American Institute of Aeronautics and Astronautics

1

Effect of a pylon on double stream jet noise from hybrid CAA computations

F. Vuillot1, N. Lupoglazoff2 , M. Huet2

ONERA, BP72, F92322, Châtillon Cedex, FRANCE

Double stream nozzle configurations are computed to evaluate the noise radiated to the far field. The method combines a LES simulation with an acoustic solver based on surface integral method. The presented works have been carried out in the framework of the AITEC national program and focus on the effect of a mast/pylon which is added to an axisymmetric configuration. The LES computations rely on hybrid meshes, which mix structured and unstructured grids, combining tetrahedra, pyramids and hexahedra. This approach has been developed in the past years at ONERA. The computations are performed on both the axisymmetric and the pylon configurations. The LES results are compared to measurements and show good agreement. The computed results for the far field noise indicate that the presence of the pylon produces an azimuthal dependence on the radiated noise, together with an overall reduction in directions located below the nozzle. These results are in good agreement with measurements performed in the course of the VITAL project.

I. Presentation and conditions of calculation

The respect of increasingly severe regulations concerning noise emission is a real challenge for any new aircraft. In particular, during take-off operations, jet noise is a major contributor to overall aircraft noise and much attention is focused on this subject. As aircraft and engine manufacturers need acoustic estimation in real life situation, the present effort consists in considering configurations of increasing complexity getting closer to actual "installed" engine configuration. In particular, the present paper is aimed at studying the effects of a mast/pylon on the radiated jet noise. This work has been supported by the French DGAC, through the national program AITEC, managed by Snecma, and makes use of aerodynamic and acoustic measurements performed in the course of the VITAL project 1-2. VITAL is a five-year project aiming to reduce aircraft engine noise and CO2 emissions, lead by Snecma and including 53 partners gathering all major European engine manufacturers: Rolls-Royce plc, MTU Aero Engines, Avio, Volvo Aero, Techspace Aero, Rolls-Royce Deutschland and ITP, and the airframer Airbus. Part of the results presented in this paper has been obtained under VITAL SP7 (Installation). The retained operating point is the "high power" (OP1) conditions which corresponds to a non zero flight velocity. The configuration that is studied concerns a high by-pass ratio double stream nozzle, equipped with a mast/pylon arrangement which penetrates inside the secondary stream. This configuration is compared to a reference nozzle which has the same lines but without any installation device (axisymmetric configuration). Figure 1 presents the two geometries. The computations presented here are made through a hybrid CAA approach developed at ONERA since 2005 3. This approach associates the CEDRE code for LES (Large Eddy Simulation) flow solution and the KIM code for the far field reconstruction through acoustic post-treatment of a large amount of stored data. Past works permitted to validate the hybrid LES/acoustic procedure against simple single stream nozzles, coplanar as well as short cowl, high by-pass ratio double stream nozzles 4-12.

Reference

Nozzle with mast/pylon

Figure 1: geometry of the reference axisymmetric and pylon configurations

1 Assistant director, CFD and aeroacoustics department, Senior Member AIAA 2 Research scientist, CFD and aeroacoustics department

16th AIAA/CEAS Aeroacoustics Conference AIAA 2010-4029

Copyright © 2010 by ONERA. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.

American Institute of Aeronautics and Astronautics

2

Data used to normalize the results are the secondary nozzle exit diameter, the primary stream exit velocity and the external stagnation pressure and temperature (Ds = 0.264 m; Up = 340.m/s; Po = 97717 Pa; To = 296.17 K). In recent years, LES has been used in several works to evaluate the noise radiated in the far field by jet flows representative of aircraft engines at take-off conditions e.g. 13-16. Quite often a hybrid CAA approach is used as the LES solution computed in the jet flow region and its near field is associated to an acoustic solver based on surface integral version of the Lighthill analogy. The LES approach permits to increase the Reynolds number of the simulations and to come closer to real life engine exhaust flows. However some care must be exercised to control numerical errors and models bias as the frequency range of the simulated unsteady flows is limited by the combined effects of the precision of the numerical schemes and the grid refinement. One of the present challenges of high Reynolds number flows is the maximum frequency that can be resolved by the combination of numerical accuracy and grid size. In simulations of real life complex geometries, the minimal grid size that can be used is under constraint from the over all grid size required to describe the geometrical complexities of the nozzle configurations that are under scrutiny. In a recent work, Bodard et al. 16 propose a combination of resolved and modelled scales to reconstruct a wider spectrum, so as to compensate for the limited frequency range of the LES simulations. In the present work, the objective is to use relatively coarse meshes, associated to parallel computing, to permit simulations of complex flows at an affordable computer cost (CPU time and memory) so that the computation duration, measured in wall clock unit, stays in terms of few days or weeks. As a consequence the grid sizing is aimed at the lower part of the spectrum (below St = 1) where most of the radiated acoustic energy lies. This approach has permitted to apply the hybrid CAA method to flows of interest to the industrial partners. In particular the numerical tools have been recently applied to study the effects of continuous or pulsed micro-jet on the jet flow development and radiated noise 17, 18.

II. Methodology The starting point of this study is the preliminary computation of the reference nozzle with a structured grid. The grid had been constructed following the guidelines employed in 10. The 3D grid is generated by rotating a 2D grid over 360 deg. A total of 61 azimuthal planes where used for this first grid while this number can be easily varied. Further, unstructured hybrid grids are constructed by importing a portion of the structured grid in the CENTAUR grid generator software. This is completed by the actual nozzle geometry and a tetrahedron mesh is generated around the geometry and the structured block. This permits to accommodate different nozzle geometries while using the same structured patch, leading to enhanced flow prediction accuracy compared to full unstructured grid. The patch is 20 Ds long; its radius varies from r/Ds ≈ 1 at the nozzle exit to r/Ds ≈ 4 at the end of the zone. Considering the pylon configuration it was decided to increase the number of azimuthal planes to 89 (50% increase) and a new 3D structured patch was generated. This structured patch also incorporates a particular treatment near the axis to avoid degenerated cells at this location. This 89 planes structured patch has been used to generate hybrid grids for both the reference and the pylon configurations. Table 1 summarizes the properties of the grids and Figs. 2 and 3 illustrate the grids. The hybrid grid for the reference configuration is not presented since it can be deduced from the presented ones.

Grid/conf. Reference

structured, 61 planes

Reference

hybrid, 89 planes

Pylon

hybrid, 89 planes

Number of elements 4 251 360 9 534 065 10 800 197

Number of hexaedra 4 251 360 3 622 208 3 463 290

Number of faces 12 884 996 22 866 132 25 305 979

Table 1: grid properties

The use of a structured patch presents the advantage to easily define the "integration" surfaces where the LES solution is to be stored for further acoustic treatment 3-6. Here these surfaces are simply defined as particular constant "J surfaces" and are naturally present in the final hybrid grid. In the present case these surfaces are defined so as to contain the pylon. Figure 4 illustrates the CFD hybrid grid and "integration surfaces" that were imbedded in the hybrid grid for the pylon configuration. The acoustic storage surfaces are placed inside the structured zone and comprise the pylon. Their radial and axial extents are similar to those tested before.

American Institute of Aeronautics and Astronautics

3

XZ cut plane

XY cut plane

YZ cut plane

YZ cut plane

Figure 2: Details of the hybrid grid for the pylon configuration

XZ cut plane

XY cut plane

YZ cut plane

YZ cut plane

Figure 3: Details of the initial structured grid for the reference configuration

American Institute of Aeronautics and Astronautics

4

Figure 4: Storage surfaces used for the far field noise radiation

Figure 5 presents the different locations were the far field microphones are placed. For the pylon configuration, the flyover (az270) and sideline (az214 and az326) microphone arcs are defined, so as to correspond to the actual position of the microphone placed in the CEPRA19 facility. Additional arcs are also considered: 90 deg. from the pylon (az0 and az180) and above it (az90). For the reference axisymmetric configuration several microphones are located at different azimuthal positions, around the jet axis, to allow averaging of the simulated power spectral densities and integrated levels. Typically 36 microphones, located every 10 degrees, are used for each angle of observation. Finally, observation angles are defined from 20 to 150 deg., counted from the jet axis (Fig. 5). They are 5 deg. apart in the forward arc and only 10 deg. apart in the rear arc.

Figure 5: Location of microphones in the far field for the nozzle with pylon. The LES computations are run in three phases, starting from rest. First run is performed at large time steps, to establish the main jet flow field. This run is followed by a run at a reduced time step to permit the flow instabilities to develop. Then the last run is dedicated to record the necessary data for the acoustic post-processing and for jet flow analysis (flow averaging and statistics). A constant time step of 2 µs has been used

American Institute of Aeronautics and Astronautics

5

for this run. This run is critical for subsequent analysis and in particular, the duration of acquired signal must be long enough to allow statistical treatment. Typical duration is in excess of 120 Ds/Up convective time units (Ds being the secondary nozzle exit diameter and Up the primary stream exit velocity). Following LES computations, the radiated far field noise is reconstructed by applying the KIM code to the flow solution stored on the surfaces. KIM code makes use of integral formulations and for jet noise applications, the Ffowcs-Williams and Hawkings porous surface formulation is used, as previous studies showed it leads to more reliable results than the Kirchhoff method, especially for hot jets.3–6 These studies also highlighted that closing the surface does not improve the prediction for the angles of interest. As mentioned above, the storage is performed on two distinct surfaces S1 and S2 to ensure that the radiated pressure levels do not depend on the position of a particular surface. Instantaneous field are stored every 10 aerodynamic time step. It corresponds to an acoustic time step ∆tac = 2 × 10-5 s and leads to a maximum resolved frequency of 25 kHz. As illustrated on Fig. 4, surfaces are kept open at both extremities; they start at the secondary stream exit and have an axial extent of 20 D. For the pylon configuration, the radii at nozzle exit are r/D = 0.9 for S1 and r/D = 1.3 for S2 while at the downstream end, their radial extents are respectively r/D = 3.8 and r/D = 4.3. Microphones are located 10 meters (38 Ds) from the nozzle exit. Once the stability of the far field results with respect to the surface location is verified, more precise computations are performed using the solution stored on one selected surface, including azimuthal averaging of the power spectral densities and overall sound pressure levels. Due to transients at the beginning and at the end of the signal received at microphones locations, only 92.88 ms of simulated time (120 Ds/Up) are available for analysis for each configuration. Signal are windowed with a Hanning window to compute Fourier transform; power spectral densities and integrated levels are then compared to experimental data on the range [200 Hz ; 25 kHz].

American Institute of Aeronautics and Astronautics

6

III. Grid comparisons The results obtained for the reference axisymmetric configuration, using the two grids are first compared. Figures 6 and 7 present the radial and axial profiles of the normalized mean (time averaged) axial velocity and resolved turbulent kinetic energy, for the two grids.

Figure 6: radial profiles of the mean axial velocity and turbulent kinetic energy for X = 2.7D, 4D, 6D, 8D, 10D,

12D. Reference configuration, for the structured (––) and hybrid (––) grids.

Figure 7: Axial profiles of the mean axial velocity and turbulent kinetic energy for r = -R, 0, +R, with R=Ds /2.

Reference configuration, for the structured (––) and hybrid (––) grids. From the left hand side of Figs. 6 and 7 it can be observed that the two grids, although they are very different (more than a factor 2 in the number of cells), provide very similar mean velocity fields. In particular very little difference can be observed on the potential core length which is estimated at Lc ≈ 6Ds. On these figures a noticeable effect of the grid can be observed on the wake just downstream of the central body. The wake of the central plug is more pronounced on the structured grid which exhibits degenerated cells in this region. From the right hand side of the figures, a visible effect of the grid construction on the velocity fluctuation is noted, especially downstream of the potential core region. The refined hybrid grid produces less turbulence along the axis which is seen as an improvement since the previous experience revealed that the structured grid construction with a reduced number of azimuthal planes, tended to overestimate the jet turbulence 10. The effect of the wake of the central plug is also visible on the turbulent kinetic energy computed along the axis which is greater with the structured grid. The decrease of the turbulent kinetic energy brought by the refined grid is also noticed at the r = ±Ds/2 positions. Far field noise spectra and integrated levels obtained from the aerodynamic simulations performed on the two grids are presented hereafter for the reference configuration without pylon. Those comparisons allow a direct investigation on the effect of the number of azimuthal planes on the radiated noise. The influence of the surface on the far field noise is detailed in section V for the grid with 89 azimuthal planes; comparisons show that results are stable and data from S1 are used for further comparisons. Thus, S1 is also considered for the investigation of

American Institute of Aeronautics and Astronautics

7

grid effect on radiated noise in this section. Computed noise is compared to the one obtained for a similar surface location with the 61 planes grid. The precise positions of both surfaces are given Table 2.

x/D = 0 x/D = 20 89 planes grid 0.90 D 3.8 D 61 planes grid 0.78 D 3.6 D

Table 2: Radial position of the storage surface S1 for the two considered grids. The methodology for the acoustic simulations has been detailed on section II for the 89 planes grid. The same method is used for the 61 planes grid; 100 ms has been stored for the post-processing and 57.5 ms are available for the spectral analysis. The computed integrated levels are represented Fig. 8 for the two grids. Levels obtained at 30 degrees are similar for both grids and roughly correspond to the peak radiation angle. With increasing angle, simulated noise from the 61 planes grid can be almost 2 dB higher than the one observed with the 89 planes grid, up to 90 degrees where the simulations give a similar noise prediction. Near 120 degrees, the pressure level difference is the opposite of the one observed below 90 degrees; the low azimuthal density grid leads to a lower noise radiation of about 2.5 dB at 130 degrees. The spectral modifications leading to those differences are visible Fig. 9.

Figure 8: Integrated levels obtained with the structured 61 azimuthal planes (–) and

the hybrid 89 azimuthal planes (–)grids.

(a) 30 degrees (b) 60 degrees

(c) 90 degrees (d) 120 degrees

Figure 9: Simulated power spectral densities obtained with the 89 azimuthal

planes grid (–) and the 61 azimuthal planes grid (–).

American Institute of Aeronautics and Astronautics

8

For all represented angles on Fig. 9, the low azimuthal density grid leads to stronger medium and high frequency levels, compared to the high density one. The decrease rate is similar between the two grids and the difference comes from levels that start decreasing at a higher frequency for the 61 planes grid. At 30 degrees the low frequency levels responsible of the main part of the radiated sound, f ≤ 1 kHz, are identical between the two simulations which explain the similar integrated levels. At 60 degrees, levels obtained with the 89 planes grid start to decrease for f = 800 Hz whereas they remain constant up to f = 2.5 kHz with 61 planes, leading to higher integrated levels for this last configuration. Spectra are similar at 90 degrees for low and medium frequencies, which is coherent with the very close integrated levels observed Fig. 8. For the last observed angle, 120 degrees, low and medium frequency levels are slightly lower for the 61 azimuthal planes grid, with the consequence of a lower integrated level of about 2 dB compared to the higher azimuthal density grid. This observation could seem erroneous because spectra look similar at low frequencies, but a computation of the integrated level on the range [200 Hz; 2 kHz] for the two configurations confirms this assumption. On this frequency range, the OASPL is 94.7 dB with the 61 planes grid and 97.15 dB with 89 planes. The higher frequencies even contribute to reduce the difference of the integrated noise between the two grids. Increasing the number of azimuthal planes in the aerodynamic grid of about 50 % thus leads to visible modifications on the radiated pressure. It corresponds to a decrease of the integrated levels below 90 degrees and an increase between 90 and 130 degrees, leading to a flatter directivity shape. Low frequency levels are not significantly modified below 90 degrees and slightly increased over this angle. The level decrease also appears for lower frequencies, leading to weaker high frequency levels. The overall effect of the refined grid is to reduce the radiated noise and this can be linked to reduced turbulence levels. Again this is viewed as an improvement as previous works concluded at the over estimation of far field levels.

IV. LES computations of the pylon configuration On the final LES run, 56000 time steps were made (following 40000 for the jet and flow instability developments) and exploited, corresponding to 140 Ds/Up. The calculations were run on ONERA Bull Novascale with 64 Itanium II processors, corresponding to a wall clock duration of the order of 360 hours.

Figure 10: Iso surfaces of instantaneous axial velocity (250, 300, 350 m/s). Comparison without and with pylon. Figure 10 presents an overall view of the jet flow for the reference and pylon configuration. On this figure, although the overall jet flows seem similar, one can notice a visible updraft effect on the configuration with the pylon installed. Figure 11 displays instantaneous views of the flow computed on the pylon configuration along two perpendicular planes. The updraft phenomenon is clearly visible and results in hot flow from the primary nozzle being entrained upward, toward the pylon. As a consequence, the Mach number if found to be higher in the area opposite to the pylon, where low temperature areas are formed as the cold ambient air is entrained and reduces the local value of the sound speed. The instantaneous pressure field is displayed on Fig. 12.

American Institute of Aeronautics and Astronautics

9

Instantaneous Mach number

Instantaneous temperature

Figure 11: Mappings in XY and XZ plans (with pylon).

Figure 12: Instantaneous pressure in XY and XZ plans (with pylon).

On Fig.12, the presence of the pylon clearly provokes a loss of symmetry in the coherent structures. Shed vortices are found to be strongly deformed leading to 180 phase shift across the secondary shear layers. Interactions with the central body can also be observed. Figure 13 proposes a pressure spectrum recorded in the secondary shear layer (at 90 deg. from the pylon) where several definite frequencies are visible and correspond to the early shedding of vortices, as well as pairing and complex interaction with the wake of the pylon. The Strouhal number associated with the observed frequencies are rather high (St in the range 2-4) indicating the early stages of instability development.

Figure 13: Computed pressure spectrum in the secondary shear layer (with pylon).

American Institute of Aeronautics and Astronautics

10

Instantaneous vorticity Mean turbulent kinetic energy Eck

Figure 14: Maps in the XY and XZ planes (with pylon). Figure 14 present the instantaneous vorticity field together with the mean turbulent kinetic energy maps. On this figure it can be noted that the secondary shear layer is the most instable and this can be linked to the difference of shearing magnitudes between the two shear layers. Indeed, for the considered operating point the velocity difference between the primary and secondary streams is somewhat reduced. On this figure, the wake of the pylon is clearly visible and produces a flow region of large (doubled) value of the turbulent kinetic energy. Measurements carried out in the VITAL project, permitted to make comparisons with the present numerical simulations, although limited measurements were performed for the considered operating point. Figure 15 proposes qualitative comparisons of the computed flow with available PIV measurements. Reasonable agreement can be observed. Figure 16 displays the axial plane maps and indicates the positions of the cross sections.

Figure 15: Mean axial velocity cross section (YZ cuttings). PIV images.

XY cut � XZ XZ cut � XY

Figure 16: Mean axial velocity and positions of the cross sections. NB: the XY plane considered in the numerical simulations corresponds to the XZ plan for the experimental campaign.

More quantitative comparisons can be made thanks to five hole probe traverses performed in the course of the VITAL project. These are presented on Figs. 17 and 18. On Fig. 17 good agreement is noted on the overall profiles while the experiment exhibits stronger heterogeneities. In particular the external boundary layer appears to be thicker in the experiment and the enhanced heterogeneities could be produced by upstream flow obstacles not reproduced in the computations. In addition, in the absence of detailed characterisation of upstream feeding

American Institute of Aeronautics and Astronautics

11

conditions, the computations use uniform and steady upstream boundary conditions which could be responsible for the observed differences.

Axial velocity

Total pressure Pi/Po

SECONDARY STREAM PRIMARY STREAM Figure 17: Radial profiles at the nozzle exits.

Farther downstream, around the 4Ds (Ds = 0,264 m) position, Fig. 18, the profiles aspect is well reproduced by the calculations, in particular the effect of the pylon and the resulting updraft mentioned above.

Axial velocity

Global pressure Pi/Po

XY CUT (pylon) AT X=4D XZ CUT (⊥) AT X=4D Figure 18: Radial profiles at the 4Ds position.

American Institute of Aeronautics and Astronautics

12

V. Far field noise radiation The instantaneous aerodynamic fields stored on the two surfaces surrounding the jet are used as input data to radiate the acoustic pressure fluctuations to the far field. In a first step, pressure time histories and power spectral densities obtained from the data of the two distinct surfaces are compared to ensure that the results do not depend on the location of the surface used. An illustration of these comparisons is given Fig. 19 for the configuration with pylon and for a microphone located 30 degrees to the jet axis at the sideline position.

(a) pressure time histories (b) power spectral densities

Figure 19: Illustration of the pressure signals from data stored on the two surfaces S1 (–) and S2 (– –). Configuration with pylon at sideline position, microphone located 30 degrees to the jet axis.

Figure 20: Computed directivities for the two surfaces S1 (–) and S2 (– –).

Configuration with pylon at sideline position. Both pressure time histories and power spectral densities obtained from the two distinct surfaces present good comparisons. Small discrepancies can be observed in the time evolution of the pressure signals, see Fig. 19 (a) and are caused by the more important attenuation of the medium and high frequency levels for S2, compared to S1. This observation is confirmed by the comparison of the power spectral densities, see Fig. 19 (b). Spectra well collapse for frequencies below 1 kHz, after what a more important decrease of the levels is noticed with data from S2. This difference is caused by the dissipation of the acoustic perturbations by the flow solver between the two surfaces. The cut-off frequency f = 1 kHz deduced from the comparison of the spectra corresponds to a Strouhal number St = 0.8, which is a value slightly higher than expected for the constructed grid. The comparison of integrated levels on the range [200 Hz ; 25 kHz], visible Fig. 20, confirms the previous observations made for spectra. Only small differences can be observed between the two surfaces and are caused by the medium and high frequency levels dissipation. It is thus clearly proved that the far field radiated pressure is almost independent of the position of the surface used for the storage. As a consequence, only results obtained from S1 are detailed and discussed in the following, for both configurations, without and with pylon.

American Institute of Aeronautics and Astronautics

13

First, the influence of the integration bandwidth is investigated at sideline position for two representative polar angles, 30 degrees and 90 degrees. Without modifying the lowest frequency of 200 Hz (St = 0.16), levels are integrated up to a maximum frequency fmax = 1 kHz (St = 0.8), which corresponds to the cut-off frequency of the grid, fmax = 4 kHz (St = 3.1) and fmax = 25 kHz (St = 19.4). Results are given Table 3.

30 degrees 90 degrees [200 Hz ; 1 kHz] -1.31 dB -5.88 dB [200 Hz ; 4 kHz] -0.14 dB -0.13 dB

[200 Hz ; 25 kHz] 0 dB 0 dB Table 3. Simulated integrated levels at sideline position with different frequency bandwidths.

The contribution of high frequency levels, greater than 4 kHz is clearly negligible at the two angles. It only leads to a small modification of nearly 0.15 dB. The influence of medium frequency levels is nevertheless different. At low angles, the contribution to radiated pressure is limited and corresponds to an increase of about 1 dB, which can be explained by the very strong low frequency levels dominating the radiation, as is can be seen on Fig. 21 (a). At 90 degrees, a level difference of 5-6 dB is noticed between the two bandwidths and is caused by the more important contribution of the medium frequency levels to the total noise. This computed spectrum is visible Fig. 21 (c).

(a) 35 degrees (b) 60 degrees (c) 90 degrees

Figure 21: Comparison of experimental (–) and computed (–) power spectral densities. Configuration with pylon, microphones at sideline position.

Computed spectra compare well with the experimental data. Illustrations of simulated and measured spectra are given Fig. 21 for the configuration with pylon, at sideline position. It can be clearly noticed that both shape and absolute levels are very similar up to the cut-off frequency. Tonal noises are numerically observed for frequencies f ~ 2 kHz, f ~ 3 kHz and f ~ 4.5 kHz. The origin of these tonal noises seems to come from an important vortex shedding in the early shear layer separating the secondary and external flows. Tonal noises have indeed already been noticed in the shear layer of the secondary stream, see Fig. 13. Such peaks have already been observed in other numerical simulations 13 with a similar methodology; the numerical methods used thus seem to overestimate this mechanism, compared to reality, which explains the fact that these tonal noises are not experimentally observed. The peak level frequency is observed at f ~ 400 Hz (St ~ 0.3), with a decrease of the levels at higher frequencies. Consequently, the dissipation observed above 1 kHz has only a small impact on the integrated values. This is confirmed by the comparison of experimental and computed OASPL, visible Fig. 22.

American Institute of Aeronautics and Astronautics

14

Figure 22: Experimental (sideline: ● ; flyover: ■) and computed (sideline: – ; flyover: –) OASPL on the range [200 Hz ; 25 kHz]. Configuration with pylon.

On this figure are also represented the measured and simulated levels at flyover position. For angles below 100 degrees, experimental sideline levels are up to 0.7 dB higher than the flyover ones; at higher angles, no differences are noticed. A similar but slightly amplified trend is observed in the simulations: near 60 degrees, levels at sideline position can be 2 dB higher than the flyover simulated ones. As visible on Fig. 23 and Fig. 24, comparisons of the spectra obtained at those two positions confirm that the level differences have the same origin in the experiments and the simulations.

(a) 35 degrees (b) 60 degrees (c) 90 degrees

Figure 23: Experimental power spectral densities at flyover (–) and sideline (–) positions.

(a) 35 degrees (b) 60 degrees (c) 90 degrees Figure 24: Simulated power spectral densities at flyover (–) and sideline (–) positions.

At 35 and 60 degrees, main spectral modifications correspond to lower medium and high frequency levels at flyover position, for both simulations and experiments. The more important noise difference between the two positions observed at 60 degrees in the simulations comes from the stronger low frequency levels at sideline, which are not observed experimentally. At 90 degrees, the only noticeable difference corresponds to slightly

American Institute of Aeronautics and Astronautics

15

higher levels for frequencies around f = 500 Hz at sideline position, observed both in measurements and simulations. All these results confirm the assumption that the pylon has a strong azimuthal effect on the radiated noise. Reference simulation without pylon also allows us to investigate the far field noise modification induced by the pylon itself, as visible on Fig. 25 for the integrated levels.

Figure 25: Simulated integrated levels without (–) and with pylon at flyover (–) and sideline (–) positions.

For almost all angles, the pylon reduces the integrated noise levels at flyover position, compared to the reference configuration. This observation is coherent with the experimental results of Bhat 19. A reduction is also visible at low angles for the sideline arc and is due to reduced low frequency levels (see fig. 26). For higher angles, levels with pylon have similar levels than the ones simulated for the reference case.

(a) 35 degrees (b) 60 degrees (c) 90 degrees

Figure 26: Simulated power spectral densities without (–)and with pylon at flyover (–) and sideline (–) positions. It is interesting to note that noise radiated just above the pylon appears to be reinforced by its presence, as already mentioned by Huber et al.2 This can be the result of the increased turbulence produced by the wake of the pylon as previously mentioned (see Fig. 14). Figures 27 and 28 below illustrates this phenomenon which is clearly evidenced by the computations.

(a) 35 degrees (b) 60 degrees (c) 90 degrees

Figure 27: Simulated power spectral densities with pylon at flyover (–) and sideline (–) positions and above the pylon (–).

American Institute of Aeronautics and Astronautics

16

Figure 28: Simulated integrated levels with pylon at flyover (–) and sideline (–) positions

and above the pylon(–).

VI. Conclusion and future works

The presented results demonstrated that relatively coarse meshes can be used to analyze the effects of installation devices, such as the pylon/mast configuration, on the radiated noise in the far field. Although the frequency range of the simulation is limited to St < 1, it permits to capture the main part of the spectrum of the radiated noise. The grid effect has been studied on a reference axisymmetric double stream nozzle and demonstrated a limited sensitivity of the numerical results to the grid design. However the increased azimuthal resolution of the finest grid was shown to improve both the aerodynamic results and the acoustic far field prediction. In particular, the previously observed over prediction of jet turbulence and far field OASPL could be reduced, thanks to a careful grid design that avoided too fast coarsening of the grid size in the jet region of the grid. Comparison with experimental results demonstrated that the absolute levels could be consistently approached by a few dB and that the method was able to predict the major effects produced by the presence of the pylon/mast configuration. In particular the presence of the pylon reduced the noise radiated below the engine while a significant increase was observed above the pylon. These results represent a major step in demonstrating the usefulness of hybrid CAA method in helping the aircraft and engine manufacturers in evaluating the impact of particular designs on the emitted noise. Further work will then be aimed at studying passive reduction devices such as chevrons and also more complex integration effects, such as the presence of wing or fuselage parts. However this latter step will necessitate meticulous works to ensure that the resulting grids preserve essential quality to capture the jet flow development that is at the origin of noise sources.

VI. Acknowledgments This work has been carried out in the framework of the national AITEC research project supported by the French DGAC and managed by Snecma. VITAL is a collaborative research project, running for five years, which aims to significantly reduce aircraft engine noise and CO2 emissions. It has a total budget of 91 million euros, including 51 million euros in funding from the European Commission. Part of the work presented in this paper was performed under WP 7.2 "Optimisation of aero-acoustic nozzle concepts" and ONERA specially contributed to the work presented in the paper. The authors are grateful to R. Davy from ONERA for his help with experimental data and to the VITAL project that permitted the use of these results for the present work.

VII. References [1] R. Davy, C. Brossard, J.M Jourdan, Y. Pioche, O. Piccin, "Installation Effects Characterization of VHBPR engine PART II: Experimental study using Particle Image Velocimetry", AIAA 2009-3253, 15th AIAA/CEAS Aeroacoustics Conference, 11 - 13 May 2009, Miami, Florida, USA [2] J. Huber, M. Omais, A. Vuillemin, R. Davy, "Characterization of Installation Effects for HBPR Engine Part IV: Assessment of Jet Acoustics", AIAA 2009-3371, 15th AIAA/CEAS Aeroacoustics Conference, 11 - 13 May 2009, Miami, Florida, USA [3] N. Lupoglazoff, A. Biancherin, F. Vuillot, G. Rahier, "Comprehensive 3D unsteady simulations of subsonic and supersonic hot jet flow-fields. Part 1: aerodynamics analysis", AIAA 2002-2599, 8th AIAA/CEAS Aeroacoustics Conference & Exhibit, June 17-19 2002, Breckenridge, CO.

American Institute of Aeronautics and Astronautics

17

[4] N. Lupoglazoff, A. Biancherin, F. Vuillot, G. Rahier, "Comprehensive 3D unsteady simulations of subsonic and supersonic hot jet flow-fields. Part 2: acoustic analysis", AIAA 2002-2600, 8th AIAA/CEAS Aeroacoustics Conference & Exhibit, June 17-19 2002, Breckenridge, CO. [5] G. Rahier, J. Prieur, F. Vuillot, N. Lupoglazoff, A. Biancherin, "Investigation of integral surface formulations for acoustic predictions of hot jets starting from unsteady aerodynamic simulations", AIAA 2003-3164, 9th AIAA/CEAS Aeroacoustics Conference & Exhibit, May 12-14 2003, Hilton Head, SC, USA [6] G. Rahier, J. Prieur, F. Vuillot, N. Lupoglazoff, A. Biancherin, "Investigation Of Integral Surface Formulations For Acoustic Post-processing Of Unsteady Aerodynamic Jet Simulations", Aerospace Science And Technology, Volume 8, Issue 6, September 2004, pp. 453-467. [7] F. Muller, F. Vuillot, G. Rahier, G. Casalis, "Modal Analysis of a Subsonic Hot Jet LES with Comparison to the Linear Stability Analysis", AIAA-2005-2886, 11th AIAA/CEAS Aeroacoustics Conference, 23-25 May 2005, Monterey, California, USA [8] N., Lupoglazoff, G. Rahier, F. Vuillot, "Application of the CEDRE unstructured flow solver to jet noise computations", 1st EUCASS, 4-7 July 2005, Moscow [9] F. Muller, F. Vuillot, G. Rahier, G. Casalis, E. Piot, "Experimental and Numerical Investigation of the Near Field Pressure of a High Subsonic Hot Jet", AIAA-2006-2535, 12th AIAA/CEAS Aeroacoustics Conference, 8-10 May, 2006, Cambridge, MA, USA [10] F. Vuillot, N. Lupoglazoff, G. Rahier, "Double-stream nozzles flow and noise computations and comparisons to experiments", AIAA-2008-0009, 46th AIAA Aerospace Sciences Meeting and Exhibit, 7-10 Jan. 2008, Reno, NV, USA [11] B. Fayard, G. Rahier, F. Vuillot, F. Kerhervé, "Flow field analysis for double stream nozzle: application to jet noise", AIAA-2008-2983, 14th AIAA/CEAS Aeroacoustics Conference, 5 - 7 May 2008, Vancouver, Canada [12] B. Fayard, G. Rahier, F. Vuillot, "Modal Analysis of Jet Flow from a Coaxial Nozzle with Central Plug", AIAA-2009-3355, 15th AIAA/CEAS Aeroacoustics Conference, 11 - 13 May 2009, Miami, Florida, USA [13] C. Bogey, S. Barré, D. Juvé, C. Bailly, "Simulation of a hot coaxial jet : direct noise prediction and flow-acoustics correlations", Phys. Fluids, Vol. 21, No. 3, 2009, 035105. [14] K. Viswanathan, M. Shur, P. Spalart, M. Strelets, "Flow and Noise Predictions for Single and Dual-Stream Beveled Nozzles", AIAA Journal, Vol. 46, No. 3, March 2008, pp. 601-626. [15] A. Uzun, M. Y. Hussaini, "High-Fidelity Numerical Simulation of a Chevron Nozzle Jet Flow", AIAA 2009-3194, 15th AIAA/CEAS Aeroacoustics Conference, 11 - 13 May 2009, Miami, Florida, USA [16] G. Bodard, C. Bailly, F. Vuillot, "Matched hybrid approaches to predict jet noise by using Large-Eddy Simulation", AIAA 2009-3316, 15th AIAA/CEAS Aeroacoustics Conference, 11 - 13 May 2009, Miami, Florida, USA [17] M. Huet, F. Vuillot, G. Rahier, "Numerical study of the influence of temperature and micro-jets on subsonic jet noise", AIAA 2008-3029, 14th AIAA/CEAS Aeroacoustics Conference, May 5-7 2008, Vancouver, Canada. [18] M. Huet, B. Fayard, G. Rahier, F. Vuillot, "Numerical Investigation of the Micro-Jets Efficiency for Jet Noise Reduction", AIAA-2009-3127, 15th AIAA/CEAS Aeroacoustics Conference, 11 - 13 May 2009, Miami, Florida, USA [19] Bhat, T. R. S., "Experimental study of acoustic characteristics of jets from dual flow nozzles", AIAA 2001-2183, 7th AIAA/CEAS Aeroacoustics Conference, May 28-30 2001, Maastricht, Netherlands.