Alexander Et Al. - 2011 - Aerodynamic Noise From Sparse Surface Roughness

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

17th AIAA/CEAS Aeroacoustics Conference(32nd AIAA Aeroacoustics Conference) AIAA 2011-2740

05 - 08 June 2011, Portland, Oregon

Aerodynamic Noise from Sparse Surface Roughness

W. Nathan Alexander1, William Devenport2


Virginia Tech, Blacksburg VA 24061

Stewart A. L. Glegg3
Florida Atlantic University, Boca Raton FL 33431

Abstract

The noise produced by fetches of 3mm hemispherical and cubic roughness in a turbulent boundary layer was
studied. The wall pressure was recorded in and around 42 element fetches of the roughness. Far field
measurements were made using a linear phased microphone array. The microphone array data was analyzed
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

using a conventional beamforming method as well as a least-squares analysis technique which can separate
individual source strengths in the multi-element fetches. The cubic elements produce much stronger far field
pressure fluctuations than the hemispherical elements. This radiated noise is produced by the inhomogeneous
wall pressure fluctuations on the surface of the roughness elements. Examination of different roughness noise
directivity models confirms the presence of separate spanwise and streamwise dipoles at each roughness
element location. The relative strengths of the sources for a single 3mm cubic element vary with frequency
and therefore produce a directivity pattern that varies with frequency. Normalized estimated mean square
unsteady drag spectra for cubic and hemispherical elements have a similar shape and magnitude to those
computed using LES by Yang & Wang (2011). For wall jet nozzle velocities of 50 and 60m/s, a single cubic
element’s estimated spanwise dipole is significantly weaker than the streamwise dipole at lower non-
dimensional frequencies. Above fh/U=2, estimated streamwise strengths are similar to the spanwise dipoles.
Source strength estimations of the multi-element fetches show that the addition of trailing cuboidal elements
enhance the streamwise noise produced by upstream elements. Also, the estimated spanwise dipole strength
of cubic elements was enhanced by the presence of spanwise adjacent elements.

I. Introduction

Roughness noise can be a significant contributor to the total sound produced by vehicles with large surface area
to perimeter ratios operating in low Mach number environments such as marine vehicles. Early theoretical work in
the field was pioneered by Howe (1984) with reference to the experimental work of Hersh (1983). Hersh measured
enhanced noise due to surface roughness in pipe flow and concluded that the noise scales as a dipole. Howe (1984)
developed a scattering theory which modeled the roughness as a random array of hemispherical bosses. Howe’s
work agreed with Hersh’s conclusion of a dipole scaling, but he was unable to perform an absolute comparison
because of unknown factors affecting the refraction of sound in the experiment.
More recent work includes that of Grissom (2007) which details the development of the Virginia Tech Anechoic
Wall Jet Facility which is a purpose built facility for the study of roughness noise. He completed measurements that
show a clear increase in noise due to wall roughness even for hydrodynamically smooth surfaces confirming the
existence of a scattering mechanism. He applied multiple inner and outer variable scalings to the far field data with
little success noting changes in spectral shape between measurements of different surface types. He concluded that a
simple scaling based on boundary layer or surface parameters will be unable to collapse these data.
Glegg & Devenport (2009) explain the reasoning for the poor collapse. They relate the radiated far field noise to
a combination of both the surface slope wavenumber spectrum and the wall pressure spectrum. Therefore, the shape
of the far field noise spectrum is dependent on the surface roughness geometry as well as the wall pressure field.
Alexander (2009) and Devenport et al. (2011) used the relation in Glegg & Devenport (2009) to collapse the far
field noise produced by several stochastic surfaces. The collapse confirmed the dipole nature of the source and was
effective up to a “break frequency” which was dependent on the surface shape. Deterministic surfaces were also

1
Graduate Research Assistant, Department of Aerospace and Ocean Engineering, AIAA Student Member.
2
Professor, Department of Aerospace and Ocean Engineering, AIAA Associate Fellow.
3
Professor, Department of Ocean Engineering, AIAA Associate Fellow.

1
American Institute of Aeronautics and Astronautics
Copyright © 2011 by William Nathan Alexander. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
studied in Alexander (2009) and Devenport et al. (2010). Devenport et al. (2010) used the relationship between the
radiated far field noise and the surface geometry to measure the wall pressure wavenumber spectrum using a
hydrodynamically smooth ribbed surface.
Liu et al. (2008) also measured noise from deterministic surfaces. They used fetches of hemispherical roughness
in an open-jet tunnel and measured the noise with a phased microphone array. They overcame the inherent
monopole source assumption in typical beamforming techniques by comparing source maps processed using the
standard techniques to a theoretical field of dipoles processed the same way. They found the results were similar and
therefore concluded that roughness noise is a dipole source.
Research on noise produced by deterministic surfaces has proven beneficial because experimental studies can
more easily by compared with computational studies. Yang & Wang (2010) performed a computational study of
roughness noise generated by fetches of hemispherical and cubic elements using large eddy simulation. Spanwise
and streamwise dipoles were generated at each roughness element, but they found the noise characteristics varied
between the two roughness geometries. The hemispherical elements produced spanwise dipoles that were stronger
than the streamwise dipoles where as the reverse was true for the cubic element case. Also, the lead row of
hemispherical elements was the weakest producer of noise and actually enhanced the production of sound from the
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

following row. In contrast, the lead row of the cubic element fetch produced the strongest noise and the source
strengths gradually decreased with each successive row.
Alexander et al. (2010) developed a least-squares method of solving for source strengths using a phased
microphone array that bypasses any assumption of monopole sources. They measured noise from varying size
fetches of 3mm cubic elements. They then calculated source strengths by using the measured cross-spectral matrix
and a transformation matrix of Greens functions for uncorrelated dipole sources in the streamwise and spanwise
direction. The results showed good agreement with single microphone measurements for frequencies above 10kHz.
Below this, source strengths were over-estimated. Also, due to the positioning of the microphone array, streamwise
source strength estimations in this study were highly uncertain.
The following study analyzes the noise from multiple fetches of both hemispherical and cubic elements. Surface
pressure measurements in and around 42 element fetches of cubic and hemispherical elements will also be
presented. A linear phased microphone array is used in the analysis to decompose the spanwise and streamwise
source strengths for individual elements of multi-element fetches using the least-squares method of Alexander et al.
(2010). The method of Alexander et al. (2010) is advanced by combining phased array measurements at multiple
sensing locations relative to the roughness. This improves the uncertainty of the estimations most notably for the
streamwise dipoles. Effects of velocity on the source strengths will be examined as well as any interaction between
elements. Differences between the noise production of the hemispherical and cubic elements will also be
investigated and compared to the findings of Yang & Wang (2011) which improves upon the LES source strength
estimations in Yang & Wang (2010).

II. Apparatus and Instrumentation

A. Wind Tunnel
Data were recorded in the Virginia Tech Anechoic Wall Jet Facility shown in Figure 1. This tunnel was
developed in 2005 and its acoustic and aerodynamic properties have been analyzed in detail by Grissom (2007),
Alexander (2009), and Devenport et al. (2011). The tunnel is powered by a centrifugal fan separated from the
settling chamber by a flexible rubber hose. There are several acoustically treated baffles in the settling chamber that
block direct noise radiation out through the nozzle. The nozzle opening can be adjusted to various heights but was
fixed at 12.7mm for this study. The air is accelerated out over a flat aluminum plate 1524mm wide and 3058mm
long. The maximum velocity decays with streamwise distance from the nozzle and then is exhausted underneath the
acoustic chamber. The wall jet design is advantageous for the study of roughness noise because far field
microphones can be placed well outside of the flow and all edges are kept away from regions of significant flow
velocity. The coordinate system used to describe roughness and microphone locations, shown in Figure 1b, has an
origin at the spanwise center of the nozzle exit.
The wall jet boundary layer profile is self-similar and can be characterized by relations similar to those given in
Narasimha et al. (1973) and Wygnanski et al. (1992) as shown in Equation 1.

2
American Institute of Aeronautics and Astronautics
Um
= 4.97 Re nj+1 Re nx
U0
y1 / 2
= 0.0335 Re mj − 2 Re mx Eq. 1
b
δ*
= 0.0156 Re pj −2 Re xp
b
where Um is the maximum velocity of the profile, Uo is the nozzle exit velocity, y1/2 is the vertical distance measured
from the wall to the point above the maximum velocity where the local flow speed has decayed to half the value of
Um, b is the nozzle height, δ* is the displacement thickness, Rej is the Reynolds number based on the nozzle exit
height, and Rex is the Reynolds number based on the streamwise distance from the nozzle exit. The values m, n, and
p are facility dependent and have been experimentally determined to be 1.0451, -0.512, and 0.888, respectively. The
boundary layer thickness, δ, and the momentum thickness, θ, can be determined as multiples of the displacement
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

thickness such that δ=15.4δ* and θ=0.74δ*.


Roughness fetches were placed starting at x=1257mm from the nozzle exit and were centered spanwise on the
flat plate. Conditions at this location range from Um=7.5-22m/s and δ=21-16mm for nozzle exit velocities 20-60m/s.
Viscous lengthscales vary from 1/25th to 1/65th of a millimeter.

B. Rough Surfaces
Fetches of hemispherical and cubic elements with roughness heights of 3mm were used in this study. This
corresponds to roughness Reynolds numbers ranging from h+=huτ/ν=73-197. Fetch sizes for this study varied from
single elements to 6x7 grid patterned arrays with a 16.5mm grid spacing from element centers. The effect of trailing
rows was studied by successively increasing the number of downstream rows from an original single row of six
elements. Table 1 details the roughness fetches studied for both types of surfaces. The fetch pattern defines the
number of spanwise by streamwise elements. The cubic element surfaces are shown in Figure 2.

Roughness Type Fetch Pattern Roughness Type Fetch Pattern


Cubic 1x1 Hemispherical 1x1
Cubic 6x1 Hemispherical 6x1
Cubic 6x2 Hemispherical 6x2
Cubic 6x3 Hemispherical 6x5
Cubic 6x4 Hemispherical 6x7
Cubic 6x5
Cubic 6x6
Cubic 6x7

Table 1. Roughness configurations.

The rough surfaces were manufactured by molding epoxy resin with backing layers of Kevlar scrim and sketch
paper. The resulting substrate was 0.30mm thick. Experiments comparing sound produced by the substrate alone
with that produced with a clean wall showed that the step perimeter produced no measurable far field noise.

C. Linear Microphone Array


Source directivity and strength analyses were conducted using a linear microphone array of 36 ½” B&K model
4190 microphones. This array is shown in Figure 3. The microphones are spaced as close as possible at 15.1mm
center-to-center and are held in position by fitting the microphones through a Delrin block that is mounted in the
wind tunnel by two end pins. The microphone data are recorded using B&K’s Pulse 14 platform along with six
B&K 3050-A LXI modules. Data are recorded at 65536Hz for 40 seconds, and the frequency spectra are calculated
as the average of 2560 to 5119 records of 1024 samples. The spectra are then frequency averaged in 1/10th octave
bands to reduce uncertainty. Background noise measured without the roughness present is subtracted from all
roughness noise measurements.

3
American Institute of Aeronautics and Astronautics
The microphone array was positioned at four orientations relative to the rough surface to record data in multiple
regions of the directivity field. These array positions are shown in Figure 4. Microphone positions were measured
using a FaroArm Fusion portable coordinate measuring machine for accuracy. The array was calibrated in
Microphone Array Position 2 using data recorded by a separate single ½” B&K 4190 microphone to cancel effects
of the Delrin block on the recorded pressures of the array’s inset microphones. This calibration was assumed
constant for all array measurement positions. Table 2 lists the central position of the linear microphone array and
receiver angle relative to the horizontal plane of the wall.

Array Position 1 Array Position 2 Array Position 3 Array Position 4


x, mm 1257 915 1366 1257
y, mm 25 585 626 597
z, mm -716 0 0 -1
θ 0° -56° -90° -90°

Table 2. Microphone array measurement positions.


Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

i. Beamforming
Source maps were generated using a conventional delay and sum beamforming method in which the inverse
Greens function for a monopole source is used in a summation at each point over a defined area using the measured
cross-spectral matrix and the known locations of the microphones. The process involves adjusting the phase and
magnitude of the measurements to account for the change in distance from each microphone to a focal point. The
cross-spectral matrix is then summed. This procedure is then repeated for many locations over a defined focus area
and the summations can be plotted producing a source map. Peaks occur in the source map when the cross-spectral
matrix adds constructively indicating the location of a source. The summation and phase adjustment were completed
using the matrix calculation from Tester & Glegg (2008) as shown in Equation 2.

( )
v
b = diag G inv CG inv
H
Eq. 2

where C is the background-subtracted cross-spectral matrix and Ginv is the inverse monopole transformation matrix
or steering matrix which applies a magnitude and phase adjustment to the individual elements of the cross-spectral
v
matrix for multiple focal points. The superscript H denotes the Hermitian transpose. The resultant b vector is the
phase delayed and summed solution for all focal points.
The Greens function for the acoustic propagation of a monopole source at yj and sensor at xm is shown in
Equation 3.
ikr
e j ,m
G j ,m = r j ,m = x m − y j Eq. 3
4πr j ,m

where k is the acoustic wavenumber and rj,m is the absolute distance between the microphone and source location.
Therefore, the elements of Ginv are defined as in Equation 4.

− ikr j , m
Ginv j ,m = r j ,m e rc M Eq. 4

where rc is a reference distance and M is the number of microphones. A reference distance of 1m was used for all
presented source maps.
Point source images for this array are shown in Figure 5 for multiple array focal areas. This figure illustrates the
linear array’s poor out-of-plane capabilities. In this example, the array can only distinguish between sources in the z-
direction.
Source convection due to the local flow velocity has been taken into account by skewing the source maps using
observed convection distances. These convection distances were determined by placing a white noise source at four
corners bounding the area where the roughness was to be applied. Separate measurements were taken with the noise
emanating from each corner with the flow off and with the flow on. The convection distances were calculated by
comparing the cross-spectral phase difference between an ideal source at the physical source location and the

4
American Institute of Aeronautics and Astronautics
measured cross-spectral phase. The convection results at the four corners were then linearly interpolated and
extrapolated over the entire source map area.
ii. Least-Squares Analysis
A least-squares analysis, as detailed in Alexander et al. (2010), was used to separate and determine the source
strengths of individual roughness elements in a fetch. This method finds the least-squares solution of source
strengths using the measured cross-spectral matrix put into a vector form and a transformation matrix of cross-
spectral Greens functions. First, the recorded pressures are written as the summation of the response due to
individual sources.

Pm = q1Gm,1 + q 2 Gm, 2 + L q j Gm, j Eq. 5

where Pm is the acoustic pressure recorded at microphone m, qj is the source strength of source j, and Gm,j is the
Greens function for acoustic propagation of source j to sensor m.
For this analysis, each roughness element was assumed to radiate uncorrelated spanwise and streamwise aligned
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

dipoles in half-space with a Greens functions defined by Equation 6.

− ik cos(θ m, j )e (1 + i kr )
ikrm , j

e iωt cos(θ m , j ) = xˆ ⋅ (x m − y j ) rm , j
m, j
Gm, j = Eq. 6
2πrm , j

where x̂ is the dipole directionality unit vector. The cross-spectrum between two microphones, m and n, can be
calculated using Equation 5 multiplying Pm by the conjugate of Pn. As the signals from the uncorrelated sources are
averaged, the cross source terms go to zero leaving only a linear combination of the source components as in
Equation 7.

Pm Pn* = q12 Gm ,1Gn*,1 + q 22 Gm, 2 Gn*, 2 + L q 2j Gm , j Gn*, j Eq. 7

This equation can be rewritten for all terms in the cross-spectral matrix as:
v v
G pp = G GG Q Eq. 8

v
where Q is a vector of mean squared source strengths, GGG is a transform matrix of cross-spectral Greens functions,
v v
and G PP is the measured cross-spectral matrix put into vector form. The source strengths, Q , are then calculated
from this heavily over-determined system of equations using a least-squares algorithm with a non-negative solution
constraint. The source convection results used in the beamforming analysis were applied to correct the physical
source positions to the observed acoustic locations.

D. Other Instrumentation
Surface pressure measurements were recorded using Sennheiser KE 4-211-2 electret microphones with modified
1/4mm pinholes. These microphones were calibrated using a white noise source and a B&K 1/8” model 4189
microphone as reference. Data from the Sennheisers were recorded with an Agilent E1432 digitizer sampling at
51200Hz. The presented spectral data are the result of 1000 averages of 2048 samples each that are then averaged
over 1/10th octave bands. The surface pressure microphones were flush mounted with the substrate of the rough
surfaces and data were recorded at several locations relative to the surrounding roughness elements. These locations
are shown in Figure 6 for the 42 element hemispherical and cubic element fetches. Measurements were completed at
9 positions in the hemispherical element fetch and 14 in the cubic element fetch. They will be referred to by the
corresponding labels C1-C14 and H1-H9.

5
American Institute of Aeronautics and Astronautics
III. Results and Discussion

1. Wall Pressure Variation


Wall pressure spectra recorded inside and around the 6x7 array of cubic elements at a nozzle exit velocity of
60m/s are shown in Figure 7. Figure 7a shows the streamwise variation of the pressure fluctuations. The measured
wall pressure spectrum half an element spacing upstream, C1, is slightly elevated at all frequencies compared to the
smooth wall data. The spectral intensities rise even further between the leading rows. The low frequency
fluctuations, below 4kHz, remain relatively constant for the interior measurement positions, C2, C3, and C4, but
higher frequencies tend back to the smooth wall spectrum with increasing streamwise position. Behind the
roughness fetch the wall pressure fluctuations do not continue this streamwise trend and are elevated at all
frequencies above the smooth wall spectrum by up to 6dB. This suggests that the boundary layer is being displaced
by the fetch of cubic roughness elements and that the presence of the downstream elements affects the upstream
pressure fluctuations. The flow reattaches downstream of the roughness fetch creating a region of increased wall
pressure fluctuations.
There is little spanwise variation in the wall pressure fluctuations through the middle of the fetch as shown in
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

Figure 7b. The lower frequency region below 5kHz has slightly less energy for the outer measurement positions
which are bound by only one roughness element, positions C6 and C9. Since the wall pressure does not vary
significantly in the spanwise direction at interior points, the analysis of measurements at positions C3 and C10-C14
will be compared as if they were recorded around a single interior element.
Figure 7c shows the wall pressure spectra at different relative locations to the surrounding elements. This
highlights the inhomogeneity of the wall pressure field. This rough flow is unlike the previously studied stochastic
roughness cases in Devenport et al. (2011) in which the recorded wall pressure fluctuations were homogeneous over
the measured frequency range. Therefore, the noise from this large deterministic roughness is unlikely to collapse
similarly on any single point wall pressure measurement. Wall pressures at all measured locations were greater than
the smooth wall values below 8kHz. Above 8kHz, the position just in front of the element recorded lower spectral
values. The measurements are consistent with the simulations of Yang & Wang (2010) which show separated
regions upstream and downstream of cubic roughness elements. Although, Yang & Wang (2010) find that the
recirculation zone downstream of the cubic elements produce less intense wall pressure fluctuations than the
separated upstream region. Figure 7c shows just the opposite for similar distances 1.2 element heights upstream and
downstream of the cuboid at positions C11 and C13.
A similar analysis of the wall pressure field created by a fetch of 3mm hemispherical elements is shown in
Figures 8a-c which is divided to analyze the streamwise, spanwise and element relative variation. The wall pressure
ahead of the fetch is slightly elevated similar to the fetch of cubic elements. The wall pressure increases and
becomes constant by the middle of the fetch at position H3. Beyond this point there is no significant wall pressure
variation in the streamwise direction. The downstream position, H5, does not record the same jump in wall pressure
as seen in Figure 7a for position C5. This suggests that the downstream hemispherical elements do not have as
significant an effect upstream as the cubic elements. The same boundary layer displacement and reattachment is not
observed for the hemispherical roughness.
Again, the same spanwise variation as observed for the cubic fetch is seen in Figure 8b. Position H8, which is
bound by a roughness element on only one side has a slightly suppressed low frequency region as compared to the
interior measurements which are constant. Therefore, the measurements at interior positions, H3 and H9, will be
compared as if they were relative to the same roughness element. These two positions vary as much as 9dB with the
measured spectral levels at all frequencies increased at the H9 position located in-line streamwise between two
roughness elements. This again agrees with the computational study of Yang & Wang (2010) showing increased
pressure fluctuations on the wall in the wakes of the roughness elements as compared to the wall pressure
fluctuations between streamwise rows.
Figure 9 compares the wall pressure spectra measured at two similar locations in the cubic and hemispherical
fetches, H4 and C4. This position is between the final two spanwise rows of the roughness. The recorded spectra are
almost identical up to 5kHz above which the cubic fetch has lower spectral values. This is an interesting finding
because of the large difference in far field noise produced by these fetches.

2. Single Point Far Field


Figures 10 and 11 show the raw and background-subtracted far field data for the 42 cubic element fetch and the
single cubic element at a range of nozzle exit velocities from 20-60m/s as recorded by a central sensor of the
microphone array at Array Position 2. This upstream position has the highest signal-to-noise ratio because it is

6
American Institute of Aeronautics and Astronautics
shielded from direct jet noise radiation from the nozzle by the acoustically treated shelf (Fig. 1). The 42 element
fetch has an 18dB signal-to-noise ratio at the highest velocity compared to 7dB for the single cube. Below
Uo=40m/s, the noise from a single cube was not discernable using this single microphone above the background
noise. Noise produced by the hemispherical elements was much lower than that created by the cubes. Even the 42
hemispherical element fetch produced no measurable noise in the single microphone autospectra at any of the
examined velocities. Combined with the wall pressure result shown in Figure 9 comparing the like wall pressure
fluctuations, this indicates that the inhomogeneous pressure fluctuations on the surface of the roughness elements
that produce the roughness noise do not scale with the pressure fluctuations on the surrounding substrate.

3. Beamformed Source Maps


Figure 12 displays source maps at 10368Hz for both the single cubic element and the 6x7 fetch of cubic elements
at a nozzle exit velocity of 60m/s obtained using the array in Position 1. The roughness noise is clearly visible above
the background, but the source location cannot be determined outside of the x-axis. Also, the spatial resolution of the
source maps is not fine enough to distinguish between multiple rows of the roughness fetch. Regardless, there are
some conclusions we can derive from these results. Figure 13 shows a diagram of the measurement and several
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

sweeps through the center of the source maps at z=0 for different roughness conditions. The magnitudes have been
multiplied by the constants shown in the figure legend so that the maps peak at similar levels. The single cubic
element map peaks at the location of the roughness element and is approximately symmetric. Noise distributions for
the 6x7 hemispherical and cubic fetches have similar shapes and peak slightly downstream of the lead row but still
ahead of the second row of elements. These results agree with those found in Alexander et al. (2010). The peak shift
downstream does not seem to be a shift of the actual peak source strength but is due to spatial broadening of the
source map. The majority of spanwise sound for both fetches seems to be produced near the leading rows of the
roughness. The source strength then weakens with each succeeding row. The source strengths differ greatly between
roughness shapes. The peak source strengths from the hemispherical element fetch are almost an order of magnitude
smaller than even the single cubic element. The array could not produce a reasonable source map for the single
hemisphere because the signal-to-noise ratio was too low.
A similar beamforming analysis was completed with the microphone array in Position 2. Figure 14 shows a
diagram and plot of spanwise sweeps through the beamformed source maps at x=1257mm. Again, the single element
map peaks at the location of the roughness and is symmetric about this peak. The source maps from the 42 element
fetches are similarly shaped but show some spanwise variation suggesting the flow from the nozzle is not perfectly
symmetric, although, the disparity is reasonably small. The hemispherical elements produce significantly less sound
in the streamwise direction as compared to the cubic elements. These data reveal some of the characteristics of the
noise but as discussed earlier cannot provide measures of the relative strengths of multiple sources or accurate
relative measures between the spanwise and streamwise sources.

4. Least-Squares Source Strength Analysis


The least-squares analysis was used to separate these sources and estimate their strength spectra. The least-
squares source strength analyses of the noise from single cubic and hemispherical elements reveal the large
difference in source strengths of the two element shapes. Figure 15 shows the estimated spanwise and streamwise
dipole strengths for both elements at a nozzle exit velocity of 60m/s. The estimated cubic element source strengths
are over an order of magnitude stronger than that of the hemisphere. Comparison of the cubic element’s spanwise
and streamwise dipole show that the streamwise dipole is strongest for all frequencies and decreases at
approximately the rate of f -3.5. This slope is shallower than the results presented in Alexander et al. (2010) of f -4.5,
but these early results only incorporated measurements from the spanwise Array Position 2 and did not include a full
calibration of the array’s response function. Although conventional beamforming was unable to produce any results
for the single hemispherical element even at the highest studied velocity, the least-squares technique did converge
over a limited frequency range. The results for the hemispherical element in Figure 15 are not as certain due to the
very low signal-to-noise ratio and thus limited the frequency range over which a non-negative least-squares solution
could be obtained. Therefore, no definitive conclusion can be made about the relative strengths of the hemisphere’s
spanwise and streamwise dipoles.
The two dipole sources for the cubic element have dissimilar spectral shapes. The spectra at 3kHz are separated
by almost an order of magnitude. At 11kHz, the spectral values are nearly equal. This means that the directivity
pattern varies with frequency. Figure 16 shows the implied radial directivity of noise in the plane of the wall at a
distance of 0.65m produced by a single cubic element at Uo=60m/s for two frequencies. At 3872Hz, the streamwise
dipole is much stronger than the spanwise so that the radiation pattern is 7dB stronger in than streamwise direction.

7
American Institute of Aeronautics and Astronautics
At 10368Hz, the dipoles are approximately equal and the directivity pattern is nearly constant in all directions
varying by only 1dB.
The source strength spectra from the single cube were determined at multiple nozzle exit velocities and are
shown in Figure 17. No estimation could be made for the 20m/s nozzle exit velocity because the signal-to-noise ratio
was too low. At all velocities, the determined spanwise strength was less than the streamwise dipole strength. The
strengths at each frequency increased by approximately three orders of magnitude with a doubling of the local
velocity from 11-22m/s which corresponds to the nozzle exit velocity increase of 30-60m/s.
These spectral data can be normalized as in Figure 18 using the undisturbed boundary layer mean velocity at the
height of the roughness element and characteristic size of the roughness, h=3mm. The velocity at the roughness
height was approximately 80% of the local maximum velocity. The Uo=40-60m/s data collapse well, but the data
from Uo=30m/s falls below the streamwise and spanwise curves. The data from the single hemispherical element
case at Uo=60m/s is included on this figure as well. These data are compared to the spanwise-averaged streamwise
and spanwise dipole strengths of Yang & Wang (2011) for the lead row of a fetch of cubic and hemispherical
elements arranged in a 4x10 pattern. The cuboidal experimental and LES results have a similar shape and
magnitude. The cubic element normalized streamwise dipole strength spectra agree well for fh/U<1 and rise slightly
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

above the LES estimations at higher frequencies. The Uo=50-60m/s spanwise results behave similar to the LES
calculation with results falling significantly below the estimated streamwise values at lower nondimensional
frequencies. Although, the non-dimensional spectra of the Uo=30-40m/s do not exhibit this behavior with
streamwise and spanwise values approximate equal at the lowest calculable non-dimensional frequencies. The
relative magnitude of the cubic and hemispherical elements’ source strengths agree with the LES study. The cubic
element is the stronger producer of noise and the normalized spectra differ by approximately two orders of
magnitude. A definitive conclusion cannot be reached concerning the relative strengths of the hemisphere’s
streamwise and spanwise dipoles because of the large uncertainty and limited range of experimental results. Overall,
experimental data agrees well with the LES study of Yang & Wang (2011). Discrepancies between the LES and
least-squares analyses could be due to differences in the experimental and simulated measurements. The LES study
was conducted for a typical boundary layer and free stream region. The experimentally measured data used a wall jet
which contains a mixing layer that could significantly affect the results.
The relative change in the cubic element’s spanwise and streamwise dipoles with velocity is analyzed by
integrating the results of Figure 17 over the normalized frequency range 0.625< fh/U <2.5. The ratio of this
integration is shown in Figure 19 displaying an increase in the relative strength of the streamwise to spanwise dipole
as the velocity is increased. The strength of the streamwise dipole grows more rapidly nearly doubling in relation to
the spanwise dipole.
Figures 20 and 21 show the estimated streamwise and spanwise source strengths for the lead row of the multi-
element cubic roughness. The least-squares technique has been shown to work for the single element, but can fail to
converge when the number of sources is increased. To reduce uncertainty, the individual source strengths given in
Figures 20 and 21 have been spanwise-averaged per element so that only one curve is given representing the lead
row of each fetch. The streamwise source strengths in Figure 20 do not vary significantly per element between the
single element and 6x1 fetch. This means that the spanwise addition of elements does not affect the streamwise
production of sound from the cuboidal elements. When a single row of trailing elements is included, the lead row’s
per element streamwise noise is actually enhanced by a factor of two. The lead row’s streamwise strength increases
further with the addition of a third row but remains approximately constant with the inclusion of successive rows.
Figure 21 suggests that the spanwise noise is enhanced by the presence of spanwise adjacent elements but that the
addition of trailing rows has no effect. A similar least-squares analysis of multi-hemispherical element fetches had
more instances of convergence errors. Therefore, no definitive conclusion is given for the variation in source
strengths throughout these surfaces.
Since there are two assumed sources for each roughness element, the strength analysis of a 42 element fetch
includes 84 sources. Figure 22 shows the normalized results of the least-squares estimation of the source strengths
for the fetch of 42 cubic elements. The results were also spanwise-averaged across the seven spanwise rows so that
only 14 curves are shown representing the streamwise and spanwise source components of each row. In general, the
relative estimated strength of the streamwise dipoles is stronger than that of the spanwise sources. The estimated
strengths are larger than that from Yang & Wang (2011) for the spanwise-averaged dipole strengths of a lead row of
cubes. The slopes differ slightly so that comparison of the streamwise and spanwise dipoles agree better at lower
non-dimensional frequencies. The spanwise results are within a factor of two for fh/U<0.8. The disagreement
increases to a factor of six at fh/U≈3. In most cases, the estimated streamwise strengths are greater than the LES
results over the entire non-dimensional frequency range the error increasing from a factor of 5 to 27 with increasing
non-dimensional frequency. The estimated streamwise strengths for the trailing rows in Figure 22 mostly fall below

8
American Institute of Aeronautics and Astronautics
the curve created by the lead row but usually remain within a factor of three. This agrees with the findings of Yang
& Wang (2011) that the lead row of cubic elements is the strongest producer of sound with source strengths
diminishing in the streamwise direction. These results also agree with the findings in Alexander et al. (2009) that
show the total non-dimensional unsteady drag spectrum of roughness fetch and therefore total produced sound
recorded in the streamwise direction of a similar roughness configuration to scale on the number of elements within
3-4dB.
The entire least-squares analysis is dependent upon the assumed presence of a streamwise and spanwise dipole
located at each roughness element location. To test this presumption, the analysis for the 42 element cubic fetch at
Uo =60m/s was completed assuming only a streamwise dipole, only a spanwise dipole, and a complete representation
including both. The results were compared to the corresponding single point measurement in Figure 10. The
comparison is shown in Figure 23. Using only a spanwise dipole, the analysis produced estimated far field results
that were far below the measured single point values since the compared position was in the streamwise direction.
Therefore, these results are not displayed in Figure 23. When assuming only a streamwise dipole, the measured far
field was overpredicted. The complete representation produced the best results agreeing with the measured far field
within 1dB. This corroborates the assumption of both spanwise and streamwise aligned dipoles that emanate from
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

each roughness element to produce the radiated far field noise.

IV. Conclusions

Wall pressure and far field measurements were made of the turbulent flow over varying size fetches of 3mm
cubic and hemispherical roughness. A linear microphone array was used to produce source maps of the roughness
noise and to estimate the individual strengths of roughness elements using a least-squares analysis technique. The
resultant data were compared with previous experimental studies and the LES-based predictions of Yang & Wang
(2011). A summary of the conclusions follow:

• Radiated noise is produced from the inhomogeneous pressure fluctuations on the surface of the roughness
elements and this cannot be inferred from the surrounding wall pressure field.

• Surface pressure measurements indicate the boundary layer flow over the 42 cubic element fetch is displaced
by the presence of the roughness and reattaches downstream. The hemispherical roughness does not show this
effect. Trailing rows of hemispherical roughness do not have as significant an upstream effect on the boundary layer
flow.

• The use of a least-squares analysis technique was improved upon from Alexander et al. (2010) by the
addition of data from multiple sensing locations. The technique was used successfully for measurements from a
single cubic roughness element case. Low signal-to-noise ratios and large source numbers increased the uncertainty
of source strength estimations from the hemispherical elements as well as the larger fetches of cubic elements.

• The existence of a spanwise dipole source is confirmed and the estimated strengths vary independently of the
streamwise dipole. Therefore, the estimated directivity pattern of an individual roughness noise source differs with
frequency.

• The least-squares analysis estimates have a similar shape and magnitude as the LES-based predictions of
Yang & Wang (2011) of streamwise and spanwise source strengths for both hemispherical and cubic elements.

• The addition of a trailing row of cuboidal elements enhances the production of streamwise noise from
upstream elements but has no effect on their spanwise source strengths. The addition of spanwise elements does
show some enhancement of noise in the spanwise direction.

Acknowledgments
The authors would like to thank the Office of Naval Research, in particular Dr. Ki-Han Kim, for their support
under grant N00014-08-1-0934.

9
American Institute of Aeronautics and Astronautics
References

Alexander W N, 2009, “Normalization of Roughness Noise on the Near-Field Wall Pressure Spectrum”, Master’s
Thesis, AOE Department, Virginia Tech. Avail: http://scholar.lib.vt.edu/theses/available/etd-06182009-104314/.

Alexander W N, Rasnick M B, Catlett M R, Devenport W, and Glegg S A L, 2009, “Boundary Layer Noise from
Discrete Roughness Elements”, AIAA/CEAS 15th Aeroacoustics Conference, Miami FL, May 11-13, AIAA-
2009-3310.

Alexander W N, Devenport W, Glegg S A L, and Van Buren R, 2010, “Directivity of Noise from Discrete Elements
in a Turbulent Boundary Layer”, AIAA/CEAS 16th Aeroacoustics Conference, Stockholm, SE, June 7-9, AIAA-
2010-3773.

Devenport W J, Wahl E A, Glegg S A L, Alexander W N, and Grissom D L, 2010, “Measuring Surface Pressure
with Far Field Acoustics”, Journal of Sound and Vibration, vol. 329, pp. 3958-3971.
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

Devenport W J, Grissom D L, Alexander W N, Smith B S, and Glegg S A L, 2011, “Measurements of Roughness


Noise”, Journal of Sound and Vibration, doi: 10.1016/j.jsv.2011.03.017.

Glegg S, and Devenport W, 2009, “Far-Field Sound from Rough-Wall Boundary Layers”, Proceedings of the Royal
Society. A, London, vol. 465, pp. 1717-1734.

Grissom D, 2007, “A Study of Sound Generated by a Turbulent Wall Jet Flow Over Rough Surfaces”, Ph.D.
Dissertation, AOE Department, Virginia Tech. Avail: http://scholar.lib.vt.edu/theses/available/etd-07192007-
123339/.

Hersh A S, 1983, “Surface Roughness Generated Flow Noise”, AIAA 8th Aeroacoustics Conference, Atlanta GA,
April 11-13, AIAA-83-0786.

Howe M, 1984, “On the Generation of Sound by Turbulent Boundary Layer Flow Over a Rough Wall”, Proceedings
of the Royal Society. A, London, vol. 395, pp. 247-263.

Liu Y, Dowling A, and Shin H C, 2008, “Measurement and Simulation of Surface Roughness Noise Using Phased
Microphone Arrays”, Journal of Sound and Vibration, vol. 314, pp.95-112.

Narasimha R, Narayan K Y, and Parthasarathy S P, 1973, “Parametric Analysis of Turbulent Wall Jets in Still Air”,
Aeronautical Journal, vol. 77, pp. 355-359.

Tester B J, and Glegg S A L, 2008, “A Review of Engine Noise Source Diagnostic Methods for Static Engine Tests,
with Phased Array and Polar Correlation Techniques”, 14th AIAA/CEAS Aeroacoustics Conference, 5-7 May
2008, Vancouver, BC, AIAA-2008-2854.

Wygnanski I, Katz Y, and Horev E, 1992, “On the Applicability of Various Scaling Laws to the Turbulent Wall
Jet”, Journal of Fluid Mechanics, vol. 234, pp. 669-690.

Yang Q and Wang M, 2010, “Boundary-Layer Noise Induced by a Roughness Patch”, 16th AIAA/CEAS
Aeroacoustics Conference, 7 - 9 June 2010, Stockholm, SE, AIAA-2010-3772.

Yang Q and Wang M, 2011, “Statistical Analysis of Acoustic-Source Field in Rough-Wall Boundary Layers” 17th
AIAA/CEAS Aeroacoustics Conference, 5-8 June 2011, Portland, OR.

10
American Institute of Aeronautics and Astronautics
4060
Aft most settling
chamber section Untreated
((acousticallyy treated)) settling
chamber Acoustically treated enclosure,
Flexible hose
section 1930mm wide
Center settling chamber section

1170
3150
with baffle (acoustically treated)

2134
3058
1247

1257
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

Large radius
L di ttrailing
ili edge
d
1247 914 1247 9.5mm thick
to promote Coanda effect
Aluminum test plate,
1524mm wide

Figure 1a. Side-view schematic of the Virginia Tech Anechoic Wall Jet Facility. All dimensions
shown in mm.

914
Acoustic baffle (to shield
acoustic instrumentation
from direct nozzle/jet noise)

330
Quarter ellipse Semi-circular trailing edge
nozzle profile
229

Flow
y
Cubic nozzle profile x Test plate

Figure 1b. Detail showing the nozzle and acoustic baffle. All dimensions shown in mm.

11
American Institute of Aeronautics and Astronautics
Streamwise S
Spanwise
i
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

Figure 2. 3mm cubic element fetches.

Mic Array

Roughness
Flow

Figure 3. Linear microphone array suspended by end-pins above the wall jet in the anechoic
chamber
chamber.

12
American Institute of Aeronautics and Astronautics
Array Position 4

Array Position 3

Array Position 2
A
Array Position
P iti 1
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

Flow

Figure 4. Four different microphone array positions in the anechoic chamber.

Plane B y
Plane C x
Microphone z
Array
1m
Plane A

Point
Source
Plane A Plane B Plane C
y, mm
z, mm
z,, mm

x, mm y, mm x, mm

Figure 5.
5 Point source image in three different planes.
planes

13
American Institute of Aeronautics and Astronautics
Flow

C1 H1

C2 H2

C14
C10 C12 H9
C13 C11
C9 C6 H8

C8 C3 C7 H7 H3 H6

C4 H4
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

C5 H5

Roughness Wall Pressure


Elements M
Measurement t Locations
L ti
a) b)

Figure 6. Wall pressure measurement locations in a 6x7 fetch of a) cubic elements b) and
hemispherical elements.

90 90 90

80 80 80

70 70 70

C1
dB

60 60 C6 60 C3
C1 C10
C7
C2 C11
C3
50 C3 50 50 C12
C8
C4 C13
C9
C5 C14
40 40 Smooth Wall 40
Smooth Wall C5
Smooth Wall
30 30 30
3 4 3 4 3 4
10 10 10 10 10 10

Freq., Hz Freq., Hz Freq., Hz


a) b) c)

Figure 7. Wall pressure variation in the a) streamwise direction, b) spanwise direction, c) and at
various roughness relative positions in the 3mm cubic element fetch.

14
American Institute of Aeronautics and Astronautics
90 90 90

80 80 80

70 70 70
dB

60 60 60
H1 H6
H2 H3 H3
H3 50 H7 50 H9
50
H4 H8 Smooth Wall
H5 Smooth Wall
40 Smooth Wall 40 40

30 30 30
3 4 3 4 3 4
10 10 10 10 10 10
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

Freq., Hz Freq., Hz Freq., Hz


a) b) c)

Figure 8. Wall pressure variation in the a) streamwise direction, b) spanwise direction, c) and at
various roughness relative positions in the 3mm hemispherical element fetch.

90

C4
80 H4
Smooth Wall

70
dB

60

50

40

30
3 4
10 10
Freq., Hz

Figure 9. Wall pressure comparison between similar measurement locations in the hemispherical
andd cubic
bi element
l fetch.
f h

15
American Institute of Aeronautics and Astronautics
50 30
Uo=20m/s
40
30m/s 20
40m/s
30
50m/s
60m/s 10
20

10 0
dB

dB
0 -10

-10
-20
20
-20

-30
-30
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

-40 -40
3 4 3 4
10 10 10 10
Freq., Hz Freq., Hz
a) b)

Figure 10. Single point far field measurement of the background noise (dashed) and noise (solid)
from a fetch of 3mm cubic elements a) raw b) and background-subtracted.

50 30

40 Uo=20m/s
20
30m/s
30 40m/s
50m/s 10
20 60m/s

10 0
dB

dB

0 -10

-10
-20
-20
-30
-30

-40 -40
3 4 3 4
10 10 10 10

Freq., Hz Freq., Hz
a) b)

Figure 11. Single point far field measurement of the background noise (dashed) and noise (solid)
from a single 3mm cubic element a) raw b) and background-subtracted.

16
American Institute of Aeronautics and Astronautics
z, mm
z, mm
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

x, mm x, mm
a) b)

Figure 12. Source maps at 10368Hz from Microphone Array Position 1 at Uo=60m/s for a) a single
cube b) and a 6x7 array of cubes in N2/(m4·Hz).

-7
x 10

7 Cube Fetch
Single Cube x10
6 Hemi Fetch x 100
Leading Edge of Roughness

Flow Mi Array
Mic A 5
N2/(m4·Hz)

Roughness Sweep 2

Fetch
1

0
1100 1150 1200 1250 1300 1350
x, mm

Figure 13. Diagram and plot of sweeps through the Position 1 source maps at z=0, 10368Hz
produced by various roughness configurations at Uo=60m/s.

17
American Institute of Aeronautics and Astronautics
-7
x 10
2.5
Cube Fetch
Single Cube x 10
Hemi Fetch x 100
2
Mic
Array

N2/((m4·Hz)
1.5

Flow
Sweep 1

0.5
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

Roughness
0
Fetch -150 -100 -50 0 50 100 150
z, mm

Figure 14. Diagram and plot of sweeps through the Position 2 source maps at x=1257mm,
10368Hz produced by various roughness configurations at Uo=60m/s.

-11
10

Cubic Element
-12
10 Hemispherical Element

-13
10 -3.5
f
N2/Hzz

-14
10

-15
10

-16
10

-17
10
4
10

Freq., Hz
Figure 15. Spanwise (dotted) and streamwise (solid) source strengths for a single 3mm cubic element
and a single 3mm hemispherical element at a nozzle exit velocity of 60m/s.

18
American Institute of Aeronautics and Astronautics
3872Hz 10368Hz
90 90
10 10
120 60 120 60
8 8

6 6
150 30 150 30
4 4

2 2

180 0 180 0

210 330 210 330

240 300 240 300


270 270
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

Figure 16. Relative directivity from a single cubic element at Uo=60m/s at various frequencies.
Angles measured from the flow direction.

-11
10

-12
10 Spanwise Dipole
Streamwise Dipole

-13
10

-14
10
N2/Hz

60m/s
-15
10
50m/s
-16
10

-17
40m/s
10
30m/s
4
10
Freq., Hz
Figure 17. Source strength comparison for the single cubic element at various nozzle exit velocities.

19
American Institute of Aeronautics and Astronautics
-2
10

LES Cuboidal Elements


-3
3
10
Spanwise Dipole
Streamwise Dipole
-4
10
N2/Hzz/(ρ2U3h5)

-5
10

-6
10
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

-7
10
LES Hemispherical Elements
-8
10
0
10
Strouhal Number, f h/U
Figure 18. Normalized source strength spectra for single cubic and hemispherical elements at various nozzle exit velocities
compared to the LES streamwise (solid) and spanwise (dashed) dipole strength estimations of Yang & Wang (2011) for
cubes (black) and hemispheres (green). Symbols same as Figure 19 with addition of downward pointing triangle for the
single hemispherical element at Uo=60m/s.

2.8
Ratio of Source Sttrength,

2.6
nwise

2.4
Streamwise/Span

2.2

1.8
Slope=0.12

1.6

1.4

10 12 14 16 18 20 22

U, m/s

Figure 19. Integrated streamwise and spanwise source strength ratio for the single cubic element
over a range
g of edge
g velocities.

20
American Institute of Aeronautics and Astronautics
0
10

-1 6x1
10 6x2
6x3
-2
10 6x4
6x5
N2//Hz/(ρ2U3h5)
6x6
-3
10 6x7
Single Cube
-4
10

-5
10
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

-6
10

-7
10
0
10
Strouhal Number, f h/U
Figure 20. Normalized spanwise-averaged streamwise dipole strengths for the lead row of varying size fetches of 3mm
cubic elements at Uo=60m/s.
-1
10
6x1
6 1
-2
10 6x2
6x3
6x4
-3
10 6x5
6x6
N2/Hz/(ρ2U3h5)

-4 6x7
10
Single Cube

-5
10

-6
10

-7
10

-8
10
0
10
Strouhal Number, f h/U

Figure 21. Normalized spanwise-averaged spanwise dipole strengths for the lead row of varying size fetches of 3mm
bi elements
cubic l t att Uo=60m/s.
60 /

21
American Institute of Aeronautics and Astronautics
-1
10

-2 Spanwise Dipole
10
Streamwise Dipole

-3
N2/Hz/(ρ2U3h5) 10

-4
10

Single Cube
-5
5
10

-6
10 Yang & Wang (2011)
Downloaded by BEIHANG UNIVERSITY on July 17, 2021 | http://arc.aiaa.org | DOI: 10.2514/6.2011-2740

Cuboidal Elements
-7
10
0
10
Strouhal Number, f h/U
Figure 22. Normalized spanwise-averaged streamwise and spanwise dipole strengths of 6x7 array of cubic elements at
Uo=60m/s, circle-1st row, square-2nd row, diamond-3rd row, upward pointing triangle-4th row, downward pointing triangle-5th
row, right pointing triangle-6th row, pentagram-7th row compared to streamwise (solid) and spanwise (dashed) dipole strengths
from LES of Yang & Wang (2011) and a single cube.

15

10

5
dB

0
w/ Spanwise Dipole
w/o Spanwise Dipole
Measured Spectrum

-5

-10
3 4
10 10
Freq., Hz

Figure 23. Comparison of far field estimations using least-squares source strength analyses for the 6x7 fetch of cubic
elements with
ith and without
itho t the assumption
ass mption of a spanwise
span ise dipole with
ith meas
measured
red single point far field data.
data

22
American Institute of Aeronautics and Astronautics

You might also like