Phased Array Aeroacoustic Measurements of an Unmanned Aerial Vehicle

Phased Array Aeroacoustic Measurements of an Unmanned Aerial Vehicle Alessandro DI MARCO 1 ; Lorenzo BURGHIGNOLI 1 ; Francesco CENTRACCHIO 1 ; Roberto CAMUSSI 1 ; Thomas AHLEFELDT 2 ; Arne HENNING 2 ; Jürg MÜLLER 3 1 University Roma TRE, Italy 2 German Aerospace Center (DLR), Germany 3 RUAG Aviation, Switzerland ABSTRACT This paper reports the results of the aeroacoustics study of an Unmanned Aerial Vehicle (UAV) experimentally investigated on a scaled prototype model. The measurement campaign were performed at the RUAG Large Subsonic Wind Tunnel in Emmen. The model was mounted on a ceiling strut in the 7-by- 5 meters test section. The aeroacoustic noise generated by the UAV was evaluated analysing the pressure fluctuations acquired through a phased array of 144 microphones installed on a flat-plate on the ground of the test section floor. Pressure fluctuations were acquired in different configurations. The main parameters varied during the tests were: the velocity of the air-flow, the incidence angle of the UAV and the presence of the landing gears. Flaps were always deployed. The directivity source contribution to the overall noise at various emission angles is determined as well, for each configuration, moving the array along the flow direction. Data were processed and then analysed in the frequency domain and using a conventional beamforming algorithm to retrieve the sound source map over the areas of interest. Source maps were deconvolved with a CLEAN-SC routine and single source contributions determined through integration of auto-powers. The resulting spectra were finally scaled to flight conditions. Keywords: Aeroacoustics, Beamforming, UAV: 21.6 1. INTRODUCTION An Unmanned Aerial Vehicle (UAV) is typically defined as an aircraft controlled autonomously or remotely without human aboard. UAVs use aerodynamic forces for vehicle lift and carry a payload (1). UAVs can be deployed quickly, operated in remote area where missions are not suitable or dangerous for manned aircraft and have relatively low operating cost. Depending on their use, UAVs range from the size of an insect to that of a commercial airliner and fly at different altitudes. Initially, the UAV development and employment were commonly associated with defence applications. Nowadays they are also increasingly found in civilian uses. UAVs are used to give real-estate buyers a better view of their property, to get packages to customers (Amazon Prime Air) or other civilian general-purpose applications such as environmental surveillance, aerial filming for agricultural management, photogrammetry and surveying (2). By virtue of their characteristics, it is expected that the growth potential of this aerial system will dramatically increase in the future (3) posing new questions for the implementation of legal regulations allowing UAV to share airspace with manned aircraft. Thus, some serious requirements will become necessary for the design and development of efficient and environmentally acceptable UAVs. One concern will be devoted to quiet technologies for reduced noise impact on community residents. In this respect, studies were concentrated on the propulsion system (4) and little attention 1 alessandro.dimarco@uniroma3.it 2 thomas.ahlefeldt@dlr.de 3 juerg.mueller@ruag.com 589

was paid on the airframe noise sources. To the author s best knowledge no works were made on the UAV airframe noise, and this motivated the present investigation. A parametric analysis on the noise generated by an UAV airframe is carried out in the closed test section of a low speed wind tunnel (WT). Wall pressure fluctuations measurement are performed with a phased array of microphones placed on the test section floor. As the data are contaminated by the hydrodynamic and acoustic contribution of the WT flow, a preliminary assessment experimental campaign on a known reference sound source is also performed. The results help to determine the aeroacoustic behaviour of the non-anechoic test section. Pressure data are analysed using beamforming techniques to locate potential sound source and are integrated over the area of interest to obtain their levels. The issue due to the Reynolds number dependency is not addressed in this paper. The paper is organized as follows. A description of the measurement campaign is given in the following section: it is divided into two sub-sections dealing respectively with the set-up of both the reference sound source and the UAV, and the processing of the acquired fluctuating pressure data. The results, in terms of spectra and source maps are presented in Section 3 for the reference sound source and the UAV. The final remarks are presented in Section 4, with future development. 2. MEASUREMENTS AND DATA PROCESSING 2.1 Set-up The data presented in this paper has been collected during a measurement campaign carried out in the RUAG Large Subsonic Wind Tunnel in Emmen. It is a continuous-flow closed wind tunnel capable of running at speeds up to and with a hard-walled 7-by-5 meters test section. It should be noted that it is not optimized for noise measurement. The initial collected data-set concerns the sounds emitted by a known reference source (a white noise and several pure tones at different frequencies), provided by the Institute of Aerodynamics and Flow Technology of DLR. The reference sound source is shown in Figure 1. Aerodynamic Fairings Metal Grid (a) (b) Figure 1: Sound source installed in WT test section (a) and its aerodynamic characteristics (b). The source is a ribbon loudspeaker, able to emit high sound pressure level in a frequency range up to (5,6). As it is thought to be used in in-flow applications, the sound source is recessed in a cavity and enclosed between symmetrical aerodynamic fairings. To prevent undesirable resonant peaks due to the cavity, a metal grid covered with silk is positioned over the loudspeaker. It is worth noting that the experimental campaign was assessed with the aim of characterizing the acoustic behavior of the WT, checking the acquisition system and validating the aeroacoustic results. The second test program was intended to investigate the mechanisms of sound generation on the airframe of an Unmanned Aerial Vehicle (UAV) in several different configurations (in terms of angles of attack and landing gear deployment) and velocities. The UAV is an old design prototype model 1:2 scaled provided by the RUAG. The UAV mounted on the ceiling strut of the RUAG WT section is shown below, in Figure 2. 581

(a) (b) Figure 2: UAV model mounted in the WT test section. The microphone array is also visible on the floor. The aeroacoustic measurements were carried out using an array of electret microphones, with microphone placement arranged as shown in Figure 3 with a maximum aperture of. The positive x-axis is defined in the direction of the air flow. The y-axis is in the cross-wise direction and the z-axis (not shown) is perpendicular to the WT test section floor. The fairing of the microphone array is made of aluminium, has a thickness of and dimensions of. The fairing edges are chamfered in order to reduce disturbances due to flow separation. All microphones are recessed behind a cone in order to reduce the influence of wall pressure fluctuations caused by the turbulent boundary-layer (TBL) and to improve the Signal to Noise Ratio (SNR)..8.6 Wind.4.2 x [m] -.2 -.4 -.6 -.8.8.6.4.2 Figure 3: Microphone locations as viewed from above. The microphone array is equipped with a carriage system to traverse it longitudinally across the floor of the WT test section. The correct positioning with respect to the model is obtained through a laser levelling system. The microphone were individually calibrated with a 423 Brüel & Kjær pistonphone before installation. In terms of frequency response for the individual microphones mounted in the array, a comparative calibration with a 1 4 inch G.R.A.S. condenser microphone was carried out in a semi-anechoic chamber before the measurement. This comparative calibration is the same as the one described in (7) and accommodates the radiation characteristic of the calibration source as well as the influence of the mounting scheme of the microphones in the array. The pressure response curves are corrected taking into account the low-pass filter applied during the measurements as described below. The microphone signals were simultaneously acquired using three GBM VIPER-48 multi-channel data acquisition units. The data acquisition is a A/D conversion system. The signals were sampled for a duration time of at a sampling frequency. All channels had an anti-aliasing filter at. In order to y [m] -.2 -.4 -.6 -.8 5811

reduce the low frequency noise due to the WT, a second-order high-pass filter (8) with adjustable cut-off frequencies was set at. In the post processing, the time signals were broken down into data blocks, with overlap and using a Hanning window. Each block was Fourier transformed resulting in frequencydomain signals with a narrowband frequency resolution of. 2.2 Data processing and extrapolation The conventional beamforming (9,1) algorithm is used in the generation of the noise source maps. It utilizes phase shifts between microphone array signals to improve the Signal-to-Noise Ratio (SNR), electronically steering the array focus to positions in space where the noise sources are presumably located. The expression of the conventional beamforming is, (1) where is the steering vector, the superscript H stands for Hermitian transpose and is the Cross Spectral Matrix (CSM). The source powers are calculated for the frequencies on a scanning plane at the grid points individuated by the vector. The assumed source model is an acoustic monopole that includes a sound convection correction. Since the boundary layer noise levels are often very high and they can affect the beamforming results, a common practice for studying aeroacoustic source of noise in closed test section wind tunnels is to remove the microphone self-noise contamination (11). This is done by removing (technically, zeroing out) the diagonal terms of (for details see Mosher (12)) and taking into account the new formulation of the steering vector. Except where otherwise specified the diagonal removal is always performed in the calculations. However, the cross-spectra may still contain incoherent noise due to the turbulent boundary layer; to overcome this problem the choice made of a long acquisition time brings positive benefits to the post processing results. The conventional beamforming technique together with the diagonal removal and eventually the spectral background noise subtraction (13) provides sufficient SNR to eliminate background noise and reverberations in closed test section (14,15). Hence, the noise sources below the background noise can be identified but, due to the number of microphones and the array configuration, the maximum dynamic range is usually limited to approximately (depending on frequency). Furthermore, the noise sources obtained applying traditional array processing methodologies are convolved with array beamforming response functions, which depend on array geometry, size and frequency (for more details, see Brooks and Humphreys (16)). Figure 4 gives the point spread function (PSF) for the array in Figure 3. Figure 4: The point spread function of the 144-microphone phased array of Figure 3. The reference sound source is fixed at the centre of the test section and the array is positioned on the test section floor at, where is the elevation/propagation angle. The figure is presented 5812

for a dynamic range of showing the low amplitude of the side-lobes and their spatial distribution for different the frequencies. The high frequency limit shown depends on the full-scale maximum frequency of interest and the model Scale Factor (SF); in this case, we have, the ratio of the maximum audible frequency and a. Given the set-up configuration and the array geometry, the PSF can be calculated and the array performances determined in terms of resolution and dynamic range. To achieve better performances, deconvolution approaches are one of the possible solutions. They can lead to an improvement of the SNR and contribute to suppress PSF side-lobe contamination in the final source map. Among the variety of algorithms developed in the past (the reader may refer to Ehrenfried et al. (17), Yardibi et al. (18) and Chu & Yang (19) for an overview) the CLEAN-SC was chosen. Such algorithm (2) is very suitable for processing microphone array data from closed wind tunnel test sections, mainly for three reasons: it is based on the spatial coherence between the main lobe and side lobes; it is able to mathematically decompose the cross-powers induced by the noise sources into its signal and noise parts; it is computationally faster with respect to other deconvolution algorithms. Sound pressure levels of the conventional and deconvolved maps are back-propagated to a reference distance of 1 m from the tested models. The scan grid used for the evaluation of the source maps consists of about points which are equally distributed with a spacing that was set depending on the analysed test model and the array resolution for the maximum frequency of interest. To quantify the noise level of a particular source, a power integration is used to compute the narrow-band and third-octave acoustic spectra generated from selected region on the model. The simplified method without auto-powers developed by Brooks & Humpreys (21) is used. This approach takes into account the diagonal removal performed on the CSM and reduce the computational time assuming the centre of the integration area as the representative point for the calculation of the normalization monopole response function PSF instead of summing all the possible PSFs over the scan grid. The integrated output of the area can be expressed by, (2) where B is the total number of grid points, P is the integrated power, is the source autopowers that take into account the diagonal removal. The subscript exp indicates the results from the experimental data. The subscript sim indicates the results from the simulated monopole source at the centre of the integration area. To avoid not physically meaningful contribution, grid point results with negative value or below the peak (22) are excluded from the integral calculation. Absolute source contributions using CLEAN-SC (2) are also calculated averaging the auto-spectra on the diagonal and compared with the source power integration (SPI) results. The full-scale flyover noise is finally obtained from the small-scale integrated spectra with methods as described in Allen & Soderman (23) and Allen et al. (9). Such method consist in correcting the data to simulate the flight conditions. Spectra are converted in Sound Pressure Level (SPL) with a reference pressure. The integrated spectra were scaled in frequency and extrapolated in amplitude to full-scale frequencies and the sound levels are reported to the actual flyover distance. The frequency adjustment is obtained as (14,23), where subscripts fs and ss correspond to full-scale and small-scale conditions. The full-scale extrapolation of the data implied the noise sources breakdown, since each source has a different scaling law (for details the reader may see e.g. Agrawal et al. (24), Dobrzynski et al.(25), Lockhard & Choudhari, 211 (26), Brooks & Humphreys, (27). The extrapolated sound amplitudes at full-scale (one unit distance far with respect to the noise source) are propagated through the application of a spherical spreading loss correction and inverse atmospheric attenuation model, obtained starting from the method described by Shields and Bass (28). 3. EXPERIMENTAL RESULTS The experimental campaigns has been conducted on both the reference sound source and the UAV scaled model presented in the previous sections. The relevant results are presented separately in the following two sections. 5813

3.1 Reference sound source The reference sound source was installed on the ceiling strut of the WT test section. In order to understand and evaluate the possible influence on the aeroacoustics results, a series of additional tests were conducted without the reference sound source and the support. The parameters varied during the experimental campaigns are summarized in Table 1. Specifically, different sound emissions were tested: three tonal noises and white noise (WN). Measurements with the loudspeaker off were made as well. The WT was run at four speeds between the maximum and half speed. To verify the directivity of the known source, the positions of the microphone array was varied from the flyover position ( ) to two extreme positions corresponding to the upstream ( ) and downstream ( ) end of the test section and to two intermediate positions ( ). Table 1: Reference sound source experimental campaign parameters. Sound Source Uninstalled Sound Source Installed WT speed, 3, 5, 6, 68, 3, 5, 6, 68 Propagation Angle -65,, +65-65, -47,, +47, +65 Sound Frequency -, WN, 6, 21, 41 Some tests were made with no-flow and the reference sound source turned on at the specified frequencies with the aim to verify the processing accuracy and the effect of the test apparatus on the acoustic field (no flow and no signal). The spectra obtained in this configuration from a microphone placed next to the array center together with the spectra at the maximum speed are presented in Figure 5. Figure 5: Electronic background noise and test sounds (array in flyover position) with WT not running. The hydrodynamic contribution due to the WT Turbulent Boundary Layer (TBL) to the broadband noise is clearly visible in the whole range of frequencies when comparing the two figures. The small peak at about is due to the vortex shedding downstream the ceiling strut. It was found comparing the spectra obtained at the same wind speed with and without the sound source installed (see Figure 6) and verifying the independence of the Strouhal number, not reported for the sake of brevity. It is also shown a broadband frequency decrease caused by the removal of the sound source and support. The source maps obtained from the noise test signals emitted by the loudspeaker are calculated following the procedure described in the previous section. Results are expressed in terms of Sound Pressure Levels propagated at 1m from the source and adopting a reference pressure. 5814

Figure 6: Spectra comparison with and without the sound source system installed (sound source off). The dynamic range of the source maps is. The flow is always from left to right. For the computation of the source maps, no shading was applied. This ensures the same array aperture and array characteristics for the comparison of the beamforming results. Unless stated otherwise source maps are calculated with the conventional beamforming algorithm, no decovolution is applied. The source maps for the loudspeaker emitting WN without wind flow and at maximum wind speed ( ) are presented in Figure 7, for comparison and validation of the beamforming algorithm. The array is at, and the sound source is clearly localized at the centre of the loudspeaker. For the dynamic range taken into consideration the maps are clean from the presence of other sources/sidelobes. (a) (b) Figure 7: Reference sound source maps at four third octave bands for the test case with WN: nowind case (a), and (b), both at. The source is correctly localized and the maximum sound levels are in good agreement except for the maximum third octave frequency band considered. The discrepancy could be attributed to the lower signal to noise ratio at high speeds and convection effects. The source maps obtained from the beamforming analysis are integrated through the source power integration method, CLEAN-SC integration is not reported for this case. The meaningful results are summarized in Figure 8 for a frequency range spanning from to. The spectra are presented as a function of the WT speed. For each speed, the position of the array microphone from the minimum to the maximum angle of propagation is varied. The growth of TBL and relative thickness contaminates the results over the whole frequency range confirming the higher influence at higher speeds. As the array is moved away from the central position, there is an 5815

overall decrease of the SPLs. This is due to the combined effect of both the directivity of the microphone and the reference sound source. Array in closer positions will lead to better agreement between the levels. 8 SPL at U nom =m/s Signal wn. 8 SPL at U nom =3m/s Signal wn. [db ref 2 Pa] 6 4 2 =-65 =-47 = =47 =65 1 4 [db ref 2 Pa] 6 4 2 =-65 =-47 = =47 =65 1 4 8 SPL at U nom =5m/s Signal wn. 8 SPL at U nom =68m/s Signal wn. [db ref 2 Pa] 6 4 2 =-65 =-47 = =47 =65 1 4 Figure 8: Sound source integrated spectra as a function of the array position and WT speed (WN signal). [db ref 2 Pa] 6 4 2 =-65 =-47 = =47 =65 1 4 3.2 UAV Two different configurations have been tested: 1) UAV with landing gears; 2) UAV without landing gears. The parameters varied during the test campaigns are reported in Table 2. The UAV flaps were always deployed with an angle of about. The rudders were also deflected at. For the UAV configuration with landing gears, the effect of the yaw angle was also tested but it is not reported since no comparative data for other model configurations is available. Table 2: Measurements parameters for the UAV test campaign. UAV with landing gears UAV without landing gears Angle of attack, +4, +8, +4, +8 WT speed 5, 6, 68 5, 6, 68 Propagation Angle, +32.5-65, -32.5,, +32.5, +65 An example of the UAV third octave sound source maps is shown in Figure 1. The maps are obtained for the two main configurations, with and without the landing gear, at the maximum WT speed in flyover position and zero incidence. This is the case with the highest SPL and the lowest SNR. The scan plane is the same in both maps. The maps have been deconvolved with CLEAN-SC to highlight the contributions due to the single sound sources. Two individual sound sources are clearly detected: the landing gears and the flap edge sides. Landing gears represent the strongest sound source; the flap edge side source is visible but attenuated with respect to landing gears, shown in Figure 9 (a). Removing the landing gears (see Figure 9 (b)) four more spots appear in the map, in correspondence of both the inner and outer gaps between the flap side-edge and the wing. The integrated spectra derived from the source maps (deconvolved with CLEAN-SC) are presented in Figure 1. Such spectra are obtained combining the effects of the landing gears deployment, WT speed and angle of attack. The microphone array is under the model at. The frequency range sweeps from to including the frequencies of interest. The peak due to the ceiling strut is present in all the configurations and it increases in frequency as the WT speed is increased. Two more peaks are visible in the low frequency range on both sides of the ceiling strut vortex shedding frequency. Their maximum SPL increases remarkably with the WT speed, and the reason of such phenomenon is still under investigation. 5816

(a) (b) Figure 9: UAV sound source maps comparison at (1/3-rd octave centre band) with landing gears (a) and without landing gears (b):,,. [db ref 2 Pa] 1 8 6 4 2 SPL at U=5m/s, = and =. with landing gears without landing gears [db ref 2 Pa] 1 8 6 4 2 SPL at U=5m/s, = and =8. with landing gears without landing gears [db ref 2 Pa] 1 3 1 4 1 8 6 4 2 SPL at U=68m/s, = and =. with landing gears without landing gears [db ref 2 Pa] 1 3 1 4 1 8 6 4 2 SPL at U=68m/s, = and =8. with landing gears without landing gears 1 3 1 4 1 3 1 4 Figure 1: Effect of the landing gears on the spectra varying the WT speed and model AoA. Landing gears introduce a broadband contribution for a higher frequency range depending on the WT speed considering the same AoA. It changes from to. As the AoA is increased this contribution tends to become lower. At two tonal components are detected that vanish at higher speed. Such components are probably related to modeled shock absorbers/oleo struts that connect the gears to the airframe. Regarding the directivity of the whole UAV, the spectra obtained from the deconvolved maps are presented in Figure 11 for the maximum WT speed and two AoA. The results show higher broadband contributions in the frequency range for positive propagation angles than for the negatives one (i.e. in front of the model): the effect is amplified for the maximum positive angle. In the low-frequency range,, there is a stronger contribution for negative propagation angle. The same trends are verified independently from the AoA. It has to be pointed out that the data was not corrected with weighting functions for the effective aperture correction and the same conclusions as in the preceding section have to be taken into account. 5817

1 SPL at = and U nom =68m/s without landing gears. 1 SPL at =8 and U nom =68m/s without landing gears. [db ref 2 Pa] 8 6 =-65 4 =-32.5 = 2 =32.5 =65 1 3 1 4 =-65 4 =-32.5 = 2 =32.5 =65 1 3 1 4 Figure 11: Effect of the microphone array position on the CLEAN-SC spectra. [db ref 2 Pa] 8 6 Finally, the obtained spectra have been scaled and extrapolated at different flight conditions (for more details see Section 2.2). As an example, let s consider the noise contribution related to the nose landing gear, shown in Figure 12, and the outer flap side-edge, in Figure 13. For both the cases, the WT measurements at with the array placed at flyover position ( have been scaled, starting from the CLEAN-SC integrated spectra, by means of suitable scaling laws, and subsequently extrapolated at for different distances. Figure 12: Nose landing gear spectra: small-scale ( ) vs. full-scale ( ) at different distances. Figure 13: Flap side-edge (outer) spectra: small-scale ( ) vs. full-scale ( ) at different distances. 4. CONCLUSIONS An extensive experimental campaign was conducted on a reference sound source and a 1:2 scaled UAV prototype installed in the RUAG large section wind tunnel facility in Emmen. The reference sound source and the acquisition system was provided by the DLR aerodynamic and flow technology institute, whereas the UAV model was available from RUAG Aviation. The objectives of the experimental campaign were to preliminary assess the aeroacoustic behavior of the wind tunnel achieved through tests on known sound signals and to perform a series of experiments on a complex aeroacoustic source for which a UAV model was used. The 5818

aeroacoustics was studied applying beamforming techniques to fluctuating pressure signals simultaneously acquired through a phased array of microphones. The array consisted of 144 microphones installed on a flat plate traversed on the WT test section floor. Different geometric and aerodynamic configuration were tested both for the reference sound source and the UAV. The main parameters for the reference sound source were the WT speed, the sound emitted by the source and the propagation angle. For the UAV, the parameters were the WT speed, the propagation angle, the model angle of attack and the model configuration. The flaps were always deployed at an angle of and the rudder at, being the set-up with landing gears in approach configuration. Data were post-processed and then analyzed with a conventional beamforming technique to obtain the source position on a scan grid plane corresponding to the areas of interest. The areas of interest were chosen as a function of the different configurations where the sound source were presumably located. The number of grid points and extensions were set accordingly and respecting the common practice in this kind of problems. Source maps were also deconvolved to clean the results from side lobes depending on the array PSF. For this, the wellknown CLEAN-SC algorithm was used. Integrated spectra were finally obtained and, in the UAV case, scaled to flight conditions. The main results obtained from the phased array analysis of the pressure fluctuations data are summarized depending upon the case. In both cases the aeroacoustics results are contaminated by the ceiling strut noise, revealed by the presence of a tonal Strouhal independent component in the pressure PSDs and due to the vortex shedding. For the reference sound source, the identification of the noise source position was successful. At high propagation angles the sound maps are stretched. The results showed a bit of discrepancy on the maximum sound levels in the high frequency range. This is probably due to the TBL of the WT that contaminates the signal with its hydrodynamic and acoustic contribution, the effect being higher as the WT speed is increased. The UAV test campaign allowed to identify the main sound sources, which are the three landing gears in tricycle arrangement and the flap edge sides. Beamforming maps deconvolved with CLEAN-SC and integrated scaled spectra confirmed known results from the literature, that the dominant airframe noise sources are the landing gears followed by the aerodynamic noise originating from the deployed high-lift devices. Landing gears contribution spans in a higher frequency range as the WT speed is increased and it is more pronounced at lower AoA. Tonal frequencies associated with the connection between the gear and the airframe are detected for the lower WT speed and the higher AoA. Directivity measurements showed a frequency range dependency being a function of the propagation angle. Positive angles significantly influence the high frequency range. The spectra are finally scaled to flight conditions showing the behaviour of two single sources as a function of the observer distance. Data from the test campaigns are still under processing to clarify the sound sources generation. Future work will be devoted to the research of new methods: to correct the data for the quantification of sound source levels at high angles of propagation and to mitigate the effect of the TBL in closed WT aeroacoustic testing. This will lead to a better understanding of the complex phenomenology involved in aeroacoustic testing in closed wind tunnels through phased microphone arrays. ACKNOWLEDGEMENTS This experimental test campaign was founded through EU funded projects EASIER (JTI-CS- 213-2-GRA-5-8) and WITTINESS (JTI-CS-213-2-GRA-2-25). The authors sincerely acknowledge RUAG and DLR personnel for their essential support during the experiments. REFERENCES 1. Bone E, Bolkcom C. Unmanned Aerial Vehicles: Background and Issues for Congress. Congressional Research Service - The Library of Congress; 23. 2. Doherty P, Granlund G, Kuchcinski K, Sandewall E, Nordberg K, Skarman E, et al. The WITAS Unmanned Aerial Vehicle Project. In: Horn W, editor. ECAI 2 Proceedings of the 14th European Conference on Artificial Intelligence. Berlin; 2. p. 747 755. 3. Park J-W, Ro K. A Prototype Design, Test and Evaluation of a Small Unmanned Aerial Vehicle for Short-range Operations. In: AIAA 3rd Unmanned Unlimited Technical Conference, Workshop and Exhibit. American Institute of Aeronautics and Astronautics; 24-6536. 5819

4. Sinibaldi G, Marino L. Experimental analysis on the noise of propellers for small UAV. Applied Acoustics. 213 Jan;74(1):79 88. 5. Kroeber S, Ehrenfried K, Koop L. Design and testing of sound sources for phased microphone array calibration. In: Second Berlin Beamforming Conference (BeBeC), Berlin, Germany; 28. 6. Kroeber S, Ehrenfried K, Koop L, Lauterbach A, Henning A. Systematic Comparison of Microphone Array Measurements in Open and Closed Wind Tunnels. In: 16th AIAA/CEAS Aeroacoustics Conference. American Institute of Aeronautics and Astronautics; 21-3734. 7. Ahlefeldt, T., and Koop, L., Microphone-Array Measurements in a Cryogenic Wind Tunnel, AIAA Journal, Vol. 48, No. 7, 28, pp. 147-1479. 8. Ahlefeldt T. Aeroacoustic measurements of a scaled half-model at high Reynolds numbers. AIAA journal 213; 51(12): 2783-2791. 9. Allen C, Blake WK, Dougherty RP, Lynch D, Soderman P, Underbrink JR. Aeroacoustic Measurements. Mueller TJ, editor. Berlin, Heidelberg: Springer Berlin Heidelberg; 22. 1. Sijtsma P. Experimental techniques for identification and characterisation of noise sources. In: Advances in Aeroacoustics and Applications, VKI Lecture Series; 24. p. 15-19. 11. Sijtsma P. Acoustic beamforming for the ranking of aircraft noise. Executive summary, NLR, 212. 12. Mosher M. Phased arrays for aeroacoustic testing - Theoretical development. In: Aeroacoustics Conference. American Institute of Aeronautics and Astronautics; 1996-1713. 13. Vaseghi SV. Advanced digital signal processing and noise reduction. John Wiley & Sons; 28. 14. Soderman P, Kafyeke F, Boudreau J, Burnside NJ, Jaeger SM, Chandrasekharan R. Airframe noise study of a Bombardier CRJ-7 aircraft model in the NASA Ames 7- by 1-foot wind tunnel. International Journal of Aeroacoustics. 24 Jan 1;3(1):1 42. 15.Liu Y, Dowling AP, Shin HC. Measurement and Simulation of Surface Roughness Noise Using Phased Microphone Arrays. Elsevier Journal of Sound and Vibration 28, 314 (1-2), pp. 95-112. 16. Brooks TF, Humphreys WM. A deconvolution approach for the mapping of acoustic sources (DAMAS) determined from phased microphone arrays. Journal of Sound and Vibration. 26 Jul;294(4 5):856 79. 17. Ehrenfried K, Koop L. Comparison of Iterative Deconvolution Algorithms for the Mapping of Acoustic Sources. AIAA Journal. 27;45(7):1584 95. 18. Yardibi T, Zawodny NS, Bahr C, Liu F, Cattafesta LN, Li J. Comparison of Microphone Array Processing Techniques for Aeroacoustic Measurements. International Journal of Aeroacoustics. 21 Jul 1;9(6):733 61. 19. Chu Z, Yang Y. Comparison of deconvolution methods for the visualization of acoustic sources based on cross-spectral imaging function beamforming. Mechanical Systems and Signal Processing. 214 Oct 3;48(1 2):44 22. 2. Sijtsma P. CLEAN based on spatial source coherence. International Journal of Aeroacoustics. 27 Dec 1;6(4):357 74. 21. Brooks T, William Humphreys J. Effect of directional array size on the measurement of airframe noise components. In: 5th AIAA/CEAS Aeroacoustics Conference and Exhibit. American Institute of Aeronautics and Astronautics; 1999-1958. 22. Oerlemans S, Broersma L, Sijtsma P. Quantification of airframe noise using microphone arrays in open and closed wind tunnels. International Journal of Aeroacoustics. 27 Dec 1;6(4):39 33. 23. Allen C, Soderman P. Scaling and extrapolating small-scale in-flow wind tunnel jet noise to full-scale flyover jet noise. In: 3rd AIAA/CEAS Aeroacoustics Conference. American Institute of Aeronautics and Astronautics; 1997-162. 24. Agrawal A, Lin K, Sawhney HS, Shim K. Fast Similarity Search in the Presence of Noise, Scaling, and Translation in Time-Series Databases. Proceedings of the 21th International Conference on Very Large Data Bases (49-51); 1995. 25. Dobrzynski W, Chow LC, Smith M, Boillot A, Dereure O, Molin N. Experimental Assessment of Low Noise Landing Gear Component Design" 15th AIAA/CEAS Aeroacoustics Conference (3th AIAA Aeroacoustics Conference); 29. 26. Lockhard DP, Choudhari MM. Variation of Slat Noise with Mach and Reynolds Numbers. 17th AIAA/CEAS Aeroacoustics Conference; 211. 27. Brooks TF, Humphreys WM. Flap-edge aeroacoustic measurements and predictions. Journal of Sound and Vibration. 8/2; 261 (23). 28.Shields FD, Bass HE. Atmospheric absorption of high frequency noise and application to fractionaloctave bands. NASA-CR-276; 1977 Jun. 582