17th AIAA/CEAS Aeroacoustics Conference(32nd AIAA Aeroacoustics Conference) 05 - 08 June 2011, Portland, Oregon
AIAA 2011-2766
Microphone-Array Measurements of a Model Scale Contra-Rotating Open Rotor in a Reverberant Open Wind-Tunnel Stefan Funke∗, Larisa Kim†, and Henri A. Siller‡ DLR, German Aerospace Center, Berlin, Germany
The DLR department of engine acoustics performed acoustic measurements of a contrarotating open-rotor (CROR) model with a linear microphone array in the T-104 open wind-tunnel at TsAGI. The microphone data is corrected for shear-layer effects using the model of an infinitely thin shear layer by Amiet. Investigations of the room-acoustics of the test hall show strong reflections from the ceiling and two side walls, which cause strong interference patterns at the microphone positions. The acoustic change between different rotor configurations, like different rotor spacings and an upstream mounted mock pylon, is shown with spatially averaged spectra. The evaluation of the source directivities is strongly influenced by the interference patterns of the reflections in the test hall. The rotor tones can be localized at the correct position with classical beamforming. The inverse method SODIX is used to model the directivities of the broadband sound-sources of the CROR. Green’s functions that take the shear layer and the first-order reflections into account were implemented. The accuracy of these correction methods could be investigated with the directive SODIX results. The shear-layer model gives accurate results for emission angles < 100◦ while all modeled sources move downstream for larger emission angles. Corrections for the first-order reflections of the test hall show an increased spatial resolution of the SODIX results and a suppression of artificial sources.
I.
Introduction
The contra-rotating open rotor (CROR) is an efficient propulsion concept promising enormous fuel savings compared to shrouded turbofan designs. The disadvantage of this concept is the high noise emission, especially the high levels of the rotor tones. Today, computational methods are used to optimize the aerodynamic as well as the aeroacoustic performance of turbomachines. Nevertheless, experimental measurements are necessary to understand noise sources mechanisms and to validate low noise solutions and prediction tools. Within the framework of the European project DREAM, the department of engine acoustics of the German Aerospace Center (DLR) performed acoustic measurements of a model-scale CROR. The CROR was designed in sub-project 3 of DREAM under the leadership of the french engine manufacturer Snecma. The forward rotor counts 12 blades while the rear rotor counts 10 blades. The hub to tip ratio is 0.38 and the rotor spacing was varied between 0.2 and 0.3 rotor diameters. The aeroacoustic measurements of the V1.1 rotor design took place in June 2010 in the large subsonic open wind-tunnel T-104 at the Central Aerohydrodynamic Institute (TsAGI) in Zhukovsky, Russia. TsAGI performed the aerodynamic measurements and acoustic measurements with highly directive tube-microphones for certain emission angles of the CROR. The DLR department of engine acoustics performed acoustic measurements using a linear array with 104 microphones that were set up outside the wind-tunnel flow on the ground of the test hall in parallel to the engine axis. The acoustic measurements were performed simultaneously with the performance tests of the CROR. ∗ Research Engineer, DLR, German Aerospace Center, Institute of Propulsion Technology, Engine Acoustics Department, M¨ uller-Breslau-Str. 8, 10623 Berlin,
[email protected] † Graduate Student at TU Berlin, Student Assistant at DLR, Engine Acoustics Department, Berlin ‡ Research Scientist, DLR, German Aerospace Center, Institute of Propulsion Technology, Engine Acoustics Department, M¨ uller-Breslau-Str. 8, 10623 Berlin,
[email protected]
1 of 11 Copyright © 2011 by the author(s). Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
American Institute of Aeronautics and Astronautics
The noise of a CROR is dominated by the blade passing tones of both rotors and the tones due to the interaction of the two rotors. The frequencies of the interaction tones can be calculated from the extended Tyler and Sofrin formulation1–3 f = h1 B1 Ω1 · h2 B2 Ω2 , (1) with h1 , h2 the harmonic numbers; B1 , B2 the number of blades; and Ω1 , Ω2 the shaft speeds of the two rotors. Experimental studies showed strong directivities of the rotor tones.4 The broadband noise of a propeller reaches maximum levels in lateral directions between 90◦ and 100◦ . In this paper we show exemplary results from CROR-measurements in a closed test-section wind-tunnel and discuss the influence of the wind-tunnel shear-layer and the room acoustics of the surrounding test-hall on the results. Absolute values can not be shown because of disclosure agreements. CROR
Flow
linear microphone array
(a) T-104 schematic: 1 - nozzle, 2 - double fan, 3 - test section, 4 . stilling chamber, 5 - aerodynamic balance cabin, 6 - operator cabin, 7 - diffusor, 8 - incoming ventilation well, 9 - exhaust ventilation well, 10 - gas-suction system, (source: www.tsagi.ru).
(b) View from the microphone array on the acoustically hard walled ground to the CROR installed in the open test-section of the T-104 wind-tunnel.
Figure 1. The T-104 open wind-tunnel at TsAGI in Zhukovsky, Russia.
II.
Measurement setup
Figure 1 shows the T-104 open wind-tunnel at TsAGI, with its closed operating-circuit and open testsection. The Nozzle diameter is 7 m and the wind tunnel was operated with different Mach numbers between 0.2 and 0.3. The distance from the nozzle to the CROR is about 7 m. Figure 1(b) shows the CROR in the VP-107 test-rig. The propellers, mounted face to face, are driven independently by two electric motors. The test hall is 18.5 m high and the plane ceiling can be assumed to be acoustically hard. Several supporting structures and technical installations make the test hall an acoustically complex environment. The volume of the test hall measures several tenthousand cubic meters. Nozzle
Collector
Ceiling Control Room
CROR
Wall
Nozzle H
Wall
CROR Microphone Array
x y h
Wall
Control Room
Wall
Microphone Array
L
z y
(a) xy-plane.
(b) yz-plane.
Figure 2. Setup of the microphone array (red line), true to scale.
2 of 11 American Institute of Aeronautics and Astronautics
A linear array with 104 1/4“ condenser microphones type Microtech-Gefell MK301 with preamplifiers MV302 was set up parallel to the engine axis with a regular spacing of ∆x = 0.15 m. All microphones were set up on an acoustically hard-walled wooden plate on the ground of the test facility in order to avoid ground reflections that would decrease the signal to noise ratio. The microphones cover view angles seen from the rotor between 70◦ and 125◦ with the 0◦ in flight direction. Each measurement took 20 seconds with a sampling rate of 25 kHz. A Fast-Fourier Transform was performed with a window length of 8192 samples using a Hanning window and 50 % overlap. This results in 122 averages and a frequency resolution of 3.1 Hz. The microphones were calibrated before each measurement.
(a) View from above the wind-tunnel nozzle.
(b) View along the linear microphone array.
Figure 3. Measurement Setup in the T-104 low speed wind tunnel at TsAGI.
III. III.A.
Effects of the test conditions on the acoustic measurements
Background noise
SPL in dB, 5 dB per Division
The background noise of the experimental setup was measured separately before the rotor measurements. The background noise consists of the wind-tunnel noise and the noise of the electrically driven test rig. The wind-tunnel noise was measured without the rotors installed in the test rig at wind-tunnel Mach-numbers 0.2, 0.25, and 0.3. The gap at the position of the deinstalled rotors was perfectly closed for these measurements. The motor noise was then measured without the wind-tunnel flow and without the rotor blades installed. The cover of the rotor gap had to be removed for the motor-noise measurements due to the rotation of the two shafts. It turned out, that the motor does not contribute any relevant acoustic energy to the background noise. The total background noise spectrum is shown in fig. 4 together with the spectrum of the CROR baseline-configuration. Both spectra are averaged over all measured microphone positions. In comparison with the CROR-baseline spectrum, the background noise dominates in the frequency range up to the first blade passing frequencies of the CROR. The blade-passing tones stand up well above the background noise. The broadband noise of the CROR shows only a small signal-to-noise ratio.
1R
1F
Background Noise
Rotor Measurement
1F+1R 2F 1F+2R 2R 2F+1R 3R 3F
5 dB
Frequency
Figure 4. SPL spectra of the baseline rotor-configuration averaged over all measured emission angles.
3 of 11 American Institute of Aeronautics and Astronautics
S
E
Ut
ϕm
Θ ϕ0
U
ct
h
O
shear layer
ϕ
y
M Figure 5. Shear-layer refraction of sound waves.
III.B.
Shear-layer effects
The shear-layer of the open jet wind-tunnel refracts and attenuates the sound waves which propagate from the CROR inside the wind-tunnel flow to the microphones outside the flow. The theory of Amiet5 will be used for the consideration of these effects. The flow is assumed to have a velocity U with uniform profile and an infinitely thin shear layer where the velocity jumps to zero. Figure 5 shows an illustration of the geometry. The measurement angle ϕm describes the direct propagation path from the source S to the microphone M if there was no wind-tunnel flow and accordingly no wind-tunnel shear-layer. The sound that is measured at the position of microphone M could also be measured at an angle ϕ0 at the shear layer position. ϕ0 therefore represents the direction of sound inside the moving medium that results from the superposition of the speed of sound and the flow speed inside the wind tunnel. Amiet defines the relations between the angles as p (1 − M cos ϕ)2 − cos2 ϕ 0 , and (2) tan ϕ = (1 − M 2 ) cos ϕ + M y cot ϕm = h cot ϕ0 + (y − h) cot ϕ. (3) Herein, M is the wind-tunnel Mach-number. The refraction angle ϕ can be calculated iteratively from equations 2 and 3. E is the retarded source position and the actual emission angle of the engine in flight Θ is given by6 cos Θ =
cos ϕ . 1 − M cos ϕ
(4)
The effect of random phase shifts due to the turbulent shear-layer was described by Koop et al.7 The quality of array-processing results, that will be shown in this paper, can help to estimate this effect for the given measurement setup. III.C.
Room-acoustic effects
Reflections from the acoustically hard walls of the test-hall were expected to influence the measured sound field at the microphone positions. In order to investigate this effect, the acoustic room-impulse responses were measured with an impulsive sound source that was placed at the position of the deinstalled CROR. Figure 6 shows the time series of an exemplary impulse response as well as a 2D plot of all measured impulse responses where the arrival times of the direct sound were set to t = 0 s. A strong reflection can be found for all microphones which arrives about 50 ms after the direct sound. This matches the time that a sound wave needs to propagate from the rotor via the ceiling of the test hall to the microphones. According to the 1/r distance-law, this reflection from the ceiling has a sound pressure level of about −7 dB with respect to the direct sound. Other reflections appear as diagonal lines in fig. 6(b). These come from the side walls at both ends of the linear microphone array that are parallel to the rotor plane, see fig. 2. The delay of a reflection from a side wall depends on the distance of a microphone to that wall. For microphones that 4 of 11 American Institute of Aeronautics and Astronautics
are close to a side wall, the corresponding first-order reflection tends to coincide with the direct sound. Due to the superposition of the direct sound and the various reflections, the measured sound field is modulated in amplitude and phase. This modulation depends on frequency and on the actual position relative to the reflecting surfaces. The coherence of the microphone signals is also affected by this modulation. As a result, the measured SPL can show strong spatial variations and the evaluation of the device under test can be difficult. In order to reduce the number of strong first-order reflections, the microphones were placed on the floor of the test hall. Thereby, the ground reflection is in phase with the direct sound which causes a doubling of the measured sound pressure which can easily be corrected in the evaluation of the data by subtracting 6 dB. direct sound reflections from side walls
ceiling
reflection from the ceiling
side walls
(a) Exemplary impulse response.
(b) Impulse responses for all microphone positions. The time of arrival of the direct sound is set to t = 0 ms for each microphone.
Figure 6. Room impulse responses measured with an impulsive sound source at the rotor position.
The acoustic environment affects the evaluation of single microphone data as well as phased array techniques. The measured impulse responses can not be used to correct for these effects, because the impulseresponse measurements could only be realized without the wind-tunnel flow. So the shear-layer effects that influence the sound propagation from the CROR-position inside the open wind-tunnel to the microphones outside the test-section are not included in the measured impulse responses. The data can be corrected for the room-acoustic effects with simulations of the room transfer functions if the source characteristics, the exact geometry of the room, and the propagation parameters are known. All these properties are highly complex for the given testing situation. Regarding the large size of the test hall and the relevant frequency range f > 1 kHz, numerical calculations of the room acoustics would need excessive computing power. An alternative analytic method is the consideration of mirror sources for walls with frequency-relevant dimensions. Guidati presented the reflection canceller (rc) that includes the reflections of a closed-section wind tunnel in a classical beamforming (cbf) approach.8 Thereby, the steering vector ~g that holds the free-field Green’s functions for all assumed monopole sources at positions ~s is extended with the Green’s functions for all corresponding mirror sources L at positions ~t. ~grc = ~gcbf (~s ) +
L X
~g (~tl )
(5)
l=1
Following this approach, the geometry of the given test hall can be approximated for three first-order reflections, the reflection from the ceiling and the reflections from the two side walls that are located near the nozzle and the collector of the wind tunnel. These are the reflections that carry the highest acoustic energies. The distances to the walls were measured using a laser distance-meter. The propagation of the sound inside the wind tunnel and through the wind-tunnel shear-layer has to be taken into account in the Green’s functions. It is possible to calculate individual Green’s functions for the different sections of the propagation path and to combine these sections in order to get the full propagation path. Figure 7 presents a ray-acoustics description for the reflections at the ceiling and the side walls. In order to describe the indirect propagation paths, each microphone position M was mirrored at the reflecting wall to the position M 0 . This approach is similar to the image source theory, but with the sound source inside the open wind tunnel it is easier to handle and the reflection angle at the wall must not be calculated. For the evaluation of single microphone data, the correction for the first-order reflections is too inaccurate in the relevant frequency range, especially due to the simplified shear-layer model, but the results from phased array techniques could improve. Results from this approach are shown in this paper. 5 of 11 American Institute of Aeronautics and Astronautics
M’
M’
R
shear layer U
S
ceiling R
side wall
ceiling
U
S
shear layer S O
O
O
R
M’
M
M
(a) Reflection at the ceiling, x-z plane.
(b) Reflection at the ceiling, y-z plane.
M
(c) Reflection at a side wall, x-z plane.
Figure 7. Exemplary propagation paths of the direct sound and the sound waves that are reflected at the walls of the test-hall. The microphone positions can be mirrored at the walls in order to simplify the determination of the propagation-path lengths.
IV.
Results
In the following, selected results of the analysis of the rotor noise will be presented. In the first step, the spatially averaged noise spectra and the directivities of certain rotor tones will be shown. Then, phased array processing techniques will be applied. The rotor tones are localized with conventional beamforming while the broadband noise is modeled with the inverse technique SODIX.9 The effects of the shear layer and the first-order reflections will be taken into account for in the Green’s functions. The results will help to evaluate the quality of the acoustic measurements in the given test environment. IV.A.
Evaluation of the individual microphone signals
The evaluation of the individual microphone signals is the most direct way to analyse the measurement data. The investigations of the room-acoustic effects in section III.C suggest strong variations of the measured SPLs between neighbouring microphones. The coherence function of the microphone signals, that is shown in fig. 8 for two exemplary rotor tones, supports this assumption of a highly modulated sound field. The effect depends on frequency and the pattern becomes more complex for higher frequencies. 20
40
mic. no. 60
80
100
20
1
40
mic. no. 60
80
100
0.9 20
0.9 20
0.8
0.8
0.7 0.6 0.5 60
0.7 40 mic. no.
mic. no.
40
0.4
0.6 0.5
60
0.4
0.3 80
0.2
0.3 80
0.2
0.1 100
1
0.1 100
0
(a) Blade-passing tone 1F of the front rotor.
0
(b) Interaction tone 1F+1R.
Figure 8. Coherence of the microphone signals for exemplary rotor tones.
In order to minimize this effect for the evaluation of the individual microphone data, the spectra of all microphones are averaged over all microphone positions to get a mean spectrum for the whole measured angular range. All spectra were corrected for shear-layer effects and they were normalized to a reference
6 of 11 American Institute of Aeronautics and Astronautics
distance of 1 m from the CROR center before averaging. The averaged spectrum of the baseline rotor-configuration was already shown in fig. 4. The broadband noise of the wind-tunnel dominates up to the frequencies of the first rotor blade-passing tones. The rotor noise itself is dominated by the rotor tones that stand out up to 15 dB well above broadband noise level. Up to five harmonic multiples of the rotor blade-passing frequencies can be seen. The most important tones are the blade-passing tones 1F and 1R and the combination tone 1F+1R. The wind-tunnel shear-layer scatters the energy in the frequency domain when the sound waves propagate through the turbulent shear-layer of the wind tunnel. This effect, the so called hay-stacking, is clearly visible for high frequencies where the rotor tones appear as broadened peaks. The averaged spectrum, as shown in fig. 4, will be used to evaluate the acoustic change between different rotor configurations. Measurements were performed with two different rotor spacings and with a mock pylon installed in front of the rotors in order to simulate the installation of the CROR in a pusher configuration. Trailing-edge blowing of varying strength was applied in order to reduce the wake-induced tonal noise. SPL in dB, 5 dB per Division
1F 1R
small rotor spacing 1F+1R 2F
large rotor spacing
2F+1R 1F+2R
2R 3R 3F
3F+1R
3F+2R
5 dB
Frequency
Figure 9. Comparison of the averaged SPL spectra for the baseline configuration and a configuration with a 40% reduced rotor spacing.
SPL in dB, 5 dB per Division
1F 1R
Without Pylon With Pylon With Pylon and Max. Blowing
1F+1R 2F 2R
1F+2R 3R
2F+1R 3F 4F
SPL in dB, 5 dB per div.
Looking at the variation of the rotor spacing, fig. 9 shows that it has only small effects on the rotor BPFs but it gives clearly higher levels of the interaction tones for the small rotor spacing. This is due to the increased interaction of the velocity field with the bound circulation. For the 2F+1R interaction tone the increase is about 8 dB. Similar results for the interaction tones were found by Janardan and Gliebe10 whereas Magliozzi11 found no change in the total EPNL for different rotor spacings.
5 dB
5 dB
0
Frequency
(a) Averaged SPL spectra
1F 20
40 60 Mass flow in %
2F
3F 80
4F 100
(b) SPLs of the front-rotor tones as functions of the blowing mass-flow, normalized to the maximum tested mass-flow.
Figure 10. Effects of a mock pylon installed in front of the CROR and varying trailing-edge blowing mass-flows.
The effect of a mock pylon installed in front of the CROR is shown in fig. 10(a). The velocity deficit in the wake of the pylon increases the SPL of the forward rotor BPF and its harmonics. According to theory and other experiments12, 13 the rear-rotor tones and the interaction tones remain unaffected. For the dominant 1F the increase is about 1.5 dB. 3F and 4F are stronger with about 5 dB, but the absolute levels are small compared with 1F. However, for a full sized engine, these tones would fall into a frequency range where the human ear is very sensitive. The comparison between the isolated configuration and the configuration with the pylon may be difficult because of the test-rig that might also affect the inflow of the CROR. The small increase of only 1.5 dB could be a sign for this. Nevertheless, trailing-edge blowing can reduce the effect of the pylon wake. Figure 10(b) shows the integrated tone levels of 1F to 4F as functions of the mass flow. The blowing works best for 3F (-4 dB) and 4F (-3 dB). 1F is reduced by 1.5 dB and 2F can be reduced even slightly below the no-pylon configuration. Overall, a saturation of the level reduction indicates an optimum blowing setting close to the highest measured mass-flow. 7 of 11 American Institute of Aeronautics and Astronautics
Not only the absolute level of the emitted noise is of interested, but also the directivity that can be measured with the linear array of microphones. Figure 11 shows the directivities of the rotor blade-passing tones 1F and 1R and the interaction tone 1F+1R for two different rotor spacings. The data are corrected for shear-layer effects and they are normalized to a reference distance of 1 m.
SPL in dB, 5 dB per Division
F, small spacing R, small spacing 2F+1R, small spacing
F, large spacing R, large spacing 2F+1R, large spacing
5 dB
forward rearward
Figure 11. Exemplary directivities of rotor tones with strong local variations up to 8 dB.
As described in section III.C, strong local variations between adjacent microphones occur due to the interference of sound waves that reach the microphones directly from the rotor and the sound waves that are reflected at the ceiling and the walls of the test hall. These variations can be up to 8 dB from one microphone to the next. Apart from this local scatter, the global directivity of the tones is less pronounced than what could be expected from experimental and analytic studies of open rotors in the acoustic far field, e.g. Parry and Crighton,4 where the tones peak at emission angles slightly over 90◦ . The results here show, that the directivity of the front rotor tone 1F shows increasing sound pressure levels towards the rear arc, as it could be expected from a single rotor with an inflow distortion. The rear rotor 1R, on the other hand, shows a maximum near Θ = 90◦ , however the change in sound pressure level with the emission angle is not very pronounced. Regarding the two different rotor configurations, the directivities of 1F and 1R are not affected by a change of rotor spacing. The overall sound pressure level of the 2F+1R interaction tone is lowered by as much as 9 dB for certain emission angles and the directivity is clearly altered when the rotor spacing is increased. IV.B.
Beamforming on the rotor tones
The phased array is a well known tool for investigations of noise sources. Figure 12 shows results from frequency-domain beamforming applied to the rotor tones 1F and 1R and the interaction tone 1F+1R. The data from the CROR baseline-configuration was used. Results from three different implementations of the Green’s functions will be shown in the following. Figure 12(a) shows results from calculations with freefield Green’s functions. Here, the rotors are located downstream of the actual CROR center x = 0 m due to the refraction of the sound waves at the wind-tunnel shear-layer. For the results shown in fig. 12(b), the Green’s functions were extended with a shear-layer model. The blade-passing tone of the forward rotor is then located at the correct position, while the blade-passing tone of the rear rotor is slightly shifted downstream of the actual position. The interaction tone 1F+1R is located closer to the rear rotor. Finally, the results in fig. 12(c) were calculated using Green’s functions which consider the shear layer effects and additionally correct for the first-order reflections from the ceiling and two side walls of the test hall. It can be noticed, that the dynamic range of 1F only slightly increases, while it is even decreasing for 1R and 1F+1R. The assumption of omnidirectional sound sources in the beamforming algorithm is one possible reason for these results. The resolution of the beamforming algorithm depends on the number of microphones which are used for the calculations. A higher number of microphones reduces the main-lobe width and therefore increases the spatial resolution. Because beamforming is an integrating method which averages over the positions of the microphones, the directivity of the sources can not be resolved. A possibility to detect the directivity of the sources with beamforming is to divide the array into subarrays and to interpolate the directivity from the individual sub-array results. The disadvantage is then again a reduced number of microphones in each sub-array and therefore a reduced spatial resolution. Thus, classical beamforming may not be the preferred method for the evaluation of acoustic measurements in the given situation of a directive sound source and a linear microphone array. 8 of 11 American Institute of Aeronautics and Astronautics
1F 1R 1F+1R
0
−15
−20
−5 arb. SPL in dB
−10
−10
−15
−20
−25 −1
−0.5
0
0.5 1 1.5 Axial Source Position in m
2
2.5
(a) Free-field Green’s-functions.
3
−25 −1
1F 1R 1F+1R
0
−5 arb. SPL in dB
−5 arb. SPL in dB
1F 1R 1F+1R
0
−10
−15
−20
−0.5
0
0.5 1 1.5 Axial Source Position in m
2
2.5
3
−25 −1
−0.5
0
0.5 1 1.5 Axial Source Position in m
2
2.5
3
(b) Free-field Green’s-functions with consid- (c) Green’s functions with consideration of eration of the shear layer. the shear layer and the first reflections from the ceiling and two side walls.
Figure 12. Beamforming results, normalized to the forward rotor BPF level. The CROR center is at x = 0 m.
IV.C.
Modeling of the directive broad-band noise-sources with SODIX
Michel and Funke9 modeled the directive broad-band noise sources of a turbofan engine with the inverse method SODIX. The so generated source model was used to extrapolate the farfield directivity of the turbofan engine. The resulting directivities agreed within 1 dB for a large spectral and angular range. The method assumes directive source strengths Djm on a source grid which are modeled in order to minimize the difference F between the modeled cross-spectra and the measured cross-spectra. 2 M J X meas X ∗ F (D) = gjm Djm Djn gjn Γmn − m,n=1 j=1
(6)
meas is the measured cross-spectrum for the microphones m and n; gjm and gjn are the Green’s funcΓmn tions for the source j and the microphones m and n. SODIX will be used here for the modeling of the CROR broadband-noise sources. In contrast to the integrating beamforming algorithm, SODIX resolves the directivities of the sound sources by determining individual source strengths from each point source to each microphone. This makes SODIX an interesting method to investigate the accuracy of the shear-layer model and the room-acoustics corrections that were applied in this investigation. The broadband noise of the CROR baseline-configuration between 1F and 1R (see fig. 4) will be used for these evaluations. All tones were thereby eliminated from this frequency range. The linear source grid is placed on the engine axis with a distance ∆x = 0.2λTOB between the point sources. λTOB is the wavelength of the corresponding one-third octave-band center frequency. Figure 13 shows the SODIX results for three different implementations of the Green’s functions. The results in fig. 13(a) were calculated using freefield Green’s-functions. The magenta line shows the positions on the engine axis where the particular microphones ”see“ the source. This position is shifted downstream due to the shear layer refraction and it was calculated for each propagation path using the refraction angle ϕ, see fig. 5. The source strengths calculated with SODIX agree very well with the assumed source positions. The dynamic range of the results is well above 10 dB, despite the poor signal to noise ratio of SNR < 3 dB. A well separated source appears at x = 2.1 m that is directed towards the large emission angles, respectively the most downstream microphones that are close to one of the side walls. The strong reflection that comes from this side wall could be the reason for the artificial source. Figure 13(b) shows results from SODIX-calculations that take the shear-layer effects into account. The shear-layer corrected Green’s functions were calculated for all propagation paths, respectively from each source-grid point to each microphone with the model described in section III.B. The directive point sources are then modeled around the real rotor position x = 0 m. Artificial sources appear for small emission angles, respectively for microphones that are close to the upstream side wall. Strong reflections from the side wall can be the reason for these sources. The shear layer model after Amiet that was used to calculate the Green’s functions seems to fail for large emission angles. For angles Θ > 100◦ the modeled source strengths move to downstream positions. The results shown in fig. 13(c) have been calculated with the additional consideration of the reflections from the ceiling and the side walls, see section III.C. According to theory, the rotor noise peaks at 90◦ < Θ < 100◦ . The overall spatial resolution appears higher than for the calculations without the consideration of the reflections. Further, the artificial sources seen from the upstream and downstream microphones
9 of 11 American Institute of Aeronautics and Astronautics
80 70 −0.5
0
0.5 1 1.5 2 Axial Source Position in m
2.5
(a) Free-field Green’s-functions.
100 90 80 70 −0.5
0
0.5 1 1.5 2 Axial Source Position in m
2.5
120 Dynamic Range 10 dB
90
110
Emission Angle rel. to CROR center in Degree
100
120 Dynamic Range 10 dB
Dynamic Range 10 dB
110
Emission Angle rel. to CROR center in Degree
Emission Angle rel. to CROR center in Degree
120
110 100 90 80 70 −0.5
0
0.5 1 1.5 2 Axial Source Position in m
2.5
(b) Free-field Green’s-functions with consid- (c) Green’s functions with consideration of eration of the shear layer. the shear layer and the first reflections from the ceiling and two side walls.
Figure 13. Broad-band noise directivities modeled from the array measurements with SODIX (0◦ in flight direction). The source positions on the engine axis seen from the array are highlighted magenta for the different Green’s functions, the real CROR center is at x = 0 m.
are attenuated. Again, the shear-layer model that was used here, seems to fail for large emission angles. Although parts of the CROR broadband-noise sources will be convected by the wind-tunnel flow, broadband noise should also be detectable at the real rotor position for large emission angles.
V.
Conclusions
The acoustic change between different rotor configurations was shown with spatially averaged spectra. However, the measured sound-pressure levels showed strong local variations. The influence of the windtunnel shear-layer and the reflections in the test hall was investigated with array processing techniques and different implementations of the Green’s functions that describe the sound propagation. The localization of the CROR sound-sources was possible with classical beamforming. The dynamic range of the results indicates only slight phase fluctuations in the turbulent shear. The Amiet shear-layer model, which was used then, was investigated with the inverse method SODIX. The modeled source distributions were reasonable for the lateral directions, but the model seems to be insufficient for large emission angles when the sound waves propagate diagonally through the shear layer at downstream positions where it is already very thick. The reflection model that was implemented in the Green’s functions can not be considered to be very accurate in the relevant frequency range. The correction of individual microphone signals is therefore clearly questionable, but for the SODIX method an improved spatial resolution and the suppression of artificial sources was observed. Further improvements could be achieved with more accurate Green’s functions which are measured with the wind-tunnel flow.
Acknowledgments The authors would like to acknowledge the excellent cooperation with Snecma, TsAGI, and Airbus. The technical support by TsAGI deserves special thanks. The assistance by our student Markus Drescher during the measurements was also much appreciated.
References 1 Enghardt, L., Zhang, Y., and Neise, W., “Experimental Verification of a Radial Mode Analysis Technique Using WallFlush Mounted Sensors,” 137th Meeting of the Acooustical Society of America, Berlin, March 15-19, 1999, 1999. 2 Neise, W. and Enghardt, L., “Technology Approach to Aero Engine Noise Reduction,” Aerospace Science and Technology, Vol. 7, 2003, pp. 352–363. 3 Tyler, J. and Sofrin, T., “Axial flow compressor noise studies,” SAE Transactions, Vol. 70, 1962, pp. 309–332. 4 Parry, A. B. and Crighton, D. G., “Asymptotic Theory of Propeller Noise Part I: Single-Rotation Propeller,” AIAA J., Vol. 27, No. 9, 1989, pp. 1184–1190. 5 Amiet, R. K., “Refraction of Sound by a Shear Layer,” J. Sound Vib., Vol. 58, 1978, pp. 467–482. 6 Ingard, U., Noise Reduction Analysis, Jones and Bartlett Publishers, 2010. 7 Koop, L., Ehrenfried, K., and Krber, S., “Investigation of the systematic phase mismatch in microphone-array analysis,”
10 of 11 American Institute of Aeronautics and Astronautics
11th AIAA/CEAS Aeroacoustics Conference, 23-25 May 2005, Monterey, California, 2005. 8 Guidati, S., Brauer, C., and Wagner, S., “The reflection canceller - Phased array measurements in a reverberating environment,” AIAA-2002-2462, 2002, 8th AIAA/CEAS Aeroacoustics Conference, Breckenridge, Colorado, June 17-19, 2002. 9 Michel, U. and Funke, S., “Noise Source Analysis of an Aeroengine with a New Inverse Method SODIX,” AIAA-2008-2860, May 2008, 14th AIAA/CEAS Aeroacoustics Conference, Vancouver, BC, Canada, May 5-7, 2008. 10 Janardan, B. A. and Gliebe, P. R., “Acoustic characteristics of counterrotating fans from model scale tests,” AIAA 12th Aeroacoustics Conference, April 10-12, 1989, San Antonio, TX , 1989. 11 Magliozzi, B., “Noise Characterisitcs of a Model Counterrotating Prop-Fan,” AIAA 11th Aeroacoustics Conference, October 19-21, 1987/Sunnyvale, California, 1987. 12 Shivashankara, B., Johnson, D., and Cuthbertson, R., “Installation Effects on Counter Rotation Propeller Noise,” 13th AIAA Aeroacoustics Conference, 22-24 Oct 1990 , 1990. 13 Ricouard, J., Julliard, E., Omais, M., Regnier, V., Parry, A. B., and Baralon, S., “Installation effects on contra-rotating open rotor noise,” 16th AIAA/CEAS Aeroacoustics Conference, 7-9 Jun 2010 , 2010.
11 of 11 American Institute of Aeronautics and Astronautics
