DEVELOPMENT OF AN ACOUSTIC MEASUREMENT CAPABILITY FOR AUTOMOTIVE TESTING IN OPEN-JET WIND TUNNELS

Transcription

1 DEVELOPMENT OF AN ACOUSTIC MEASUREMENT CAPABILITY FOR AUTOMOTIVE TESTING IN OPEN-JET WIND TUNNELS Author: Brianne Y. Williams Advisor: Dr. Colin P. Britcher Old Dominion University, Norfolk, VA. 359 Automobile aeroacoustics are still in their beginning stages of understanding and controlling of the noise sources using mainly exploratory experimental methods and idealized models. The difficulty lies in the mechanisms that generate noise. Sound sources are generated from unsteady aerodynamics which arise from turbulent boundary layers and from regions of separated flow over the vehicle. These regions are subjected to unsteady pressures which are interpreted by the driver as undesirable noise. In order to reduce the noise around an automobile the individual sources are best identified by performing full-scale wind tunnel testing. This can be achieved by deploying a microphone phased array. A microphone phased array permits discrimination of low-level sources in the presence of diffuse background noise by means of strong spatial variations in array sensitivity. The aim of this study is to demonstrate the technology for microphone phased arrays for source localization in the NASA Langley Research Center 4 by wind tunnel in order to permit suppression of undesirable noises and improve the acoustic measurement capability for automobile testing. Superscript *T complex conjugate Nomenclature A constant B m shear layer correction factor c speed of sound (in s - ) d shear layer thickness (in) e steering vector, see Eq. (6) h source to shear layer distance (in) k wavenumber (in - ) M total number of microphones in array N number of blocks of data P pressure (Pa) P(e) steered powered response R spatial correlation matrix r radial distance (in) r c corrected distance (in) r m radial distance between the source and the m th sensor (in) r' m radial distance between the m th sensor and arbitrary focus point (in) r mic radial line-of-sight distance (in) r p radial distance between the origin and arbitrary focus point (in) r r retarded radius (in) r s radial distance between the source and the origin (in) r radial distance from source to shear layer (in)

2 r t W Ws Y ω A radial distance from shear layer to observer (in) time (s) theoretical array response at wavenumber, k, see Eq. (4) data window weighting constant Fourier transform of data block radial frequency (rad s-) I. INTRODUCTION UTOMOBILE aeroacoustics are still in their beginning stages of understanding and control of the noise sources using exploratory experimental methods and idealized models. Previous research has conducted noise measurements during road tests., This is very difficult to do because of accidental disturbances, such as sidewinds, and the presence of engine and tire noise. Therefore, wind tunnels must be used. For accurate acoustic measurements, the wind tunnel needs to have low background sound pressure levels especially at low frequencies. In recent years, this has led to the development of special wind tunnels under the class "quiet airflow" tunnels, such as the Boeing Low Speed Aeroacoustic Facility, the NASA Langley Quiet Flow Facility, German-Dutch Wind Tunnel Low-Speed Facility (DNW), and other specialty automotive wind tunnels.3-5 However, the majority of wind tunnels were designed for aerodynamic tests as the primary objective. These "noisy" wind tunnels mask the vehicle's sources and may affect the actual aerodynamic noise contributed by a particular feature. Non-acoustically treated wind tunnels have test section background noise that is dominated by broadband and tonal noises. The sources of the broadband noise are: broadband fan noise, turning vanes, screens, heat exchangers, boundary layer noise, shear layer turbulence, and noise from shear layers impinging on the collector (open-jet test sections)6,7. Examples of tonal background sources in wind tunnels are: fan blade passing frequency, fan blade/stator interaction, electromagnetic motor noise, vortex shedding noise, self-noise, and flow interactions of an open jet nozzle, collector, and the wind tunnel circuit.6 Other problems are the wall boundary layer and microphone self noise, especially in closed circuit wind tunnels due to their reverberant environments. These problems are usually solved by placing the acoustic instrumentation in the walls of the test section. Or for an open-jet wind tunnel, such as the NASA Langley Research Center 4 by Wind Tunnel, the acoustic instrumentation is placed outside the shear layer. The problems of using a traditional wind tunnel do not outway the benefit of conducting aeroacoustic and aerodynamic tests simultaneously, where the exact aerodynamic conditions the test was conducted under is recorded. The NASA Langley 4 by wind tunnel, pictured in Fig., is an closed circuit wind tunnel that can run open- or closed-jet with flow speeds up to 348 ft/sec and a Reynolds number ranging from 0 to. 06 per foot. The tunnel is 50 feet in length with a test section size of feet operating with a drive power of,000 Hp. The wind tunnel will be used for commercial automobile testing applications conducted by Old Dominion University. The aerospace industry has been developing new test techniques, such as microphone phased arrays and acoustic mirrors, and data processing algorithms, such as delay-and-sum beamforming, frequency domain beamforming, and most recently the deconvolution approach for the mapping of acoustic sources (DAMAS).8-0 A microphone phased array permits discrimination of low-level sources in the presence of diffuse Figure. NASA Langley 4 by Wind Tunnel. background noise by means of strong spatial variations in array sensitivity. The aim of this study is to demonstrate that phased array technology is capable of source location for automotive technology.

3 II. BACKGROUND - AUTOMOBILE Aeroacoustics Aerodynamic forces such as lift and drag are generated when an automobile is in motion. As the airflow moves over the vehicle it interacts with the surface causing aerodynamic noise. At low speeds, less than 30 mph (48 km/h), the sound levels are very low compared to the engine and tire noise. At higher speeds, around 60 mph (97 km/h), the aerodynamic noise becomes dominant. Thus at cruising speeds the driver and passenger(s) will experience annoyances. It will also lead to driver fatigue especially during long highway trips. It is clear to see that these noise levels must be suppressed in order to satisfy the occupants. Buchheim, Dobryznski, Mankau, and Schwab demonstrated that there is no direct correlation between low aerodynamic drag and low sound levels. This is due to the fact that aerodynamic drag on an automobile depends on the exterior airflow over the rear where the wake separates from the vehicle. The noise perceived by the occupants is mostly dependent on the details of the exterior airflow; for example, small openings around the doors and windows. These openings will add to the noise but not to the drag of the vehicle. The mechanisms that generate noise are quiet complex but they must be well understood in order to successfully design quieter automobiles. Since it is quite difficult to model the actual sound source there are three types of idealized models: monopole, dipole, and quadrupole sources. The monopole source results from unsteady volumetric flow and it is the dominant noise source at low Mach numbers. It is also generated by the pressure fluctuations that may travel through leaks in the car creating a "secondary" monopole source. The monopole is the most dominant type of sound source since it is tonal in nature. The dipole results from unsteady pressures acting on a rigid surface; for example, turbulent flow impinging on a surface. The quadrupole results from two fluid elements colliding with one another causing unsteady internal stresses in the fluid. This is similar to a turbulent shear layer. There are three primary types of characteristic noise: aspiration noise, cavity noise, and exterior noise. Aspiration noise occurs when air passes through a direct flow path such as a leak or gap. This is caused by the pressure difference between the exterior and interior of the vehicle. The exact causes of aspiration noise are complex phenomena, such as, external flow interaction with cavities, separation, boundary layer turbulence, unsteadiness, or other factors. It is also modeled as a monopole source. Cavity noise is due to regions of the vehicle located in high velocity flow, such as, the A-pillar and exterior rearview mirrors. Cavity noise works on the feedback and resonance phenomena. It has been explained by Callister and George as such: "A disturbance is shed from the front edge of the cavity and is convected at the local flow velocity. This disturbance impinges on the rear edge of the cavity. This generates an acoustic wave that propagates in all directions. When the acoustic wave reaches the front edge of the cavity, it may trigger the shedding of another disturbance. In this way, the shear layer develops a preferred frequency." An example of this phenomenon is when a 3-inch long (787.4 mm) side window is opened and the vehicle is traveling at 60 mph (96.6 km/h) there will be a preferred frequency of approximately 40 Hz. Although this low frequency will be difficult to hear with human ears, it will still be felt through vibrations. Finally, the exterior noise is due to exterior pressure fluctuations on the vehicle caused by the airflow over the surface. Since the flow around a vehicle is turbulent the pressure on the surface is unsteady in time. This type of noise is generally dipole-like. Since the automobile is not a rigid structure this type of noise causes the panels to vibrate and radiate into the interior. It should be pointed out that even if the flow around the vehicle remained perfectly attached there will still be noise generated because the boundary layer is still turbulent. On the other hand, if the flow separated the noise generated will be more intense. Watanabe et. al. examined the effect of body shapes on wind generated noises. The study showed that windshield contours influences the noise, when leaks are not present. It also proved that separated airflow has maximum sound pressure levels at the separation point and at the reattachment point. III. ARRAY SIGNAL PROCESSING ALGORITHMS Arrays are a distribution of transducers, receivers, transmitters, or elements which perform both functions, laid out in a certain spatial pattern; for example: linear, rectangular, spiral, or irregular shape. In this study the array will be made up of microphones which serve as transducers. In a phased array, phase delays are applied to the signals received by the individual array elements to steer or focus the array. Steering means the array "looks" in a particular direction and this implies plane wave propagation a farfield source approximation. Whereas, focusing the array means the phase delays were selected to "point in 3

4 space" and implies spherical wave propagation a near-field source approximation. This latter technique will be required for this application. Beamforming algorithms are array signal processing algorithms that focus or steer the array. It results in the array directivity pattern, which is the directional response of the array, and it is determined by the array design and the processing algorithm used. Beamforming algorithms exploit the usage of the wave equation to describe the propagation of energy through space. Generally, a mathematical solution will involve Bessel functions and associated Legendre polynomials. However, that is not of interest because the problem under consideration suggests that the sound wave is a symmetric spherical wave a monopole point source assumption. The monopole assumption means that a single spherical sound source radiates a spherical wave that is only a function of the radial distance, r, and time but not the angular coordinates from the source. Thus the wave equation becomes, r r r p r c t p () A solution for a propagating pressure wave in all direction is given by, p m ( t) A r m exp jω t () rm c where p m (t) is the measured pressure at the m th microphone, r m is the distance from the source to the m th microphone, the term (t-r m /c) is the time delay, and A is a constant. In order to focus on a source, the individual microphone outputs are delayed by the appropriate amounts and added together to reinforce the signal with respect to noise or waves propagating in different directions. This is known as delay-and-sum beamforming. This method allows for the microphone phased array to be electronically steered for source localization while providing noise rejection. Using the delay-and-sum beamforming technique the ideal array response for a simple ideal monopole source can be expressed as, ( r r ) ( r r ) M r r r s s p m m W( ω, x p, xs) exp jω r m m c (4) where r m is the radial distance between the source and the m th sensor, r' m is the radial distance between the m th sensor and the arbitrary focus point, r p is the radial distance between the origin and the arbitrary focus point, and r s is the radial distance between the source and the origin. Delay-and-sum beamforming is the oldest and simplest array processing algorithm. It works on the principle that if a propagating signal is present in an array, the sensor outputs are delayed and then added together. This reinforces the signal with respect to noise or waves propagating in different directions. This method will provide the ideal array response for a given source and frequency. It mathematically represents the spatial filtering of the array graphically at specific frequency (or wavenumber) and allows one to study the beamwidth (related to the precision and resolution of the array), mainlobe, and sidelobe structures. The frequency-domain beamforming algorithm is similar to the delay-and-sum method but developed completely in the frequency domain. The sensor signals correspond to a spatiotemporal filter that delays the signals by a prescribed amount, then adds them, and outputs the results. In the frequency domain the delay corresponds to a linear phase shift. It works on the principle that the "calculations are entirely in the frequency domain by Fourier transforming the inputs, applying the spatiotemporal filter, and inverse transforming the result". 6 The steps are also detailed by Humphreys, Brooks, Hunter and Meadows. 4 The spatial correlation matrix is formed from the raw data from the array by taking the Fast Fourier Transform (FFT) of the individual matrix elements; Eq. (5a and 5b). This step is completed after the raw data is converted from voltage to engineering units (Pascals) using microphone sensitivity data based on the calibration process. The calibration is based on a frequency of khz using a sound pressure level meter. 4 The focus or steering is done electronically; the physical array is not moved. 4

5 R R R R K O R R M M M MM (5a) with R ij NW N [ ik jk ] * ( f ) Y ( f ) Y ( f ) s k (5b) where N is the number of blocks of data, W s is the data window weighting constant, and Y is an FFT data block. The spatial correlation matrix is a Hermitian and is dependent on the temporal frequency. Since the wind tunnel is an open-jet tunnel there needs to be a shear layer refraction correction factors applied to the data after the spatial correlation matrix is determined. The open-jet facility has a problem that the sound must pass through the open-jet shear layer before reaching the microphone array. Since the microphone array will be placed outside the shear layer, shear layer refraction corrections must be applied to the beamforming algorithms. The basic idea works on the assumption that the fluid density is unchanged across the shear layer. As sound reaches the shear layer it is refracted and scattered in the process. This results in both an amplitude and angle change in the field. 7-9 The amplitude correction is represented in Eq. (6) by the B M term. The angle correction is represented by the second term in the exponential; see Eq. (6). The steering vector, denoted by e, is a sequence of exponents chosen to cancel the wave signal's propagation related phase shift. The phase shifts focus the beam's assumed propagation direction to the wave's direction propagation. It also models the signal's propagation characteristics in the frequency domain. ( r r ) r r s m m B exp jω + ω rm c e M r ( s rm rm ) r BM exp jω + ω rm c path path r c mic r c mic, shear M, shear (6) where r m is the radial distance between the source and the m th sensor, r' m is the radial distance between the m th sensor and arbitrary focus point, r mic is the line-of-sight distance from the source to the microphone, r p is the radial distance between the origin and arbitrary focus point, r s is the radial distance between the source and the origin, r path r + r is the wavefront travel distance, r mic is the line-of-sight distance from the source to the microphone, and c is the speed of sound. Finally, using the steering vector and the spatial correlation matrix the steered array output power spectrum at the user-defined steering location is obtained as, Ñ ( ) ( * T e e ( R R ) e * ) data background M (7) *T It should be noted that e denotes the conjugate transpose of the steering vector. The background spectrum is subtracted out to improve the results. The R is obtained without tunnel flow, where the acquisition system noise dominates the record output. background 5

6 IV. MICROPHONE PHASED ARRAY DESIGN When designing microphone phased arrays for existing facilities the array must 40 meet certain array performance requirements. 30 For automobile testing the desired frequencies of interest range from 50 to 0,000 Hz. The 0 desired spatial resolution is foot and the ) array's position relative to the device under.n 0 test is 9 ft away to the side. The data i( 0 acquisition board available for testing has a y -0 channel count limitation of 64 channels. Underbrink's reversed-logarithmic multi-0 arm spiral array design was used to develop -30 the array for the 4 by because the array design does not require a lot of sensors (i.e. -40 microphones) for a relatively large aperture size.3-5 This is very beneficial since there is x (in.) a channel count limitation. Also, the sensor arrangement will improve the beamwidth. Figure. Microphone Phased Array Design This is especially important for low frequency source mapping applications, where physical dimensions and the number of sensors available are constraining factors. In short, reducing the beamwidth at the lowest frequency of interest will improve the resolution of the array. The reversed-logarithmic multi-arm spiral changes spacing along the spiral such that the arc-length between the elements increase logarithmically from the outside in. This array design is very beneficial for low frequency source mapping applications. This is because the outer radius of the array dedicates the resolution (i.e. beamwidth) at the lowest frequency on interest; however, at the sacrifice of higher average sidelobe levels compared to other multi-arm designs. Underbrink demonstrated about a 0% reduction in beamwidth over the linear-spaced array design.4 For the given design parameters the final array design has the following dimensions: an inner radius of 4 inches, an outer radius of 48 inches, a spiral angle of 76, and a total of 63 microphones, as illustrated on Fig.. Appendix I lists the tabulated microphone coordinates. At the smallest frequency of interest (50 Hz) the spatial resolution is 4.8 feet. The foot spatial resolution could not be achieved because one reaches a limit in the array design. The array pattern was drilled onto 64 inch diameter plywood panel that is _ inch thick. At each location of the microphone, a _ inch hole was drilled. The panel was finished with gray paint in order to prevent moisture which could cause the panel to warp. A front view of the array is shown on Fig.3. Microphone holders were made out of a _ inch diameter fiberglass rod with approximately _ inch holes drilled in to hold the microphones in place. The panel was then back with supporting members and attached to a movable tripod. Figure 3. Physical Microphone Phased Array Also the electrical box was attached to the back of the array. V. RESULTS The array was benched tested with a tonal source (i.e. PA speaker) at a known source location. The speaker was placed 08 inches away from the array located at the array's origin. The array was tested at /3-octave frequencies ranging from 50 Hz to 0,000 Hz. The results, computed with frequency-domain beamforming, are shown for octave band frequencies (500 Hz to 4000 Hz) in terms of normalized pressure, Fig. (4-7). The experimental results clearly show that the beamwidth is a function of frequency; also, the 6

7 beamwidth decreases as the frequency increases. Interestingly at 4000 Hz there is no mainlobe present. This is due to undersampling issues in the data acquisition board and possibly mode effects of the speaker. However, the algorithm was able to show the sidelobe structures. Figure 4. Figure 6. Array pattern at 500 Hz octave Array pattern at 000 Hz octave Figure 5. Figure 7. Array pattern at 000 Hz octave Array pattern at 4000 Hz octave An acoustic test was performed with an automobile in the Langley Full Scale Tunnel. The array was placed outside of the shear layer. The automobile was then placed 08 inches away from the array. The wind tunnel was operated at 80 rpm. Array patterns were computed for /3-octave bands. Within each band the patterns were summed together. This step was performed since the automobile noise sources are typically broadband in nature. Therefore, results are shown, below, for octave bands of 500 Hz, 000 Hz, and 000 Hz. Higher frequencies were neglected due to undersampling issues of the data acquisition board. Sources at lower frequencies were masked by the background noise of the wind tunnel. Figure 9. Array pattern at 500 Hz octave Figure 0. Array pattern at 000 Hz octave 7

8 The results show that the dominant noise source around the car window occurs at 000 Hz (Fig. 0). This is possibly due separated airflow over the passenger side window (closed). Maximum sound pressure levels would occur at the separation and reattachment points of the airflow. There is a source at 500 Hz; however, it is not as intense as the 000 Hz (refer to Fig. 0). Figure, also, shows a sound source. However it shows more low-level broadband noise compared to Fig. 0. VI. SUMMARY It is necessary to understand what type of noise is Figure. Array pattern at 000 Hz octave expected around the exterior of an automobile. The mechanisms that generate sound on the exterior of the automobile are quite complex phenomena involving unsteady aerodynamics. Thus, idealized models are used to simplify the complexities. Using these idealized models array signal processing algorithms have been developed delay-and-sum beamforming and frequency-domain beamforming. Delay-and-sum beamforming algorithm was used to develop the sensor locations of the array and to also obtain theoretical array patterns. The frequency-domain beamforming was used to process the experimental results with a known tonal source location and during a wind tunnel test. The experimental results with a known tonal source location clearly demonstrates the array ability to electronically steering to the source location and "focus" on it. During the wind tunnel the array was capable of discriminating against other source and background noise, and locates noise sources around the automobile. 8

9 APPENDIX I ARRAY DESIGN COORDINATES Table. Microphone Phased Array Coordinates Mic. Number X (inches) Y (inches) Mic. Number X (inches) Y (inches) ACKNOWLEDGEMENTS The author would like to thank Dr. Colin P. Britcher from Old Dominion University for advising this study. A special thank you goes out to John Bledsoe, from Langley Full Scale Tunnel, for designing the microphone circuitry. Also a very special thank you to William Humphreys, from NASA Langley Research Center, for assistance in understanding the theory into microphone phased arrays. LITERATURE CITED George, A.R., and Carr, J.F., "Recent Advances in Understanding Automobile Aerodynamic Noise" st Joint CEAS/AIAA: Aeroacoustics Conferences, CEAS/AIAA , 995. Callister, J.R., and George, A.R. "Wind Noise", Aerodynamics of Road Vehicles, W.H. Hucho, SAE International, Warrendale, PA, 4 th ed., Herkes, W.H., and Stoker, R.W. "Wind Tunnel Measurements of the Airframe Noise of a High-speed Civil Transport", AIAA Paper , Humphreys, W.M., Brooks, T.F., Hunter, W.W., and Meadows, K.R. "Design and Use of Microphone Directional Arrays for Aeroacoustic Measurements", AIAA Paper , Eitelberg, G., and Eckert, D. "Some Developments in Experimental Techniques of the German Dutch Wind Tunnels (DNW)", AIAA Paper , Duell, E., Walter, J., Arnette, S., and Yen, J., "Recent Advances in Large-Scale Aeroacoustic Wind Tunnels", AIAA Paper , Duell, E., Yen, J., Walter, J., and Arnette, S., "Boundary Layer Noise in Aeroacoustic Wind Tunnels", AIAA Paper , Brooks, T.F., and Humphreys, W.M., "A Deconvolution Approach for the Mapping of Acoustic Sources (DAMAS) Determined from Phased Microphone Arrays", AIAA Paper , Brooks, T.F., and Humphreys, W.M. "Three-Dimensional Application of DAMAS Methodology for Aeroacoustic Noise Source Definition", AIAA Paper , Dougherty, R.P., "Extensions of DAMAS and Benefits and Limitations of Deconvolution in Beamforming", AIAA Paper , 005. Buchheim, R., Dobryzynski, W., Mankau, H., Schwabe, D., "Vehicle Interior Noise Related to External Aerodynamics", International Journal of Vehicle Design. SP3, pp ,

10 Watanabe, M., Harita, M., Hayashi, E., "The Effect of Body Shapes on Wind Noise", SAE Paper No , Society of Automotive Engineers, Warrendale, Pa, Underbrink, J.R. "Aeroacoustic Phased Array Testing in Low Speed Wind Tunnels", Aeroacoustic Measurements, T.J. Mueller (Ed.), Springer-Verlag, New York, Underbrink, J.R. "Practical Considerations in Focused Array Design for Passive Broadband Source Mapping Applications", Master's Thesis, The Pennsylvania State University, State College, PA, Underbrink, J.R. Circularly Symmetric, Zero Redundancy, Planar Array Having Broad Frequency Range Applications. U.S. Patent No. 6,05,4, Johnson, D.H., and Dudgeon, D.E. Array Signal Processing: Concepts and Techniques. Prentice Hall, NJ, Amiet, R.K., "Correction of Open-Jet Wind-Tunnel Measurements for Shear Layer Refraction", AIAA Paper , Schlinker, R.H., and Amiet, R.K., "Shear Layer Refraction and Scattering of Sound", AIAA Paper , Amiet, R.K., "Refraction of Sound by a Shear Layer", Journal of Sound and Vibration, v. 58, no. 3, pp , Brooks, T.F., Marcolini, M.A., and Pope, D.S., "A Directional Array Approach for the Measurement of Rotor Noise Source Distributions with Controlled Spatial Resolution", Journal of Sound and Vibration, v., no., pp. 9-97, 987. Marcolini, M.A., and Brooks, T.F., "Rotor Noise Measurement Using a Directional Microphone Array", Journal of the American Helicopter Society, pp. -, 99. 0