European Congress on Computational Methods in Applied Sciences and Engineering ECCOMAS 2004 P. Neittaanmäki, T. Rossi, S. Korotov, E. Oñate, J. Périaux, and D. Knörzer (eds.) Jyväskylä, 24 28 July 2004 LANDING GEARS AERODYNAMIC INTERACTION NOISE Werner Dobrzynski *, Michael Pott-Pollenske *, Dave Foot and Michael Goodwin * Deutsches Zentrum für Luft- und Raumfahrt e.v. (DLR) Institut für Aerodynamik und Strömungstechnik Lilienthalplatz 7, 38108 Braunschweig, Germany e-mail: werner.dobrzynski@dlr.de, web page: www.dlr.de Aerodynamic Design & Data Domain Airbus UK Filton, Bristol BS 99 7AR, Great Britain e-mails: dave.foot@airbus.com, michael.goodwin@airbus.com Key words: Landing gear wake flow, Airframe noise, Landing gear noise, Interaction noise. Abstract. Airframe noise is generated as a result of the interaction of turbulent flow with different airframe components, i.e. the high lift devices and landing gears in particular, and may dominate over engine noise in the approach phase of large commercial aircraft. This paper describes the landing gears interaction noise research work in the EC co-financed project Significantly Lower Community Exposure to Aircraft Noise (SILENCER). Systematic steady and unsteady landing gear wake measurements were performed on differently scaled A340 type centre and main landing gears. The analysis of wake measurements showed almost isotropic turbulence characteristics for short distances downstream. Noise measurements were performed for a main landing gear being exposed to the wake flow of a centre landing gear. It turned out that only low frequency interaction noise is likely to be radiated while at high frequencies a noise reduction is obtained due to reduced mean wake flow speed.
1 INTRODUCTION Due to the successful efforts in the past to reduce engine noise, aerodynamic noise originating from the turbulent flow around the entire aircraft structure, i.e. high lift devices and landing gears in particular, tends to dominate in the approach phase [1, 2]. While high lift devices noise is most important for narrow-body aircraft, wide-body aircraft need large landing gears and thus often suffer from related aerodynamic noise. Stretched wide body aircraft may even need additional centre landing gears (CLG). In case of an unfavourable arrangement of main and centre landing gears excess aerodynamic noise is expected to be generated due to the impingement of the turbulent wake flow from the upstream located CLG onto the downstream located main landing gear (MLG). Therefore gear positions must be properly chosen in order to avoid such detrimental effects. Since aerodynamic noise from such complex 3D landing gear structures can still not be described numerically, a dedicated aeroacoustic experiment was set up. The objectives of this study were: Quantification of interaction noise characteristics and Definition of low noise design guidelines. To quantify this type of excess noise and develop design rules for low noise landing gear arrangements, scale model landing gear wake measurements were conducted in the Airbus- UK Filton low speed wind tunnel (LSWT) and in the Aeroacoustic Wind Tunnel Braunschweig (AWB) of DLR. In the latter additional noise measurements were performed. Wake measurements were performed for gear models of different scale in order to obtain an extended data base to develop/improve interaction noise prediction schemes [3] based on physical grounds and identify possible Reynolds number effects. 2 EXPERIMENTS The gear model is an A340-300 CLG in its fully extended position. The landing gear model fidelity was determined by the smallest gear replica (1/10 scale). As is presented in Fig. 1, all the principal gear features were reproduced, i.e. main fitting, drag stay, torque links, wheels, tyres and doors. Tests were performed with the gear in its baseline configuration and with the bay doors and/or the aft door removed, respectively (see Fig. 1). The undercarriage bay cavity was not represented. Transition tripping (all configurations) was confined to the cylindrical components of the CLG, (i.e. main leg and drag stay) using zig-zag-tape, positioned at 45 along the cylinder relative to the airflow direction. Tripping was applied to both the 1/3 scaled and the 1/10 scaled CLG models. Fig. 1 Fidelity of 1/10 scaled CLG model 2
Wake measurements were performed on a 1/3 scaled model in the Filton LSWT and on a 1/10 scaled model in the AWB. 2.1 Filton LSWT wake measurements on a 1/3 scaled CLG The Wake Traverse Rig assembly was installed at the downstream end of the test section. The pitch / yaw rake consisted of 12 x 5-hole probes, spaced at 0.04 m pitch on a vertical rake. The bottom probe in its lowest position was approximately 40 mm from the tunnel floor, and the area of tunnel scanned was 1.6 m wide x 1.44 m high from floor. Pressures were measured using a 64- channel PSI module rated at 5 psi. The test set-up is depicted in Fig. 2. Wake data were taken at 4 different axial positions downstream of the CLG assembly relative to the wake traverse rig, i.e. for X/D = 3, 4, 5 and 7.5 with wheel diameter D. Tests were conducted for a Fig. 2 Filton low speed wind tunnel 1/3 scaled CLG wake test set-up range of tunnel velocities V = 18, 40, 60 and 70 m/s. The dynamic pressures from each 5- hole probe were measured to determine local flow angularity and velocities in X, Y and Z directions. DLR performed unsteady wake measurements by means of a rake, consisting of 4 cross hot wire sensors with a pitch of 0.04 m. Data were taken for the same downstream positions where steady wake data were acquired. Scan planes were 1.24 m in width and 0.66 m in height with a grid point distance of 0.04 m. Because of rake vibrations at high wind tunnel speed, in contrast to the preliminary test matrix, the highest velocity (70 m/s) was reduced to 60 m/s, thus data were taken for velocities of 20, 40 and 60 m/s, respectively. Also unsteady surface pressures were acquired at selected positions on the gear. 2.2 AWB wake measurements on a 1/10 scaled CLG The experiments were performed with the CLG model mounted on a flat plate, which was adapted to the upper wall of the wind tunnel nozzle. Fig. 3 shows the installed landing gear in a side view. Because the Filton data analysis showed that no correlation could be detected between the different unsteady surface pressure sensors, only one unsteady surface pressure sensor of type Kulite LQ-3H-062-15A was installed in the front area of the main fitting at the same position as was selected for the 1/3 scaled gear model. Data were acquired up to 20 khz. 3
Fig. 3 AWB wind tunnel 1/10 scaled CLG wake test set-up The position of the one unsteady pressure sensor is depicted in Fig. 4. Steady and unsteady wake measurements were performed by means of a rake, consisting of 4 cross hot wire sensors with a probe to probe distance of 0.020 m. Studies showed, that a closer distance of the probes, which would have been desirable due to the geometrically small model scale, is not practicable because of mutual disturbances. Data were acquired at similar wind speeds and axial positions as was defined in the 1/3 scaled tests, i.e. for V = 20, 40 and 60 m/s and separation distances of X/D = 3, 4, 5. The largest distance of X/D = 7.5 could not be realized due to mechanical constraints. Fig. 4 Location of the unsteady pressure sensor on the main fitting 3 WAKE FLOW CHARACTERISTICS 3.1 Steady state wake characteristics In order to understand the wake effects it is best to look at the results as obtained for the 1/3 scaled model at the closest measurement plane and the highest free-stream velocity. Fig. 5 shows the axial velocity distribution (V x ) as a function of free-stream velocity (V = 70 m/s) with X/D = 3. (In contrast to the test configuration in the following graphs the gear is shown for its real orientation). The largest deceleration of the flow is behind the upper leg (main fitting) area, where the velocity ratio gets to around 0.4. Most flow blockage is in the region above the wheel axle and between the bay doors, where the average velocity ratio is in the region of 0.65. There is also a noticeable region of deceleration, circular in shape, below the axle and between the tyres. Fig. 6 is a plot of the span-wise (horizontal) velocity ratio (V y / V ). The span-wise flow pattern is nearly symmetric. Strongest span-wise flows are in the inboard direction behind the wheels and behind the small gaps between upper wheels and bay doors. The peak V y / V is close to ± 0.1 (i.e. ± 7 m/s). Fig. 7 shows the distribution of velocity ratio in the vertical direction. A strong upward flow component (V z / V = -0.2) was measured behind the lower leg (just above the axle and between the wheels). Three areas of downward flow were identified downstream of the bay door / upper wheel gap on both sides and below the axle. In these areas the V z / V ratio is close to 0.1. Results for all 4 measurement planes show that the wake decay is rapid compared to, for example, typical wake decay behind a wing. The landing gear is a relatively bluff body in the 4
flow inducing negligible lift thus the overall energy transferred into the wake is local and not sustained. The wake energy level is found to be near zero at a distance of 5-wheel diameters downstream of the model. It can be stated that the non-dimensional wake properties are very similar for the 3 main free-stream velocities tested (V = 40, 60, 70 m/s). While the removal of the aft door did not result in a marked change of the gear wake characteristics, a subtle change to the shape and extent of the axial (V x ) velocity distribution is obtained when the bay doors were removed. The discussion of corresponding detailed results, however, is beyond the scope of this paper. RUN 53824 BAY DOORS ON AFT DOOR ON ART. DETAIL OFF V = 70 m/s X/D = 3 Fig. 5 Distribution of axial velocity component V x (view against flow direction) 5
-VY +VY RUN 53824 BAY DOORS ON AFT DOOR ON ART. DETAIL OFF V = 70 m/s X/D = 3 Fig. 6 Distribution of lateral velocity component V y (view against flow direction) RUN 53824 BAY DOORS ON AFT DOOR ON ART. DETAIL OFF V = 70 m/s X/D = 3 -VZ +VZ Fig. 7 Distribution of vertical velocity component V z (view against flow direction) 6
3.2 Unsteady wake characteristics To determine the characteristics of the turbulent wake flow, unsteady velocity data were acquired up to frequencies of 20 khz. In Fig. 8 the overall turbulence intensity is plotted for the baseline test case at free stream flow velocity of 60 m/s and the closest CLG position, X/D = 3 (i.e. X = 1.37 m). In this graph the gear is shown in its inverted test position. A similar turbulence distribution is obtained as was shown above for the corresponding distribution of axial mean velocities (see Fig. 5). Highest turbulence intensities are observed downstream of the main fitting with values of up to 35%. Turbulence intensities up to 23% occur in the circular region above the tyres. Although the Filton wind tunnel is a low turbulence wind tunnel, turbulence values up to 6 % were measured at the boundaries of the measurement plane. This result is due to low frequency fluctuations as a result of sensor rake vibrations. The spatial development of the wake flow in terms of overall turbulence intensity (20 Hz to 20 khz) is presented in Fig. 9 in terms of lateral traces across the gear wake (for Z = 900 mm, see Fig. 8) for different downstream measurement planes. Two characteristic turbulence peaks are observed at lateral positions (Y = ± 50 mm), which could be associated with vortex shedding off the main fitting (also visible in Fig. 8). These peaks are preserved even for far downstream positions where the overall turbulence level has decreased from 35% at X/D = 3 (X = 1.37 m) to 15%.at X/D = 7.5 (X = 3.5 m). An estimate of the isotropy of measured wake turbulence can be provided through a comparison with the von Karman power spectrum of normalised isotropic flow turbulence. For this comparison the integral length scales of turbulent flow fluctuations must be determined from the autocorrelation function. Fig. 9 Fig. 8 Overall turbulence intensity for V 4 = 60 m/s and X/D = 3 Lateral distribution of turbulence intensity for different downstream separation distances 7
The result of this analysis is presented in Fig. 10. Normalised spectra of wake flow turbulence (with integral length scale L) exhibit good agreement with the von Karman spectrum. This is true for all downstream test positions X. Therefore it can be stated that gear wake flow turbulence can be considered isotropic even for the most upstream measurement plane at X/D = 3 (X = 1.37 m). Fig. 10 Normalized turbulence spectra for all positions downstream of the CLG 3.3 Comparison of test results for different model scales The comparison of wake flow characteristics as acquired downstream of CLG models of similar fidelity but different scale provides the following results: Mean flow and turbulence wake patterns are similar Axial mean flow deficit downstream of the main fitting is less pronounced at the smaller model scale Turbulence decay with increasing separation distance behind the gear is similar Normalised turbulence spectra compare reasonable well with von Karman spectrum of isotropic turbulence For both gear models unsteady surface pressures were measured at the upstream facing surface of the main fitting, which is exposed to the wake flow of the drag stay (see Fig. 4).. The analysis of this data showed good agreement between corresponding unsteady surface pressure spectra if those are normalized on a Strouhal number basis accounting for the geometrical model scale. 8
4 INTERACTION NOISE CHARACTERISTICS Finally interaction noise measurements were performed to (i) quantify the effects on landing gear aerodynamic noise if a downstream installed main landing gear would be exposed to the unsteady wake flow of an upstream located centre landing gear and to (ii) develop design rules for low noise landing gear arrangements. Dedicated noise tests were performed for 1/10 scaled CLG and MLG models in the AWB. 4.1 Test set-up and measurement techniques According to the objectives of this test the following general test set-up features were realized: The gear bay was not reproduced. The fuselage underbelly was not reproduced. The gear inclination corresponds to a zero aircraft angle-of-attack. No particular wind tunnel boundary layer condition is required (no wind tunnel boundary layer scoop). Interaction noise measurements were performed for 3 different stream-wise ( x) and 4 different lateral ( y) separation distances between the two gear s main legs, referenced to the upstream CLG s wheel diameter D (Fig. 11). For these altogether 12 different gear arrangements noise data were taken for 3 different wind tunnel speeds, i.e. 20; 40; and 60 m/s. To support the model gears the upper wind tunnel nozzle wall was extended by means of a flat plate to which the gears were attached. While the downstream MLG was mounted at a fixed position the upstream CLG could be attached at different lateral and streamwise positions. Flow Upstream Centre Landing Gear Test Parameters: Streamwise separation: x / D = 3; 4 and 4.7 Lateral Separation: y / D = 0; 0.5; 1.0 and 2.1 Flow speed: Fig. 11 Gear Wake v = 20; 40 and 60 m/s The CLG and the MLG typically feature vastly different lengths due to their different positions at the aircraft (i.e. under the fuselage or under the wing, respectively). In order to simulate the typical wake flow interaction conditions for the flat plate test set-up, the CLG leg was extended accordingly. Since only the wheel/bogie wake interaction condition was of interest, the MLG leg door was removed for this interaction noise study (Fig. 12). Farfield noise data were taken by means of 5 out-of-flow installed microphones, distributed underneath the gears covering a range of polar radiation angles between 56 and 140 (Fig. 13). For noise source localisation an acoustic mirror was applied (not shown in Fig. 13). D x Interaction noise test configuration Downstream Main Landing Gear y 9
Noise data were acquired for frequencies up to 25 khz. In order to determine the noise source characteristics from wind tunnel test data, the following corrections were applied to farfield measured noise data: Background noise correction, shear layer correction and source convection, microphone directivity, atmospheric absorption and convective amplification. Finally farfield noise levels were referenced towards a constant radiation radius. From the data analysis it turned out that a sufficiently high signal to noise ratio was obtained for frequencies above 100 Hz. Fig. 14 illustrates the effect of velocity on either the noise originating from the CLG or MLG, respectively. For frequencies above 600 Hz both noise level spectra exhibit a frequency pattern of successive level peaks and valleys which do not change with velocity. This phenomenon can be explained through reflection effects from the supporting wall. Different from a real landing gear structure and it s correspondingly extended source distribution in the test case under consideration the noise sources are concentrated in the Fig. 12 Arrangement of landing gear models in the AWB respective wheel bogie area so that reflection caused interference effects are more pronounced and do not smear out. For the CLG alone case an additional tone mechanism occurs, which follows a constant Strouhal number relationship (f/v = constant): this is related to wake vortex shedding from the extended centre landing gear leg. In order to quantify the effects on noise of gear-wake/gear interaction phenomena, the sum of the sound energies as radiated from both the individually installed CLG and MLG was taken as relevant noise reference. For both the CLG and the MLG, noise spectra were found to scale on a Strouhal number basis and levels to increase with flow speed corresponding to a 6 th -power law. However, due to reflection interference effects a normalized data representation would feature strong level variation (see Fig. 18), which would make the data interpretation more difficult. Therefore the discussion of interaction noise effects will be based on the results as obtained for the highest test speed (60 m/s) only. Flow CLG ϕ MLG M1 M2 M3 M4 M5 Farfield microphones Fig. 13 Schematic of test set-up and microphone positions 10
f/v = constant f = const f = const r/2 r 0.72 m Fig. 14 Effect of flow speed on CLG and MLG alone aerodynamic noise Figs. 15 to 17 present typical results of this study in terms of measured (and corrected) noise spectra for different gear arrangements relative to the corresponding reference noise spectra. The following effects are found for the highest test speed of 60 m/s and forward radiation direction: Significant excess interaction noise is observed to occur at low frequencies (< 300 Hz) Low frequency interaction noise is highest for in-line gear arrangement and diminishes with increasing lateral separation distance: In-line gear arrangement results in a significant noise reduction for frequencies >500 Hz: High frequency noise reduction diminishes with increasing lateral separation distance. High frequency noise reduction is most pronounced for large axial separation distances. Similar results are obtained for different speeds and radiation angles. The noise reduction due to reduced wake velocity is highest in the forward arc but can still be observed also Y/D = 0 Fig. 15 Effect of axial separation distance on interaction noise spectra for Y/D = 0 11
for rear arc radiation directions. In an effort to explain these unexpected results two opposing effects must be looked at in more detail: From the gear wake survey it is known that at a downstream position of X/D = 3 the velocity deficit in the wake (see Fig. 5) corresponds to about 0.66 (local speed re inflow speed). Since aerodynamic noise intensity (dipole sources) is decreasing (or increasing) with the 6 th power of the local velocity this would provide a noise reduction at the downstream gear on the order of 11 db, for the inline gear arrangement in particular. Since the noise level originating from the upstream gear essentially is unchanged, a combined noise reduction on the order of 3 db can be expected. From the same wake survey it is also known that compared to free stream conditions - the turbulent kinetic energy in the wake flow is about 20 to 30 db higher. As a consequence a 20 to 30 db noise increase would be expected based on the assumption that there is a linear relationship between turbulent kinetic energy, correlation length scales and sound intensity. From these arguments one would expect a net noise increase on the order of say 20 db, which, however, is not confirmed through the experimental results. It can be stated that the effect of flow turbulence on noise increase is overestimated. One reason could be that the noise generation in a complex gear structure to a large extent is already due to inflow turbulence/solid body interaction effects. Therefore additional inflow turbulence will only lead to a limited noise increase, i.e. at gear components which so far were exposed to clean (low turbulence) inflow conditions. X/D=4.7 4 3 Y/D = 0.5 Fig. 16 Effect of axial separation distance on interaction noise spectra for Y/D = 0.5 Y/D = 1.0 Fig. 17 Effect of axial separation distance on interaction noise spectra for Y/D = 1 12
5. CONCLUSIONS This study was performed to quantify and characterise possible excess noise levels due to interaction effects between the turbulent wake of an upstream located centre landing gear onto a downstream installed main landing gear. The results obtained from scale model experiments show that small changes of the upstream gear configuration have no effect on the gear wake characteristics. An initial comparison between measured and calculated mean flow wake characteristics shows promising agreement. From testing of differently scaled gear models no Reynolds-number effect was found. Gear wake turbulence turned out to be almost isotropic even for the closest downstream distance, corresponding to three wheel diameters behind the CLG main fitting. Noise results are in contrast to what was expected. When transposed to full scale it can be concluded that no excess interaction noise of any practical relevance will occur, almost independent of the selected gear arrangement. In detail the following two major results were obtained: Maximum excess interaction noise of up to 6 db for full scale frequencies below 30 Hz. A broadband noise reduction on the order of 2 db is achieved for full scale frequencies above 100 Hz. As a consequence no low noise design rules can be established from the results of this study. However, a large experimental data base was generated which can be used to validate results from both CFD calculations of wake characteristics and semi-empirical interaction noise predictions. REFERENCES [1] W. Dobrzynski and M. Pott-Pollenske. Slat noise source studies for farfield noise prediction. AIAA/CEAS 2001-2158, 2001. [2] L. C. Chow and W. Dobrzynski. Landing gears airframe noise research study. 9th International Congress on Sound and Vibration, 2002. [3] M. Smith et.al.. Prediction method for aerodynamic noise from landing gears. AIAA/CEAS 98-2228, 1998. 13