroacoustic noise source (e.g.impulsive rotor noise, shear layers, engine exhaust,and low frequency detaching phenomena), highfrequency broadband noise (e.g. trailing edge noise)and mechanical noise (e.g. engine noise) is notmapped in the current numerical tool. Therefore, anextraction of the impulsive main rotor and tail rotortonal content from the measured signal is performedusing the blade passing frequency information ofboth rotors. This signal is further denoted as MR+TRsignal representing the impulsive noise sources,whereas the total signal is denoted as total. Thismeans that the numerically computed signal level isexpected to be between the impulsive noise MR+TRsignal and the total signal.The simulated data consists of 5 main rotor revolu-tions equal to a time window of ≈0.8 s. To enablea comparison to the full approach flight, which lastsaround 30 seconds, the data is assumed to be peri-odic and chained together to gain a database for theCAA computation over the required time window. Asalready described, the flow field is 1/5 rev periodic forthe main noise emitting components, the main andtail rotor. The integration surface is moved accordingto the flight test trajectory over the microphone arraywhich enables the consideration of the Doppler effect.The amplitude of the free-stream computed pressuretime signal at the microphones is doubled to get acomparable signal to the ground plate (GP) micro-phone, representing the occurring total reflection ofthe pressure waves. According to the flight trajectory,the atmospheric damping of an undisturbed environ-ment is taken into account.

Tone corrected Perceived Noise Level timehistories (PNLT)The comparison of the tone corrected PerceivedNoise Level (PNLT) time histories for the three con-sidered ground plate microphones is shown in Figure17. The abscissa is centred around the fly-by time at0 s (helicopter vertical over the center microphone),whereby negative values represent the approachand positive the departure flight phase. Besidesthe simulation result (blue), the PNLT data for theflight test - total (red) and MR+TR extracted signal(green) are shown. The shown solid flight test datais averaged over 5 approach flight signals and theminimum and maximum occurring PNLT levels areshown in the dashed lines.Comparing the flight test signal for all microphonesduring the approach it can be seen that this flightphase is mainly dominated by main rotor and tailrotor noise. At the fly-by position a higher differencebetween the flight test total and MR+TR signalis encountered, indicating higher proportions ofnon-rotor induced noise. Taking the simulated PNLT

(a) Retreating blade side

(b) Center

(c) Advancing blade side

Fig. 17. Comparison of PNLT time history of flighttest microphones.

time histories into account, the approach phase anddeparture phase are captured very well with goodaccordance in the absolute PNTL value as well as thegradient. The slight deviation to the average PNTLflight test signals almost always stays within the flighttest variations. The simulated values remain betweenthe flight test MR+TR and total signal showing goodaccordance to the expected characteristic. Onlyparticular deviations from the signals are found asin case at the fly-by time of the retreating bladeside (cf. Figure 17(a)) and marginally at the centermicrophone (cf. Figure 17(a)).At low level PNLT values, caused by large distance ofthe helicopter to the microphones, a slight flatteningof the PNLT shape is found in case of the simulation

(notable in cf. Figure 17(a) t>10 s). This results ofthe overall low noise level, where numerically causedbackground noise increases in proportion as well aslonger noise transport through the atmosphere, whichincreases uncertainty in the tool chains atmosphericcorrection. The influence of the tail rotor to the PNLTvalues is notable in the flight test as well as in the sim-ulation data. As seen in the tail rotor noise footprint inFigure 14(c), the noise is mainly produced sidewaysof the helicopter with higher noise emitted starboard,whereas nearly no noise is radiated directly belowthe helicopter. The same character is seen on theretreating and advancing blade sided microphonewith peaks shortly after the fly-by position. As in thenoise footprint, a higher influence at the advancingblade side microphone is present as well as noinfluence to the center microphone.For the computation of the certification relevantEffective Perceived Noise Level (EPNL) value, a goodaccordance of the maximum PNLT value (PNLTmax)is requisite. Apart from the retreating blade side,the PNLTmax values are in general captured withvalues between the TR+MR and the total signal. Forthe retreating blade side an underestimation of themaximum value of >0.5 TPNdB is present. Hence,the overall noise at this microphone is considerablylower compared to the other microphones, whichincreases the demand on the numerical tool chainadditionally.Further mandatory key values of PNLT time-historyare the 10dB-down times, defining the integrationboundaries for the computation of the EPNL value.Those are determined based on the PNLTmax valueand mark the time where initially and lastly the PNLTvalue exceed PNLTmax-10 TPNdB. The respectivetimes are highlighted with dots in the PNLT timehistories in Figure 17.Due to the underestimation of the PNLTmax value atthe retreating blade side microphone, the 10 TPNdB-down times show higher deviation to the flighttest. In the case of the other two microphones,the 10 TPNdB-down times are captured very wellwith only slight variation, since PNLTmax as wellas approach and departure gradients show goodagreement with the flight test data.

Effective Perceived Noise Level (EPNL)The comparison of the EPNL value is shown in Figure18 in terms of the deviation of the flight test averageto the simulated results. The red bars indicate therespective deviation to the minimal and maximaloccurring EPNL values of the considered flight cases.The simulation overestimates the MR+TR signaland underestimates the total signal in case of thecenter and advancing blade side microphone. Thischaracteristic is attributed to the high fidelity URANScomputation disregarding high frequency broadband

(a) Retreating blade side

(b) Center

(c) Advancing blade side

Fig. 18. Comparison of full helicopter CAA-CFDcomputed ICAO approach flight EPNL values tototal and MR+TR filtered experimental signal.

noise and of course mechanical noise sources.In case of the retreating blade side a general under-estimation of the EPNL value is found with marginaldifference to the MR+TR signal. In this case thePNLT time history explains the small deviation ofthe EPNL. The underestimation of the PNLTmaxvalue and the overestimation of the 10 TPNdB-downtimes cancel each other by integration. Therefore,the result is to be assessed carefully. At the otherhand, the notably lower absolute EPNL value at thismicrophone constrains the relevance of its slight errorfor a global assessment of the result.

Pressure-time recordsBesides the ICAO certification relevant values,sections of pressure time histories of the centermicrophone during the approach and fly-by time areshown in Figure 19. The flight test considered ispicked out of the previously averaged flights. During

the approach a very good agreement between thesimulated and measured signal is present. Thestrength of the BVI events are captured as well asthe basic noise, which is dominated by the main rotorfrequency. Only a slight overestimation of the first BVIevent is found. At the fly-by time spot, an excellentaccordance can be found as well. The generalmagnitudes as well as the higher harmonic partsare clearly visible. A deviation of the simulation ispresent in an additional positive BVI caused pressuredisturbance, which is missing in case of the flight testdata. In the experimental data a slight positive peak,underestimated by the simulation, is present beforethe first common present BVI event occurs. However,this two events can not clearly be connected.

(a) Approach

(b) Fly-by

Fig. 19. Comparison of the pressure time signalsduring approach and fly-by situation for the centermicrophone.

BVI proportion during approach flightAs a further comparison, the BVI share of thepressure time signal is determined and compared.Figure 20 shows the evaluation for the BVI dominatedmicrophone at the center position. The dashed linesmark the total Sound Pressure Level (L) and the solidlines the BVI part detected by the wavelet filter.As in the case of the PNLT time history, a very goodaccordance with the overall noise level is present.During the early approach and later departure phase

a deviation between flight test and simulated BVIpart is present. This is mainly caused by backgroundnoise in the flight test data, wherefore partially noiseevents could not be assigned clearly to BVI.However, the previously illustrated BVI noise radia-tion character (cf. Figure 13) is also notable in thetime history of the simulated microphone data, withtwo BVI peaks and a plateau at the fly-by position(-5 s<t<5 s). This characteristic is also found in theflight test data, where as in the case of the simulation,the approach peak directly before the fly-by situationshows the maximum BVI noise share. The shortBVI noise reduction followed by another peak is inaccordance with the simulation results. In the timephase around the fly-by time, the total BVI relatednoise shows good accordance in absolute value aswell as relative to the total noise level.

Fig. 20. Comparison of sound pressure level andBVI assigned noise proportion at the center mi-crophone.

Comparison and assessment of earlier iso-lated rotor investigationIn the following, an assessment of the prior investi-gated isolated rotor is performed. The isolated rotorwas computed with the same numerical tool chainand the same CFD mesh resolution. A detailed de-scription of the investigation and its results includinga comparison to the experimental data can be foundin Reference 3.Even if the results with ENPL deviation of around3.5 dB to the flight test data were outstandingimprovements compared to former aeroacousticsimulations, the next step to simulate the completehelicopter was a necessarily consequence to resolvethe remaining offset.The present evaluation of the full helicopter sim-ulation revealed the influence of the helicoptercomponents to the noise footprint in terms of shadingand reflection effects. The overall noise emission wasseen to reduce (∼2 dB) due to the presence of thefuselage reflected by a lower PNLT value comparedto the isolated rotor investigation.

Besides the influence of the geometry components,a slight difference (∼1 dB) is introduced due to thetrim strategy. The isolated rotor investigation wasbased on a 3 component trim. The trim objectives(thrust, pitch and roll moment) were pre-computedusing low fidelity flight mechanic tools, since no flighttest measurements for this quantities are available.In case of the 6-DOF full helicopter trim, pre-computed trim objectives are obsolete. The approachenables the computation of a fully CFD based flightstate determination, whereby the CFD shows high re-liably computed loads. The remaining shortcomingsare located in the dynamic blade model and CFDenvironmental settings (e.g. wind).The results of this investigation campaign shows thatreliable results for the complex aeroacoustic noiseemission of a helicopter demands a high fidelity CFDcomputation in combination with the consideration ofpreferably every aeroacoustic influencing geometrycomponent. However, it has to be noted that theisolated rotor simulation still gave reasonable resultswith a much lower simulation complexity (80 M vs.190 M grid cells, 3-DOF vs. 6-DOF trim).

7 CONCLUSIONS

An assessment of the CFD-CSD-CAA based repro-duction of helicopter noise emission in a BVI relevantdescent flight phase is presented. Basis for the vali-dation is given by a ICAO noise certification flight testcampaign of the H145 helicopter.In contrast to former simulations focusing on 3-DOFtrimmed main rotors only, the current CFD simulationcovers the complete helicopter geometry with a 6-DOF trim. Hereby, the influence of all helicopter com-ponents to the noise emission as well as the inclusionof noise sources beside the main rotor are consid-ered.Using a high resolution with 190 mio overall grid cellsin combination with the 5th order WENO-Z methodfor the CFD computation ensures a detailed mappingof the noise generating phenomena, prominently BVI.The solution of the CFD flow field is forwarded to theCAA simulation to investigate the far field noise radi-ation.The typical noise radiation character is found for thehelicopter structure, and the BVI related noise couldbe assigned effectively using a wavelet filter for iden-tification. The influence of the geometry components,especially the fuselage, could be resolved clearly withstrong reflection and shading effects of the advancingblade side BVI events. Pursuing the effects, a influ-ence of ±3 dB to BVI related noise is identified.Investigating the approach flight noise of the heli-copter shows very good accordance with the flighttest data. The PNLT time histories are met in shape

and absolute values within the flight test data varia-tion. The followed certification relevant EPNL com-putation shows high accordance as well, where thenumerical signal produces values between the flighttest main rotor + tail rotor signal and the total sig-nal. Taking the neglected high frequency broadband(URANS) and mechanical noise sources (e.g. enginenoise) into account, this is the required range.Compared to a previous isolated rotor investigation,the advantage of simulating the complete helicopterbecame evident. It was shown that with the compo-nent influences, the noise footprint of the helicopterchanges partially considerably. As a result of this,the investigated flight test microphones changes instrength and shape reduce the deviation from flighttest from 3.5 dB overestimation to less than 1.0 dB.This shows the necessity to consider the full heli-copter geometry for high fidelity aeroacoustic noiseemission results. The marginal deviations show thevery good capability of the presented numerical toolchain to reflect the aeroacoustic characteristics of thehelicopter in this complex flight state.Next steps in investigating aeroacoustic helicopterflow fields will be the consideration of hybrid LES-URANS methods to include first broadband noisecomponents.The available high fidelity flow field constitutes an im-provement of knowledge to industry. Based on thosedetailed investigations of the aeroacoustic influence,developments may further improve the aeroacousticnoise emission e.g. by absorbing surface materials onthe fuselage to reduce constructive reflection. How-ever, currently available computational resources inindustry prohibit comparable simulations, thus this re-mains a task for research facilities in the near future.

ACKNOWLEDGEMENTS

This work was performed within the federal researchproject ECO-HC2, which is lead by Airbus HelicopterDeutschland GmbH.The authors would like to thank Airbus HelicoptersDeutschland GmbH for the esteemed cooperationwithin this project and beyond, as well as providingus experimental data to enable this investigation. Wewould like to express our thanks to the Federal Min-istry of Economics and Technology for providing usthe resources to realize this project. Furthermore, theinvestigation is based on the long-standing coopera-tion with the German Aerospace Center (DLR) mak-ing us their CFD code FLOWer available for advance-ments and research purpose, which we would like tothank for.Further acknowledgement is made to the High Perfor-mance Computing Center in Stuttgart who providedus with support and service to perform the compu-

tations on their high performance computing systemHornet.

REFERENCES[1] Boyd, D. D. J., “HART-II Acoustic Predictions us-

ing a Coupled CFD/CSD Method,” Proceedingsof the American Helicopter Society, 65th AnnualForum, Grapevine, Texas, 2009.

[2] Ahmad, J. U. and Chaderjian, N. M., “High-OrderAccurate CFD/CSD Simulation of the UH-60 Ro-tor in Forward Flight,” 29th AIAA Applied Aerody-namics Conference, Honolulu, Hawaii, 2011.

[3] Kowarsch, U., Lippert, D., Schneider, S., Keßler,M., and Kramer, E., “Aeroacoustic Simulation ofan EC145T2 Rotor in descent flight,” 71th Amer-ican Helicopter Society Annual Forum, VirginiaBeach, VA, 2015.

[4] Kroll, N., Eisfeld, B., and Bleeke, H., “The Navier-Stokes code FLOWer,” Notes on Numerical FluidMechanics, 1999, pp. 58–71.

[5] Wilcox, D., “Multiscale Model for TurbulentFlows,” AIAA Journal, Vol. 26, No. 11, 1988,pp. 1311–1320.

[6] Jameson, A., “Time dependent calculations us-ing multigrid, with applications to unsteady flowspast airfoils and wings,” AIAA 10th Computa-tional Fluid Dynamics Conference, Honolulu, HI,1991.

[7] Eurocopter Deutschland GmbH, Internal Report,Manual for the FLOWer HELI Version, version2008.1 rel 1012 edition, 2011.

[8] Kowarsch, U., Oehrle, C., Hollands, M., Keßler,M., and Kramer, E., High Performance Comput-ing in Science and Engineering 13, Springer Ver-lag, 2013, Chapter Computation of HelicopterPhenomena Using a Higher Order Method, pp.423–438.

[9] Kowarsch, U., Keßler, M., and Kramer, E., “HighOrder CFD-Simulation of the Rotor-FuselageInteraction,” 39th European Rotorcraft Forum,Moscow, 2013.

[10] Kowarsch, U., Keßler, M., and Kramer, E., “CFD-Simulation of the Rotor Head Influence to theRotor-Fuselage Interaction,” 40th European Ro-torcraft Forum, Southampton, 2014.

[11] Borges, R., Carmona, M., Costa, B., andDon, W. S., “An improved weighted essentiallynon-oscillatory scheme for hyperbolic conserva-tion laws,” Journal of Computational Physics,Vol. 227, (6), 2008, pp. 3191 – 3211.

[12] Toro, E. F., Riemann Solvers and NumericalMethods for Fluid Dynamics, Springer Verlag,Berlin, 1997.

[13] Johnson, W., CAMRAD II Comprehensive ana-lytical model of rotorcraft aerodynamics and dy-namics, fourth edition, 2009.

[14] Keßler, M. and Wagner, S., “Source-TimeDominant Aeroacoustics,” Computers & Fluids,Vol. 33, 2004, pp. 791–800.

[15] International Civil Aviation Organization, En-vironmental Technical Manual - Volume I:Procedures for the Noise Certification of Aircraft,doc. 9501 edition, 2012.

[16] Gareton, V., Gervais, M., and Heger, R.,“Acoustic Design and Testing of the EurocopterEC145T2 and EC175B - a harmonized Franco-German Approach,” 39th European RotorcraftForum, Moscow, 2013.

[17] Stephenson, J. and Greenwood, E., “Effects ofVehicle Weight and True Versus Indicated Air-speed on BVI Noise During Steady DescendingFlight,” 71th American Helicopter Society AnnualForum, Virginia Beach, VA, 2015.