Professional Documents
Culture Documents
CrossMark
View Export
Online Citation
Effect of Mach number on the aeroacoustic feedback loop generating airfoil tonal noise
Physics of Fluids (September 2022)
Acoustic receptivity in the airfoil boundary layer: An experimental study in a closed wind tunnel
Physics of Fluids (April 2023)
AFFILIATIONS
1
Department of Mechanics and Aerospace Engineering, Southern University of Science and Technology, Shenzhen 518055, China
2
Guangdong Provincial Key Laboratory of Turbulence Research and Applications, Southern University of Science and Technology,
Shenzhen 518055, China
3
Center for Complex Flows and Soft Matter Research, Southern University of Science and Technology, Shenzhen 518055, China
4
Department of Naval Architecture and Ocean Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China
a)
Electronic mail: yangyn@sustech.edu.cn
b)
Author to whom correspondence should be addressed: liuy@sustech.edu.cn
ABSTRACT
a flat plate). As a consequence, the average scaling law fs u1:5 1 was Some controversy exists in the literature about the flow around a
proposed.1 However, as pointed out by Tam,14 the wake vortices of symmetric NACA 0012 airfoil at zero AoA. In a numerical study, Tam
streamlined airfoils were fundamentally different from those of a bluff and Ju19 observed only a single tone and explained its presence by a
body and simple vortex shedding cannot explain multiple tones. Kevin–Helmholtz type wake instability. The authors noted, however,
Therefore, a feedback mechanism was proposed by Tam.14 Tam14 that this result may not be representative for experimental studies since
conjectured that the perturbations from the boundary layer propagate the simulated flow was free of environmental perturbations (free
downstream and grow due to instability, resulting in a sufficiently stream turbulence and acoustic). More recently, Ricciardi et al.,20 sim-
strong lateral vibration in a highly localized region of the wake acting ulated a NACA 0012 airfoil at a ¼ 0 and Re ¼ 1:0 105 . Vortex
as a noise source. The acoustic waves emitted in the upstream direc- roll-up upstream of the trailing edge was observed on one side, even
tion, in turn, enhanced the oscillation of the boundary layer. In this though the free stream and grids were symmetric for the case of
model, the noise source region in the wake and the receptivity point at a ¼ 0 . The frequency of the main tone was proposed to be related to
the trailing edge form a feedback loop. The phase condition necessary flow instabilities that led to the development of vortical structures, and
for the existence of such a self-excited feedback loop was consistent the side tones were attributed to the flapping effect of the separation
with the existence of several distinct tones at a given velocity. By apply- bubble. Pr€obsting et al.,3 showed experimental results of a NACA 0012
ing the neutral stability curve for a flat plate to the data of Paterson under similar conditions at incidence near zero. Also these experimen-
et al.,1 Tam14 found that all the experimental data points were con- tal results showed vortex roll-up upstream of the trailing edge and a
tained within the frequency band required by stability consideration. minor asymmetry in the mean flow field, with flow separation further
Since then, the concept of the feedback mechanisms has been upstream on one side for some Reynolds number and on the opposite
refined,6,7,11 with the main difference being the locations of the noise side for other Reynolds numbers. As in those experiments, the inclina-
source and receptivity region. The noise source region was suggested tion of the airfoil was not changed between measurements at different
to be at the trailing and the receptivity point assumed to be at the max- Reynolds number; the separation behavior indicates a bifurcation
imum velocity point6 or at the separation point11 around the airfoil. rather than a bias error due to misalignment or flow curvature.
The instability mechanism was later postulated to be the To assess the features of the flow contributing to the noise gener-
Tollmien–Schlichting (T–S) waves,15 which are convected down- ation mechanism, POD analysis has been proven to be an efficient
stream, amplified through a separation bubble, and noise is radiated as tool for data reduction of complex flow fields.21 POD analysis has
the resulting vortices pass the discontinuity posed by the trailing edge been applied to the velocity field for a cylinder wake and slat cove by
FIG. 1. Experimental setup. (a) Setup inside the anechoic chamber; (b) illumination area of the PIV setup (bottom view).
on the aerodynamic and acoustic performance of the wind tunnel, the resulting in zero lift (CL ¼ 0), the CL values for two larger, geometric
reader may refer to Yang et al.25 angles of attack (62 ) were interpolated linearly to obtain an estimate
A NACA 0012 airfoil with a span of 0.55 and chord length c of for the angle corresponding to a zero lift condition (aCL ¼0 ). The angle
0.3 m was chosen as the test model and mounted vertically between was consequently adjusted to the interpolated value using a stepper
the two end plates. motor. Measurement during the actual experiment resulted in a lift
In order to compare the present test conditions with previous coefficient of CL ¼ 0:0042 (uncertainty of 0.0015 with 95% confidence
studies, a summary of the airfoil type, chord length, test section size, level) for this condition. With an estimated CL a slope of 0.042 , the
and Reynolds number is shown in Table I. The majority of previous actual angle of attack can be estimated to be equivalent to 0.1 .
studies focused on the NACA 0012 airfoil,1,2,6,7,9,11,18,20 which is also The model consists of acrylic glass with a polished surface, ren-
adopted in our research. Airfoils with the same chord length were dering the model transparent and thus allowing for light transmission.
spacing of 0.38 mm. The sampling rate of the PIV measurement was where U(X, t) is the instantaneous, U ðX; tÞ the time-averaged, and
2645 Hz, and the duration was 2 s. The PIV measurement was syn- U 0 ðX; tÞ the fluctuating flow field. The proper decomposition is then
chronized with the acoustic measurements allowing for time- applied on the auto-correlation matrix CNs Ns as reported in
correlation of the two signals. Sirovich,32 Kourentis and Konstantinidis,33 Schrijer et al.,34 Chen,35
Yang et al.,36 which is defined as
C. Uncertainty analysis ðð
1
The uncertainty analysis includes bias and random errors. The Cij ¼ U 0 ðX; ti ÞU 0 ðX; tj ÞdX; (2)
NS
bias error mainly considers the peak-locking as a result of the large
imaging aperture (f/4), which yields diffraction spots smaller than the where Ns is the number of snapshots; the integral is evaluated here as a
pixel diameter. This effect was mitigated by slightly defocusing the summation for the discrete measurement points.
images,26 adjusting the particle image diameter to approximately 2 The eigenvalues kn and the eigenvectors An ðXÞ of the auto-
pixels. The verification was performed by inspecting the histogram of correlation matrix satisfy
particles image displacement.27 The focal setup was adjusted until no
CNs Ns An ðXÞ ¼ kn An ðXÞ: (3)
obvious peak-locking was noticed, and thus the bias error could be
regarded negligible in the current study. The error due to the finite The POD mode, which represents the coherent flow pattern, is
spatial resolution depends on the characteristic length to be accurately determined by
measured in the PIV velocity fields, which is the distance between dis-
0
crete vortices in this case. It was reported that the measurement was Un ðXÞ ¼ Rn¼N
n¼1 U ðX; tn ÞAn ðXÞ:
S
(4)
within 95% precision with a multipass cross correlation algorithm and Furthermore, the measurement data contain temporal informa-
window deformation when the window size is smaller than 0.6 times tion that can be retrieved by projecting each snapshot onto the basis
the length of the scale to be measured.28 Therefore, with a 1.52 window functions,
size, flow structures down to 2.53 mm (1/8 of the distance between
ðð
adjacent shedding vortices) can be measured within 95% precision.
The random error in the PIV data was estimated by the image cn ðtÞ ¼ U 0 ðX; tm ÞUn ðXÞdX; (5)
matching method as introduced by Sciacchitano et al.,29 which
where the coefficient cn indicates the correlation between the nth basis
and –0.03 as shown in Fig. 3(b). At the same chordwise position, the
wall-parallel velocity at the lower side is slightly higher than that on
the upper side, which is a result of the earlier separation on the upper
side. The boundary layer thickness ðd99 Þ on the upper and lower sides
is approximately equal at 0:0228c at xt ¼ 0:03c. It has been shown
earlier that the tone generation occurred together with inflections in
the measured velocity profile.7 These inflections are clearly visible in
the boundary layer velocity profiles at xt =c ¼ 0:15; 0:11, and
–0.07 as shown in Fig. 3(b).
Significant streamwise and vertical velocity fluctuations are
observed in the boundary layer and near wake as shown in Fig. 4. The
contours for both velocity components are approximately symmetric
with the maximum fluctuation magnitude located more upstream on
the upper side than on the lower side. The maximum velocity fluctua-
tions appear at xt =c ¼ 0:94c and xt =c ¼ 0:97c on the upper and lower
sides, respectively, which are downstream of the reattachment point of
the separation bubble. There is also another local maximum of the ver-
tical velocity fluctuation located 0:05c downstream of the airfoil trail-
FIG. 2. Noise spectra of the airfoil and facility effects. ing edge, which was observed at the same position by Chong and
Joseph.8 Figure 5 shows that the fluctuation profiles at xt =c ¼ 0:15;
frequency or introduce new tones to the spectrum. Therefore, it can be 0:11; 0:07, and –0.03 are extracted from the above contours. The
0
stated that the PIV system has little effect on the tonal noise generation values of ut;rms show evidence of three peaks, which are discernible at
of the airfoil. xt =c ¼ −0.15 as indicated by red circles. This triple-peak structure of
the streamwise velocity fluctuations is one of the features often
0
IV. FLOW FIELD NEAR THE AIRFOIL TRAILING EDGE reported for tonal noise emission.2,7,9 The values of un;rms have one
A. Time-averaged flow field peak in the measurement domain and the peak value increases from
FIG. 3. (a) Contour of time-averaged flow field colored by horizontal velocity component u. Nodal points of separation bubbles are indicated by N1 and N2 on the upper and
lower sides, respectively. (b) Boundary layer velocity profiles at xt =c ¼ 0:15; 0:11; 0:07, and –0.03.
0 0 0
FIG. 4. Velocity fluctuation contours. (a) urms ; (b) vrms , and the dashed lines indicate the position with maximum un;rms values.
bubble of a SD7003 airfoil with two velocity components from PIV 0.88 < x=c < 0.94, vortex roll-up is observed. Spatial coherence
measurement has been reported.39 In the present case, vortices form- reduces from x=c ¼ 0:94, which may be attributed to the vortex
ing over the separation bubble have a larger spanwise coherence. The breakup, as the vortex is convected downstream.39,40 The location of
upstream vortex ridges are uniform in space and time domains. vortex breakup corresponds to the location of accelerated increase in
For x=c 0.88, i.e., before the nodal point of the separation velocity fluctuations seen in Fig. 4. The vortex breakup on the pressure
bubble, the vorticity is concentrated in the laminar shear layer. For surface was also reported for a NACA 0012 airfoil at a ¼ 6 .17
Considering the vortex breakup as observed in the PIV results and the
significance of spanwise coherence for the aeroacoustic emissions at
the trailing edge, this indicates that results obtained from numerical
studies based on the 2D flow assumption have to be considered with
care as also noted by Pr€obsting et al.9 and Turner and Kim.41
The time history of the velocity showed periodic fluctuations.
Frequency spectra are used to analyze the energy content further. The
0
PSD values of un are shown in Fig. 7 for the probes located on the line
0
along maximum un;rms as indicated in Fig. 3(a) (the contour in Fig. 3 is
0
colored by vrms , but the dashed lines are plotted for the maximum
0
value of un;rms ). In order to achieve a similar frequency resolution as
FIG. 6. Space-time development of the vortex structures visualized by the k2 for PSDp0 p0 , the number of samples per window for estimating
criterion. PSDu0n ;u0n was set to 660 with an overlap of 50%, which yielded a
0
FIG. 7. PSD values of velocity fluctuation un . (a) Upper side and (b) lower side.
0
frequency resolution of 2.583 Hz. The velocity un was chosen for anal- the main tone frequency fc=u1 ¼ 6.31. Amplitude modulation with
ysis as its contribution to the airfoil tonal noise generation prevails lower frequency corresponds to the frequency interval of discrete tones
over that of the wall-tangential velocity components.42 Both, similarity fc=u1 ’ 0.5.
between the two velocity spectra and between the velocity spectra and
the spectrum of acoustic pressure (Fig. 2) is observed. Discrete peaks C. Convective velocity
can be identified at fc=u1 ¼ 5.76, 6.31, 6.86, and 7.41. The broadband Section IV B described the vortex roll-up and convection domi-
spectra level increases downstream from xt =c ¼ 0:12 to –0.03, nating the flow field. Therefore, the spatiotemporal evolution of the
which is consistent with the observation that the velocity fluctuation is
have observed, based on the results of a numerical simulation, that the sources,9,17,22 is utilized in this study. This method is based on the
emitted acoustic wave is weak in comparison to the case with out-of- cross correlation between the vertical velocity component measured
phase vortices. The results presented here confirm this description by PIV of vðxref ; tn Þ and the acoustic pressure pðx; tn Þ at the micro-
experimentally. It should be mentioned that the shear layer without phone position x. The normalized cross correlation function with a
vortex roll-up is also reported to be responsible for the low amplitude time shift s is defined by
region of sound pressure,7,9 yet the shear layer can be regarded as dis- 1 N
tributed vortex and they are “in-phase” on the two sides, and thus this Rn¼1 vðxref ; tn Þp0 ðx; tn þ sÞ
Rv;p ðsÞ ¼ N pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; (8)
explanation is a special case for in-phase vortices. The lower acoustic
hv2 ðxref ; tÞihp02 ðx; tÞi
emission in the case of in-phase vortex shedding from the two sides
can be explained with the reduced pressure difference between the two where N is the number of PIV measurements rather than that of the
sides of the airfoil, and thus a lower pressure disturbances scattered by sound pressure measurements, as the latter sampling frequency is sig-
the discontinuity posed by the trailing edge as suggested by the model nificantly higher. The correlation map in Fig. 17 has an alternative pat-
of Amiet.43 tern of negative and positive values, which is similar to the results of
airfoils9,17 and cylinders22 with tonal noise generations. The region
with strong periodic patterns of the correlation coefficient was
C. Comparison of noise source region with POD mode reported to be dependent on the separation region.17 Comparing the
pattern correlation map of qv;p in Fig. 17 with the first and second POD mode
A method for combining simultaneous PIV and acoustic mea- patterns as shown in Figs. 13(a) and 13(b), a high similarity in terms
surements, which has been used for the analysis of aeroacoustic of periodic and asymmetric patterns is noticed. The POD modes
FIG. 14. Time histories of sound pressure and POD mode time coefficients. (a)
Sound pressure; (b) POD mode coefficient c1; (c) c3; and (d) c5. The dashed lines
represent the velocity fluctuations in the flow field, which are the
source of hydrodynamic pressure fluctuations scattering noise into the
far field at the trailing edge. The region with strong velocity fluctua-
tions has a strong correlation with the far field noise, and thus the first
and second POD modes with higher turbulent energies show a similar
pattern to the correlation map.
The correlation coefficient between the first POD mode and
sound pressure is shown in Fig. 18. The maximum magnitude of the
correlation coefficient reaches 0.84 for the first mode, which is larger
than that (0.4) between a sing velocity probe and sound pressure as
shown in Fig. 17. This behavior implies that the POD analysis has FIG. 15. Flow fields in (a)–(d) correspond to instants of “J1” “J4” as represented
in Fig. 14, respectively.
greatly suppressed the background noise22 and filtered out flow field
fluctuations that are incoherent with the large-scale motion.30 In the
case of the third mode, the correlation coefficient reaches a value of wall-parallel velocity fluctuations are observed, which were previ-
approximately 0.24 at its maximum which means its contribution to ously described as necessary conditions for the tonal noise genera-
the noise emission is relatively weak. tion on the airfoil. The instantaneous flow fields show the
development of shear layer, vortex roll-up, and vortex breakup
VI. CONCLUSION
from the upstream to the downstream region near the airfoil trail-
For the NACA 0012 airfoil subject to this study, the sound pres- ing edge. The coherent vortex shedding is explained by the acous-
sure spectrum shows a broadband hump superimposed by discrete tic feedback mechanism as outlined in the introduction. The
tones, which is consistent with results reported widely in literature. shape of the wall-normal velocity spectra agree well with the
The interval of the discrete tones agrees well with the feedback model acoustic ones. The energy content increases downstream for all
suggested by Arbey and Bataille.6 broadband components, yet the tone magnitude reaches a maxi-
The time-averaged flow field shows laminar separation bub- mum value around xt =c ¼ −0.07. Corresponding to the vortex
bles on both the upper and lower sides. The inflection point shedding, the convective velocity is around 0.42 u1 , and the wave-
on the time-averaged velocity profile and triple-peaks for the length is estimated to be 0.07c.
FIG. 17. Contour of the correlation coefficient between the vertical velocity compo-
nent and the acoustic pressure signal. The time delay has been corrected for the
propagation time between the trailing edge and the location of the microphone.
FIG. 16. Flow fields in (a)–(d) correspond to instants of “K1” “K4” as repre- noise generation. The correlation map between the vertical velocity
sented in Fig. 14, respectively. component and the far-field sound pressure is also calculated and
compared with the POD mode pattern. It is found that the correla-
tion map has a great similarity with the first and second POD
POD analysis revealed the most energetic modes of the fluctu- modes, which further confirms that the far-field noise emission is
ating field. Time history comparison and correlation with the far- strongly dependent on the most energetic POD modes.
field sound pressure linked these modes to the acoustic emission. The results thus provide empirical evidence for nature of the
Amplitude modulation was observed for the coefficients of the first interaction of vortex shedding from the two sides of the airfoil and its
(c1) and third (c3) POD modes, which is consistent with the far- influence on noise emission as described by Desquesnes et al.2 based
field sound pressure time history. The first and second modes rep- on simulation results.
resent out-of-phase vortex shedding on the upper and lower sides,
which accounts for the high amplitude region of noise emission. ACKNOWLEDGMENTS
On the contrary, the third and fourth modes represent in-phase
vortex shedding on the two sides, which accounts for the low Yannian Yang and Yu Liu would like to thank National Natural
amplitude region of noise emission. The higher modes which only Science Foundation of China (Grant Nos. 11802114 and 92052105), the
show stochastic contributions to the turbulent kinetic energy are Department of Science and Technology of Guangdong Province (Grant
regarded as insignificant for the flow field dynamics relating to No. 2019B21203001), the Natural Science Foundation of Guangdong
22
Province (Grant No. 2018A0303130142), and Science, Technology, and A. Henning, K. Kaepernick, K. Ehrenfried, L. Koop, and A. Dillmann,
Innovation Commission of Shenzhen Municipality (Grant No. “Investigation of aeroacoustic noise generation by simultaneous particle image
velocimetry and microphone measurements,” Exp. Fluid 45, 1073–1085 (2008).
RCBS20200714114941227) for the support. 23
J. H. M. Ribeiro and W. Wolf, “Identification of coherent structures in the flow
Stefan Pr€obsting and Ye Li would like to thank National past a NACA0012 airfoil via proper orthogonal decomposition,” Phys. Fluids
Natural Science Foundation of China (Grant Nos. 51761135012 29, 085104 (2017).
and 11872248) and Ministry of Science and Technology of China 24
T. Ricciardi, J. Ribeiro, and W. Wolf, “Analysis of coherent structures in large-
(Grant No. 2017YFE0132000) for the support. eddy simulations of a NACA0012 airfoil,” AIAA Paper No. 2019-0320 (2019).
25
Y. Yang, Y. Liu, R. Liu, C. Shen, P. Zhang, R. Wei, X. Liu, and P. Xu, “Design,
DATA AVAILABILITY validation, and benchmark tests of the aeroacoustic wind tunnel in SUSTech,”
The data that support the findings of this study are available Appl. Acoust. 175, 107847 (2021).
26
M. Raffel, C. Willert, W. S, and J. Kompenhans, Parcile Image Velocimetry: A
from the corresponding author upon reasonable request.
Practical Guide (Springer, 1998).
27
B. Wieneke, “Piv uncertainty quantification from correlation statistics,” Meas.
REFERENCES Sci. Technol. 26, 074002 (2015).
1
R. W. Paterson, P. Vogt, M. R. Fink, and C. Munch, “Vortex noise of isolated
28
F. F. J. Schrijer and F. Scarano, “Effect of predictor-corrector filtering on the
airfoils,” J. Aircraft 10(5), 296–302 (1973). stability and spatial resolution of iterative PIV interrogation,” Exp. Fluid 45(5),
2
G. Desquesnes, M. Terracol, and P. Sagaut, “Numerical investigation of the 927–941 (2008).
tone noise mechanism over laminar airfoils,” J. Fluid Mech. 591, 155–182
29
A. Sciacchitano, B. Wieneke, and F. Scarano, “PIV uncertainty quantification
(2007). by image matching,” Meas. Sci. Technol. 24, 045302 (2013).
3
S. Pr€obsting, F. Scarano, and S. C. Morris, “Regimes of tonal noise on an airfoil
30
B. van Oudheusden, F. Scarano, N. van Hinsberg, and D. Watt, “Phase-
at moderate reynolds number,” J. Fluid Mech. 780, 407–438 (2015). resolved characterization of vortex shedding in the near wake of a square-
4
V. Golubev, “Recent advances in acoustics of transitional airfoils with section cylinder at incidence,” Exp. Fluid 39, 86–98 (2005).
31
feedback-loop interactions: A review,” Appl. Sci. 11(3), 1057 (2021). Y. Yang, T. Zhou, A. Sciacchitano, L. L. M. Veldhuis, and G. Eitelberg,
5
L. Nguyen, V. Golubev, R. Mankbadi, G. Yakhina, and M. Roger, “Numerical “Experimental investigation of the impact of a propeller on a streamwise
investigation of tonal trailing-edge noise radiated by low Reynolds number air- impinging vortex,” Aerosp. Sci. Technol. 69, 582–594 (2017).
foils,” Appl. Sci. 11(5), 2257 (2021).
32
L. Sirovich, “Turbulence and the dynamics of coherent structures, part I:
6
H. Arbey and J. Bataille, “Noise generated by airfoil profiles placed in a uni- Coherent structures,” Q. Appl. Math. 45, 561–571 (1987).
form laminar flow,” J. Fluid Mech. 134, 33–47 (1983).
33
L. Kourentis and E. Konstantinidis, “Uncovering large-scale coherent struc-
7
E. C. Nash, M. V. Lowson, and A. McAlpine, “Boundary-layer instability noise tures in natural and forced turbulent wakes by combining PIV, POD, and