Вы находитесь на странице: 1из 13

International Journal of Heat and Fluid Flow 82 (2020) 108560

Contents lists available at ScienceDirect

International Journal of Heat and Fluid Flow


journal homepage: www.elsevier.com/locate/ijhff

The reduction of noise induced by flow over an open cavity T


a b a,⁎
Seung Joong Kim , Wei-Xi Huang , Hyung Jin Sung
a
Department of Mechanical Engineering, KAIST 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Korea
b
Department of Engineering Mechanics, Tsinghua University Beijing 100084, China

ARTICLE INFO ABSTRACT

Keywords: Large eddy simulations were performed and Lighthill's analogy applied in order to predict the noise induced by
Large eddy simulation the flow of a turbulent boundary layer (ReD = 12,000 or Reθ = 300) over an open cavity (length/depth = 1).
Open cavity The effects of three simple geometrical changes, namely lids at the upstream edge (C10 and C20) and a
Self-sustained oscillations chamfered edge (C23) on the downstream side of the cavity, on the internal circulation flows and shear-layer
Flow-induced noise
flow oscillations were determined and their effectiveness in reducing flow-induced noise was assessed. These
Lighthill's analogy
small modifications were found to weaken the circulating flows inside the cavity and to suppress the pressure
fluctuations near its downstream edge. As the lid length increases, the turbulence intensity inside the cavity is
suppressed. The acoustic sources are concentrated at the downstream edge. The wall pressure fluctuations at the
fundamental frequency are slightly higher for C23. The energy densities at the fundamental frequency are lower
by 87.5% (C10), 30% (C20), and 18.7% (C23) than that of C00. The contributions to the acoustic directivity
patterns of the dipole and quadrupole sources are oriented toward 135 and 45° respectively. The ratios of
decrease of the dipole and quadrupole sources are 93% (C10), 39% (C20), and 23% (C23) with respect to C00.

1. Introduction explained by using the Rossiter mechanism (Rossiter, 1964; Heller and
Bliss, 1975). Small periodical vortical structures separate at the up-
As conveyances such as ground, flight, and marine vehicles cruise stream edge, collide with each other at the downstream edge, and
through fluids, their surface discontinuities and non-uniform geome- amplify shear-layer instabilities while circulating inside the cavity
tries disturb the progress of the turbulent boundary layer over their (Tam and Block, 1978; Rockwell and Naudascher, 1979; Keller and
surfaces. The large unsteady pressure fluctuations that arise due to Escudier, 1983). Ahuja and Mendosa (1995) found in wind-tunnel ex-
these irregular geometries are not only the primary sources of flow- periments that there is a resonant frequency of fluctuations, which is
induced noise but also the root causes of interior noise (Dowell, 1969; affected by not only the cavity dimensions but also by the character-
Howe, 1998; Ask and Davidson, 2010). Even simple, small and uneven istics of the incoming boundary layer, particularly its thickness. The
geometries, such as those due to a panel mismatch gap, the protrusion acoustic intensity increases with the thickness of the boundary layer.
of rivets and bolts, or low tolerance between different kinds of parts, Rockwell et al. (2003) provided experimental evidence for the gen-
generate unpleasant noise with a unique peak frequency eration of flow-induced noise when the incoming flow is fully turbulent.
(Ffowcs Williams, 1965; Wang et al., 1996; Kim and Sung, 2018). To Computational studies have focused on the accurate prediction of the
examine the characteristics of such flows and investigate the me- characteristics of self-sustained oscillations at relatively low Mach
chanism of flow-induced noise generation i.e., to determine the nature numbers. Rowley et al. (2002) simulated short two-dimensional cavities
of unsteady pressure fluctuations in complex turbulent flows, highly by directly solving the compressible Navier‒Stokes equations in order
accurate flow simulations are required (Jeon et al., 2003). To this end, to examine their wake mode oscillations. Larsson et al. (2004) simu-
an open cavity with a simple geometry that exhibits energetic self- lated the production of flow-induced noise inside an open cavity with
sustained oscillations of pressure and velocity is often employed to L/D (length/depth) = 4 at Mach number 0.15 by using an identical
explore the mechanisms of flow-induced noise generation and methods simulation method. The far-field radiation flow noise sources were
for noise reduction. identified by using an acoustic analogy of the modified Curle's (1955)
Numerous studies of the noise induced by flows over an open cavity integral equation, and compared with directly obtained pressure fluc-
have been carried out. The acoustic feedback phenomenon was first tuations. The wall pressure fluctuations on the downstream cavity wall


Corresponding author.
E-mail address: hjsung@kaist.ac.kr (H.J. Sung).

https://doi.org/10.1016/j.ijheatfluidflow.2020.108560
Received 29 October 2019; Received in revised form 22 December 2019; Accepted 7 February 2020
Available online 29 February 2020
0142-727X/ © 2020 Elsevier Inc. All rights reserved.
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

are the primary contributor to the intensity of the acoustic sound, and u¯ i
= 0,
are reflected from the downstream edge to the upstream edge. xi (1)
Howe (1997) developed a linear theoretical analysis of cavity tonal
noise. This analysis, in combination with the use of a tailored Green's u¯ i u¯ i u¯ j 1 p¯ ( ij ij )
function, has since become the fundamental method for accounting for + = + ,
t xj xi xj (2)
the monopole contribution to the far-field resonance frequency. The
flow over a conjugated forward–backward facing step at where xi are the directions in Cartesian coordinates, ui are the corre-
Reh = 1.7 × 105 was investigated by Addad et al. (2003), who used a sponding velocity components, p is the pressure, and
hybrid approach to obtain the far-field acoustic noise. The near-field ij = [( u¯ i / xj + u¯ j / x i ) 2/3( u¯k / xk ) ij] is the molecular viscosity
acoustic noise was explored with large eddy simulations (LESs), and the stress tensor. The bars over the physical variables denote the perfor-
linearized Euler equations were adopted to calculate the pressure pro- mance of the grid filtering operation, G(x, xʹ). The subgrid-scale (SGS)
pagation at far-field locations. Hao et al. (2013) performed LESs of an stress tensor, ij = ui uj ui u j is resolved with the dynamic
incompressible flow with Reθ = 4755 to examine the sound generated Smagorinsky model, as proposed by Germano et al. (1991). The cal-
by the flow over small gaps. They investigated 9 cases with variations in culation of the SGS for a turbulent viscosity uses an equation that is
the gap design parameters, namely the height of the upstream wall, the identical to that of the standard Smagorinsky model,
width of the gap, and the height of the downstream wall. The acoustic
noise sources of the turbulent flows were extracted by using Lighthill's SGS = (CS )2 2S¯ij S¯ij , (3)
analogy (1952), and found to be generally concentrated at the down-
stream corner. The fundamental bases of the generated sounds were where CS is the Smagorinsky constant calculated dynamically by ap-
calculated by using an approximate tailored Green's function and the plying a test filter, Δ is the test filter width (or grid spacing), and
boundary-element method. Kim and Sung (2006) performed direct S¯ij = ( u¯ i / x j + u¯ j / x i )/2 is the filtered strain-rate tensor. A 3rd-order
numerical simulations of a flow passing over a small bump. Far-field monotonic upwind scheme was used for the conservation law
noise generation was predicted for various bump heights by using (van Leer, 1979) and a 2nd-order temporal discretization scheme was
Lighthill's analogy and Curle's integral equation. Lighthill's acoustic used to obtain accurate numerical data.
sources were found to be highest in the vicinity of the downstream
edge, and the dipole sources were found to make the dominant con-
tribution to the intensity of the far-field acoustic noise. The intensity of 2.2. Aeroacoustic theory
the dipole source contribution increases with the bump height.
Chatellier et al. (2004) and Lee et al. (2010) confirmed the nature of Lighthill (1952) showed that an acoustic source inside a turbulent
flows over an open cavity and identified the pressure spectral peaks. flow field is completely converted into pressure fluctuations over the
Forestier et al. (2003) and Marsden et al. (2012a, 2012b) determined free flow field region. The far-field density fluctuations, ρʹ(x, t), that
the flow regimes by varying the cavity geometry at relatively high result from the turbulent interaction can be obtained by rearranging the
Reynolds number. Navier−Stokes equation to obtain the following equation:
Many research groups have reported success in the reduction of 2 2 2T
ij
noise induced by flow over an open cavity. Sarohia & Massier (1977), c02 = ,
t2 x i2 xi xj (4)
Lamp & Chokani (1996), Zhang et al. (1999), Raman & Raghu (2004),
and Arunajatesan et al. (2009) used massive flow injection at a base
where Tij = ui uj ij + (p ) ij is the Lighthill's stress tensor. c0,
co2
wall and an upstream edge to control the internal circulation flows and
pʹ, δij, and τij are the ambient speed of sound, the pressure fluctuations,
shear-layer flow oscillations. Similarly, Sarno & Franke (1994) and
the Kronecker delta function, and the viscous stress tensor, respectively.
Ukeiley et al. (2004) utilized mechanically operated flaps and fixed
Curle (1955) derived the integral solutions for the encounter between
fences placed on an upstream edge to cancel out the oscillations in a
flows in the free field and fixed rigid bodies by using the framework of
shear-layer flow. These methods provide significant reduction in flow-
Lighthill's acoustic analogy to predict the far-field density fluctuations:
induced noise. However, they are difficult to apply in engineering ap-
plications because of the complexity of their constituents, such as 1 2 Tij (y, )
hinges, flaps, additional parts for flow injection, and flow controllers. (x , t ) = 2
dV (y )
4 c0 x i x j V r
The objective of the present study was to test methods for the reduction
1 Pij (y , )
of flow-induced noise that use only simple geometrical changes near the + nj dS (y ),
upstream and downstream edges of an open cavity: lids at the upstream 4 c02 x i S r (5)
edge (C10 and C20) and a chamfered edge (C23) on its downstream
where x and t are the position vector of an observer point and its cor-
side. To this end, we carried out LESs to examine turbulent internal
responding time, i.e., the current time of observation, respectively. y
circulation flows, shear-layer flow oscillations, and the acoustic
and τ are the position vector of a source point and its time of origina-
sources. The far-field pressure fluctuations were determined by using
tion; the source time can be obtained by subtracting the retarded time
Lighthill's acoustic analogy and Curle's equation due to advantages for
from the current time, τ = t – r / c0. Pij = pδij – τij, r = |x – y| is the
the far-field noise calculation and the wave pattern analysis
distance between a source point and an observer point, and nj is a unit
(Gloerfelt et al., 2003). This paper contains four major sections. In
directional vector on the rigid solid surface. For numerical simulations,
Section 2, our numerical simulation methods and the theoretical
Larsson et al. (2004) rewrote Curle's integral equation in a form that
background are described; Section 3 presents the simulation results and
takes the derivatives inside the integral. In the far-field region, the
the acoustic characteristics. Our summary and conclusions are provided
density fluctuations can be replaced with the pressure fluctuations, ρʹ(x,
in Section 4.
t) = pʹ(x, t) / c02. Further, the spatial derivative of an arbitrary function
f (τ) can be converted to the temporal derivative form as follows,
2. Computational methodology
f( ) f 1 r f
2.1. Numerical method = = ,
xi xi c0 xi (6)

The governing equations of LES for incompressible turbulent flows where ∂r / ∂xi is a unit directional vector from a source to an observer
are point. Therefore, Eq. (5) yields

2
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

p (x , t ) symmetric boundary condition was applied at an upper limit surface, y/


li l j 3l i l j 3l i l j D = 6. The grids in the streamwise and wall-normal directions were
1 ij ij
= T¨ij + Tij + Tij dV non-uniform, and the grid was uniform in the spanwise direction. The
4 v c02 r c0 r 2 r3
total number of grids was 631 × 265 × 221 (Nx × Ny × Nz) for the
li l j 3l i l j ij 3l i l j ij simulation domain of the baseline case. The grid spacing was non-di-
T¨ij + Tij + Tij dV (y )
c02 r c0 r 2 r3 mensionalized by the incoming shear velocity (uτ). The grid spacings in
1 p ij ij p ij ij
the streamwise (Δx+) and wall-normal (Δy+) directions were varied
+ l i nj + dS (y ) from 3.8 to 49.1 and from 1.1 to 50.2 respectively. The grid spacing in
4 c0 r r2 (7)
S
the spanwise direction (Δz+) was 20.4. In this study, a time step
The volume integral term and the surface integral term on the RHS Δt = 0.006D/U0 was employed to capture the energetic flow motion
of Eq. (7) signify the volume quadrupole and surface dipole sources inside the cavity (Park and Sung, 1995; Kim et al., 2010). The max-
respectively. In the present study, the LESs and acoustic simulations imum Courant–Friedrichs–Lewy (CFL) number and y+ were approxi-
were carried out by using the commercial software Star-CCM+ with mately 0.5 and 1.2 respectively at the downstream edge.
user-defined functions to implement the modified Curle's integral
equation. The volume integrals were performed over the entire fluid
2.4. Grid validation and design parameters
domain, and the bottom surfaces, including cavity surfaces, were ob-
tained with surface integrals. The monopole contribution for the far-
To ascertain the grid convergence, the normalized mean velocities
field location was negligible since the incoming free stream Mach
and rms turbulence intensities in the x and y directions were compared
number is approximately 0.02 (Howe, 2004).
with experimental data (ReD = 11,650) of Kang et al. (2008). The data
were time- and spanwise-averaged, and evaluated at a streamwise lo-
2.3. Computational domain and boundary conditions cation (x/D = 0.5) from y/D = 0 to 1.2. The mean streamwise velocity
profile (U̅/U0) is in overall good agreement with the experiment,
In three-dimensional simulations of flows over an open cavity, whereas the free stream velocity is overestimated (Fig. 2(a)). The ver-
periodic instabilities in the spanwise direction can be induced in ac- tical mean velocity profile (V̅ /U0) is slightly underestimated in the re-
cordance with the spanwise length of the simulation domain gion 0 < y/D < 0.8 (Fig. 2(b)). The turbulence intensity profiles in the
(Meseguer–Garrido et al., 2014). Further, Suponitsky et al. (2005) re- streamwise (urms/U0) and vertical (vrms/U0) directions are shown in
ported that domain extensions in the spanwise direction result in very Figs. 2(c) and (d). Although the experimental results are slightly un-
small changes in the amplitude of the shear-layer oscillations, but their derestimated in the region beyond y/D = 1, the differences between
frequencies are unaffected: they simulated flow at ReD = 5000 over an them and the simulated results are relatively small.
open cavity with L/D = 4 for z/D values of 2, 4, and 8. Previous studies Self-sustained oscillations inside cavities are the primary sources of
have used the following parameters: L/D = 2 with z/D = 6D at flow-induced noise (Rossiter, 1964). To reduce the noise induced by the
ReD = 3360 (Chang et al., 2006) and L/D = 1 with Lz = 4D at flow over the open cavity, several simple geometrical modifications of
ReD = 12,000 (Lee et al., 2010). In the present study, the aspect ratio of the upstream and downstream edges were tested. In this paper, we
the cavity length (L) to depth (D) was fixed at unity (L/D = 1), as consider only three of the modifications among the numerous trial
shown in Fig. 1. The domain height and width were 6D (at 0 ≤ y/ geometries we tested in this study (Fig. 3): lids at the upstream edge
D ≤ 6) and 4D (at –2 ≤ z/D ≤ 2) respectively. The inlet and outlet (C10 and C20) of the cavity and a chamfered edge (C23) on its down-
were located at x/D = ‒4 and 45 respectively. The outlet boundary was stream side. The results of the other trials are summarized in the Ap-
placed far from the downstream edge of the cavity to eliminate the pendix. In the first modification, a lid is installed at the upstream edge
influence of pressure reflections due to the outflow boundary condition. (C10 or C20). The introduction of the lid weakens flow oscillations by
The incoming boundary layer momentum thickness (θ) was D/ preventing interactions between the separated shear layer and the flow
θ = 40, and the boundary layer thickness (δ) was D/δ = 3.89. The circulating inside the open cavity (Guo and Sung, 1997; Jung and
Reynolds number based on the cavity depth (D = 25 mm) was Sung, 2016). Note that the present lid length is slightly shorter than that
ReD = 12,000. The no-slip condition was imposed at the bottom surface of Jung & Sung, which reduces the impact on the cavity. In the second
(y/D = 1) and at the cavity surfaces, and the periodic boundary con- modification, the intensity of the circulating flows is suppressed by
dition was employed at all surfaces in the spanwise direction. The installing a chamfered edge on the downstream side of the cavity (C23);

Fig. 1. Schematic diagram of the flow geometry; the enlarged views display the grid systems generated near the edges, and the front view shows the boundary
conditions.

3
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

Fig. 2. Comparison of velocity profiles at x/D = 0.5. (a) U̅/U0, (b) V̅ /U0, (c) urms/U0 and (d) vrms/U0.

the size of the chamfered edge (l2/D = h2/D) was selected according to Table 1
the size of the vortical structure in front of the downstream edge. These Design parameters of the modifications.
modifications are represented here with the notation “Cαβ”, where α upstream edge downstream edge
and β indicate upstream and downstream respectively and C denotes
the cavity. 0, 1, 2, and 3 indicate the baseline, a short lid, a long lid, and l1/D h1/D l2/D h2/D
a chamfered edge, respectively. For instance, case C00 is the baseline,
C10 0.04 0.008 - -
and case C23 is that in which there is a long lid at the upstream edge of
C20 0.08 ↑ - -
the cavity and a chamfered edge on its downstream side. The char- C23 ↑ ↑ 0.04 0.04
acteristic design parameters are listed in Table 1. Subscripts 1 and 2
denote upstream and downstream respectively.
three small circulating flows (rotating in a counterclockwise direction).
The separated shear-layer flow developed at the upstream edge im-
3. Results and discussion
pinges on the downstream edge. For C00, C10, and C20, the flow di-
rection changes along the side wall at x/D = 1. For C23, there is a
3.1. Flow characteristics
smooth protrusion in the separated flow at the downstream edge due to
the chamfered edge. The main flow circulates strongly inside the cavity,
Before proceeding further, it is important to determine the char-
but reattaches at the middle of the wall (y/D = 0 and x/D = 0). The
acteristics of the flow near the open cavity. Fig. 4 shows the mean
small circulating flows near the upstream edge are significantly
streamlines averaged over time in the spanwise direction. The results
strengthened when the lid length is increased.
for the baseline case (C00) are included for comparison. In all cases, the
Fig. 5 shows the instantaneous streamwise velocity fluctuations (u/
trends in the distributions of the mean streamline are similar, except at
U0) at z/D = 0. The separation of the shear layer due to Kelvin–-
the upstream and downstream edges. The flows inside the cavity consist
Helmholtz instability gradually develops from the middle of the cavity
of one main circulating flow (rotating in a clockwise direction), and

Fig. 3. Schematic diagrams of the modifications: (a) C10 (a short lid at the upstream edge), (b) C20 (a long lid at the upstream edge), and (c) C23 (a long lid at the
upstream edge and a chamfered edge on the downstream side).

4
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

Fig. 4. Mean streamlines around the cavity: (a) C00, (b) C10, (c) C20, and (d) C23.

(x/D, y/D) ≈ (0.5, 1). Close to the downstream edge, the layer is split coefficient, Cf = τw/(ρU0²/2), and rms wall pressure fluctuations, prms/
into two different directions of flow: one attaches to the downstream (ρU0²/2), are displayed in Fig. 6 in the streamwise wall coordinate (s).
edge and the other forms a circulating flow inside the cavity. In the As shown in the inset in Fig. 6(b), s begins from the origin (x/D, y/
presence of the chamfered edge, most of the flow does not enter the D) = (0, 0), and extends along the cavity wall. Positive values (on the
cavity, but passes over the downstream edge. Self-sustaining oscilla- red dashed line) arise in the streamwise direction, and negative values
tions arise when the circulating flow encounters the separated shear (on the blue dashed line) arise in the reverse direction. For example, for
layer. The new inflows accelerate the circulatory motions because of C00, the values s = –1, 1, and 2 correspond to (x/D, y/D) = (0, 1), (1,
the instability of the separated shear layer. C00, C10, and C20 exhibit 0), and (1, 1), respectively. Cf at the inlet is slightly higher at the up-
similar behavior at the downstream edge, whereas that of C23 is sig- stream edge (s = –1), as shown in the close-up on the left hand side of
nificantly different. The streamwise velocity fluctuations of C23 in the Fig. 6(a). Beyond this point, Cf drops sharply after separation at the
region (y/D < 1) are attenuated, especially at the bottom wall (y/ upstream edge. At every corner of the cavity (s = –1, 0, and 1) except
D = 0). for the downstream edge (s = 2), Cf converges to zero due to the
To visualize the circulating flows inside the cavity, the skin friction weakened wall shear stress. The circulating flow is weakened when the

Fig. 5. Instantaneous streamwise velocity fluctuations, u/U0, at z/D = 0: (a) C00, (b) C10, (c) C20, and (d) C23.

5
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

Fig. 6. Distributions of (a) Cf and (b) prms along the streamwise wall coordinate (s).

lid length is increased, and is weakest for C23. When the separated Helmholtz instability, and forms a large-scale vortical structure in front
shear layer impinges on the downstream edge, then there is a larger and of the downstream edge. The large-scale vortical structure collides with
sharper increase in Cf. For C00, C10, and C20, the maximum Cf is ob- the downstream edge. The patterns for C00, C10, and C20 are similar
tained at the downstream edge (s = 2), but the maxima in Cf arise at (see Cf at s = 2 in Fig. 6(a)). For C23, an elliptical vortical structure is
two different points for C23 (s = 1.96 and 2.017), i.e. at the top and formed and proceeds downstream along the edge. A more energetic
bottom chamfered edges. Fig. 6(b) shows the rms of the wall pressure flow motion is observed behind the downstream edge, but the flow (①)
fluctuations. In the region–1.5 < s < 1.71, the wall pressure fluctua- into the cavity is decreased. Thus, the circulating flows inside the cavity
tions are relatively constant. The wall pressure fluctuations at the over the path of the cavity wall (② and ③) are significantly weakened.
downstream edge (s = 2) are sharply increased by the impingement of Fig. 9 shows the vorticity strength in the z-direction (ωk) at three
the separated shear layer. For C00, an initial gradual peak at s = 1.976 different locations (y/D = 1.06, 1, and 0.94). The vorticity strength was
and a second sharp peak at s = 2 are evident. For C10 and C20, the averaged over time and in the spanwise direction. Negative and positive
intensities of the first peaks are reduced, and there is no second peak values correspond to clockwise and counterclockwise rotations re-
due to the addition of a lid. For C23, there is a sharp peak at the bottom spectively. Note that the measurement ranges in the streamwise di-
edge (s = 1.96) with a high intensity; the intensity of this peak rapidly rection (y/D = 1) are dissimilar along the vertical direction and the
decreases along the chamfered surface. chamfered edge: 0.025 < x/D < 0.975 (C00), 0.064 < x/D < 0.975
Fig. 7 displays the time histories of the pressure fluctuations, p/ (C10), 0.103 < x/D < 0.975 (C20), and 0.103 < x/D < 1.016 (C23).
(ρU0²/2), for the range–2 < z/D < 2 at (x/D, y/D) = (0.5, 1), at which At y/D = 1.06 (Fig. 9(a)), the negative vorticity strength after se-
the flow instability begins. The pressure fluctuations vary in the tem- paration decreases in the region beyond x/D = 0.5. The rise in the
poral domain, but not in the spanwise direction. The flow separation at vorticity strength is delayed as the lid length increases. The rise in the
the upstream edge is synchronized in the spanwise direction. The vorticity strength is most delayed for C23; the rises in vorticity strength
temporal pressure fluctuations have a constant periodicity, i.e. ΔtU0 occur in the order C00 ≈ C10 > C20 > C23. For y/D = 1, the negative
/δ = 4.59 (C00), 4.42 (C10), 4.33 (C20), and 4.42 (C23), which is vorticity strength slightly increases after separation for C00 and C10. In
related to the distance between the upstream and downstream edges. all cases, the vorticity strength diminishes consistently until x/D = 0.5
When the lid length is increased, the periodicity is slightly shortened, as is reached; beyond this region, it recovers smoothly before the down-
is evident for the sequence C00, C10, and C20. When we compare the stream edge is reached. For C23, the vorticity strength remains en-
results for C20 and C23, we can see that the space expansion effect of ergetic with a lower value than in the other cases, i.e., the variation in
the chamfered edge induces an increased period. The pressure fluc- the vorticity strength is least for C23. For y/D = 0.94, a positive vor-
tuations gradually weaken as the lid length increases, and the most ticity value arises near the upstream edge in all cases due to the
weakened fluctuations are observed when the lid length is as long as the counterclockwise rotating flow at the upstream corner, and then it
chamfered edge (C23). decreases smoothly until x/D = 0.275, after which the strength is
Fig. 8 shows the instantaneous vortical structures at z/D = 0. The slightly decreased by changing the sign to negative up to x/D = 0.75
contour level was nondimensionalized by the boundary layer thickness (C00 and C10) and 0.85 (C20 and C23). Beyond this region, the ne-
(δ) and the free stream velocity (U0) to give ωkδ/U0. The vortex sheet gative vorticity is sharply strengthened within a short distance. The
induced by the separated shear layer rolls up due to the Kelvin- negative vorticity due to the separated shear layer (①) is impinged by

6
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

Fig. 7. Time history pressure fluctuations for the range–2 < z/D < 2 at (x/D, y/D) = (0.5, 1).

the large circulating flow (②) inside the cavity at a specific location (③). two pressure spectra for each modification case obtained at two dif-
Meanwhile, the negative vorticity strength is reduced instantaneously ferent locations, which are averaged in the spanwise direction. The
by the circulating flow at the intersecting point, but it recovers due to reference pressure level is 2 × 10−5 Pa for the calculation of the sound
the impact of the shear layer flow. For C23, the circulation flow inside pressure level (SPL). Figs. 10–12 show the results for the pressure
the cavity is weakened, as shown in Fig. 6. The impact on the separated fluctuations over the cavity in the separated and reattached wall re-
shear layer is relatively small, which results in a delayed intersecting gions, and for the circulating flow inside the cavity, respectively. Fig. 10
point with small fluctuations. shows the frequency spectra of the pressure fluctuations. The locations
of the minimum ((x/D, y/D) = (0.5, 1)) and maximum ((0.95, 1) for
3.2. Spectral analysis C00, C10, and C20 and (0.99, 1) for C23) values of the vortical strength
were selected (see Fig. 9(b)). At these locations, sharp peaks and the
Relatively long time measurement data were employed in the fundamental frequency and its corresponding harmonic frequencies are
spectral analysis. The frequency resolution was selected as Δfδ/ evident. The fundamental frequencies at these different locations are
U0 = 0.00427, which is sufficient to distinguish the peak frequencies. nearly identical: fδ/U0 = 0.218 (C00), 0.226 (C10), 0.231 (C20), and
Our spectral analysis is presented in three figures. Each figure contains 0.226 (C23).

Fig. 8. Instantaneous vortical structures, ωkδ/U0, at z/D = 0: (a) C00, (b) C10, (c) C20, and (d) C23.

7
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

Fig. 9. Vorticity strength (ωkδ/U0) at (a) y/D = 1.06, (b) y/D = 1, and (c) y/D = 0.94.

Fig. 10. Spectra of pressure fluctuations: (a) at the midpoint of the cavity and (b) near the downstream edge.

As shown in Fig. 10(a), the sound pressure levels (SPLs) are similar respectively, as shown in the insets in Figs. 11(a) and (b). The peak
for C00, C10, and C20 until the second peak. The peak frequency of C20 frequencies coincide exactly with those of the pressure fluctuations
quickly decreases in the high frequency region. The overall SPL of C23 shown in Fig. 10. At the upstream edge (Fig. 11(a)), the results for C10
is lower than in the other cases, which exhibit similar values. The and C20 are similar, whereas C00 exhibits high SPL in the high fre-
characteristics of the spectra near the downstream edge in Fig. 10(b) quency region. The lowest SPL is evident for C23, but there are slight
are similar to those of the midpoint of the cavity, except for the de- increases in the frequency range above fδ/U0 = 1. As expected, the
coupling of the spectra for C00 and C10 after the second peak. At the sound pressure levels at the downstream edge (see Fig. 11(b)) are ap-
downstream edge (Fig. 10(b)), the overall sound pressure level is ap- proximately 20 dB larger than those at the upstream edge due to en-
proximately 10 dB larger than that of the midpoint of the cavity ergetic turbulent motions. For C00, C10, and C20, the first peak values
(Fig. 10(a)). In some frequency ranges, the sound pressure level is 20 dB are almost identical even in the low frequency range. However, the
larger than elsewhere. This result could be due to the flow instability intensities of the peaks of C10 and C20 rapidly decrease after the
resulting from impingement, i.e., it could be due to energetic turbulent second peak. Above fδ/U0 = 1, C20 follows C23, but C10 collapses with
structures near the downstream edge. the same ratio of decrease as C00. The fundamental peak of C23 is 5 dB
Fig. 11 shows the spectra of the wall pressure fluctuations at the larger than that of C00. Thus the chamfered edge on the downstream
upstream and downstream edges. The measurement points were at the side of the cavity affects the localized area because of the oscillating
separation point and the maximum prms locations (see Fig. 6(b)) separated flow (see Figs. 6(b) and 11(b)).

8
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

Fig. 11. Spectra of the wall pressure fluctuations (a) at the upstream edge and (b) near the downstream edge.

Fig. 12. Frequency spectra of the wall pressure fluctuations at (a) (x/D, y/D) = (1, 0.75), and (b) (x/D, y/D) = (0.6, 0).

The spectra of the wall pressure fluctuations inside the cavity are 3.3. Acoustic sources and far-field noise
shown in Fig. 12. The circulating flow affects all walls of the cavity,
sequentially x/D = 1, y/D = 0, and x/D = 0. Two measurement points, Fig. 13 shows the distributions of instantaneous rms of the second
located at relatively high Cf (see Fig. 6(a)), were selected: (x/D, y/ derivative of Lighthill stress tensor (Tij), as in Eq. (4); the displayed
D) = (1, 0.75) and (0.6, 0). At the first point (Fig. 12(a)), the SPL is physical quantity was nondimensionalized by multiplication with δ2/
decreased approximately 5 to 10 dB at the impinging point of the (ρ0U02). The distribution is weak in the initial instability (see Fig. 9). In
downstream edge. At the second point (Fig. 12(b)), the SPL is reduced 5 all cases, the maximum intensity arises at the downstream edge where
to 8 dB of the first point. No significant differences are evident in the the wall pressure fluctuations are dominant (see Fig. 6(b)). The max-
wall spectra results, except in the high frequency region. In all cases, imum intensity of C23 is weaker than those of the other cases. The
the SPL decreases in the high frequency region, with a similar ratio of maximum intensity is split into the top and bottom chamfered edges. As
decrease at the two locations. For C23, the SPL at the fundamental given by Eq. (5), the quadrupole noise source has the form of a volume
frequency is similar to those of the other cases, but its level is sig- integral of Lighthill's stress tensor. A decrease in the acoustic source
nificantly lower at around the first fundamental frequency. implies that the contribution of the quadrupole noise source to the far-
field pressure fluctuations is weakened.

9
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

Fig. 13. Distributions of the instantaneous rms of the second derivative of Lighthill stress tensor for (a) C00, (b) C10, (c) C20, and (d) C23.

Fig. 14 illustrates the contours of the SPL of the wall pressure using Curle's integral equation at the far-field location (X1/D, X2/D, X3/
fluctuations at each fundamental frequency: fδ/U0 = 0.218 (C00), D) = (0, 85, 0). The coordinate system is expressed as (X1, X2, X3) to
0.226 (C10), 0.231 (C20), and 0.226 (C23). The temporal pressure distinguish the obtained results from those for the fluid domain, i.e., the
fluctuations up to t = 61D/U0 at all of the central nodal points were results obtained in these coordinates are the predicted acoustic pressure
stored and analyzed for the entire surface at each specific frequency. fluctuations outside the fluid domain. The wall pressure fluctuations for
The contour level represents the SPL from 90 to 100 with 20 segments. calculating far-field acoustic noise were measured up to a time of 61D/
The locations of the maximum SPL are qualitatively in good agreement U0. Note that this time range coincides with that of Section 3.2 to
with those of the maximum wall pressure fluctuations shown in prevent frequency distortion. As for the pressure spectra in Figs. 10–12,
Fig. 6(b). For C00, the locations are widely spread around the down- fundamental and harmonic frequencies are clearly evident in the far-
stream edge. When we compare the results for C00 and C10 with C20, it field location. The harmonic frequency of C23 is rapidly attenuated in
is evident that the SPL around the downstream edge is gradually wea- the high frequency range. The SPLs obtained in the fluid domain are
kened as the lid length increases. The maximum regions of C20 and C23 similar at the fundamental frequency (see Section 3.2), but those of the
are small. In particular, the SPLs are concentrated at the bottom of the far field are significantly different: the values for C00, C10, C20, and
chamfered edge for C23. As shown by Eq. (5), the dipole noise source is C23 are 22.89, 22.31, 19.36, 16.85 dB, respectively. For C23, the vo-
closely associated with the wall pressure fluctuations (Kim et al., 2019). lumetric contribution of the Lighthill acoustic source and the area
A decrease in the maximum and its corresponding area indicates that contribution of the wall pressure fluctuations are significantly reduced
the contribution of the dipole source to the far-field acoustic pressure (see Figs. 6(b), 13, and 14). The SPL of C23 is predicted to be the
fluctuations has weakened. lowest.
Fig. 15 shows the predicted acoustic pressure spectra obtained by Fig. 16 depicts the contributions of the dipole and quadrupole

Fig. 14. Sound pressure level of the wall pressure for (a) C00, (b) C10, (c) C20, and (d) C23.

10
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

Fig. 15. Spectra of the far-field acoustic pressure fluctuations at (X1/D, X2/D, X3/D) = (0, 85, 0) for (a) C00, (b) C10, (c) C20, and (d) C23.

acoustic sources to the acoustic directivity patterns of the fundamental (C20), and 76.7% (C23) with respect to C00. The dipole source affects
frequency. The calculation points for the far-field locations were se- the far-field acoustic pressure fluctuations to a greater extent than the
lected at 2° intervals over an 80D semicircle centered at (X1/D, X2/D, quadrupole source.
X3/D) = (0, 5, 0) on the X1-X2 plane; the starting point (0°) is located at
(80, 5, 0). The main directivity patterns due to the dipole source are 4. Conclusion
oriented at approximately 135° (opposite to the flow direction), as
shown in Fig. 16(a). This result could be due to the wall pressure LESs have been performed to examine the turbulent flow and flow-
fluctuations from the downstream edge of x/D = 1. The dipole con- induced noise generation over an open cavity. The Reynolds number
tribution to the directivity patterns gradually weakens as the lid length based on the cavity depth was ReD = 12,000. The turbulent flow
increases. The power spectral densities decrease by 9.1% (C10), 62.8% around the cavity was characterized in terms of the skin friction coef-
(C20), and 78.1% (C23) with respect to C00; the average ratios of de- ficient, wall pressure fluctuations, and the vorticity strength. The
crease are 7.2% (C10), 61.2% (C20), and 77.4% (C23) with respect to acoustic sources arising from the turbulent flow were calculated by
C00. The second radiation patterns due to the dipole source are weakly using Lighthill's analogy, and the far-field acoustic pressure fluctuations
distributed between 0 and 30° along the flow direction. This distribu- were predicted with Curle's integral equation. Three different mod-
tion is generated by the wall pressure fluctuations at y/D = 1 near the ifications with two design strategies were tested to assess their effects
downstream edge. C00 and C10 have similar patterns, but the intensity on flow-induced noise. The turbulent boundary layer is split into two
of C23 is extremely low. Fig. 16(b) shows the directivity patterns due to different flow directions after impingement at the downstream edge,
the quadrupole source. The main directivity patterns are oriented at one of which forms a circulating flow inside the cavity. The flow in-
45°. As for the dipole distribution, C23 exhibits the least intense ra- stability is amplified and self-sustained by the circulating flow. For C10
diation pattern; the average ratios of decrease are 5.9% (C10), 59.8% and C20, the intensity of the circulating flow is attenuated when the lid

Fig. 16. The acoustic directivity patterns by the contributions of the (a) dipole and (b) quadrupole acoustic sources at the fundamental frequency.

11
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

length is increased. Since the lid at the upstream edge delays the flow radiation directions; the ratios of decrease of the dipole and quadrupole
separation, the energetic turbulent structures are more weakly devel- sources were found to be approximately 92.8% (C10), 38.8% (C20), and
oped in C10 and C20 than in C00. The wall pressure fluctuations are 22.6% (C23) with respect to C00. Thus flow-induced noise can be re-
also weakened, which results in the reduction of the flow-induced duced by introducing a long lid and a chamfered edge. These small
noise. For C23, massive flows pass over the chamfered edge, so the modifications were found to weaken the circulating flows inside the
circulating flow is significantly decreased. The pressure fluctuations at a cavity and to suppress the pressure fluctuations near its downstream
single point in the case of C23 are similar or slightly larger than in the edge.
other cases, whereas the volumetric contribution of the Lighthill
acoustic source and the area contributions of the wall pressure fluc- Declaration of Competing Interests
tuations are significantly reduced. At far-field locations, the predicted
frequency coincides with the frequency of the wall pressure fluctua- The authors declare that they have no known competing financial
tions. The sound pressure levels of the fundamental frequency are interests or personal relationships that could have appeared to influ-
22.89 dB (C00), 22.31 dB (C10), 19.36 dB (C20), and 16.85 dB (C23), ence the work reported in this paper.
respectively. The dipole source contributes more strongly to the far-
field acoustic pressure fluctuations than the quadrupole source. The Acknowledgements
acoustic radiations due to the dipole and quadrupole sources are mainly
oriented at 135 and 45° respectively. For C23, the power spectral This study was supported by grants from the National Research
density at the fundamental frequency is significantly decreased in all Foundation of Korea (NRF) (No. 2019M3C1B7025091).

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.ijheatfluidflow.2020.108560.

Appendix

Fig. 17 shows the contributions of the dipole and quadrupole sources to the acoustic directivity patterns at the fundamental frequency for three
additional designs (C11, C22, and C13). The results for C00 and C23 are included for comparison. As described above, C11 has a short lid/a short lid,
C22 has a long lid/a long lid, and C13 has a short lid/a chamfered edge; here, the design features are specified as upstream edge / downstream edge.
As shown in Fig. 17(a), the main radiation pattern due to the dipole source is oriented at 135°. This comparison indicates that the lids at the
downstream edge in these cases significantly increase the flow-induced noise; the power spectral densities in the main directivity angle direction of
C11, C22, and C13 are 36.4% higher than that of C10, 144.4% higher than that of C20, and 149.3% higher than that of C23, respectively. In contrast
with upstream lids, lids at the downstream edge have negative influences on the flow-induced noise with increases in their lengths. The chamfered
edge plays a major role in reducing the flow-induced noise. Fig. 17(b) shows the contributions to the directivity patterns of the quadrupole source.
For C11 and C22, the main directivity pattern is expanded over C00, with distortion particularly evident in the direction of 55° for C11. C13 exhibits
the weakest power spectral density in all angle directions of these three cases, but its efficiency is not good as that of C23. When we compare C13 and
C23, it is evident that the maximum efficiency can be obtained when a chamfered edge is combined with an upstream long lid.

Fig. 17. The acoustic directivity patterns by the contributions of the (a) dipole and (b) quadrupole acoustic sources at the fundamental frequency.

References leading-edge blowing. AIAA J 45 (5), 1132–1144.


Ask, J., Davidson, L., 2010. Flow and dipole source evaluation of a generic SUV. J. Fluids
Eng. 132 (5), 05111.
Addad, Y., Laurence, D., Talotte, C., Jacob, M.C., 2003. Large eddy simulation of a for- Chang, K.C., Constantinescu, G., Park, O.S., 2006. Analysis of the flow and mass transfer
ward–backward facing step for acoustic source identification. Int. J. Heat Fluid Flow processes for the incompressible flow past an open cavity with a laminar and a fully
24, 562–571. turbulent incoming boundary layer. J. Fluid Mech. 561, 113–145.
Ahuja, K.K., Mendosa, J., 1995. Effects of cavity dimensions, boundary layer and tem- Chatellier, L., Laumonier, J., Gervais, Y., 2004. Theoretical and experimental investiga-
perature on cavity noise with emphasis on benchmark data to validate computational tions of low Mach number turbulent cavity flows. Exp. Fluids 36, 728–740.
aeroacoustics code. NASA-CR-4653. https://ntrs.nasa.gov/search.jsp?R= Curle, N., 1955. The influence of solid boundaries upon aerodynamic sound. Proc. R. Soc.
19950018459. A 231, 505–514.
Arunajatesan, S., Kannepalli, C., Sinha, N., 2009. Suppression of cavity loads using Dowell, E.H., 1969. Nonlinear flutter of curved plates. AIAA J 7 (3), 424–431.

12
S.J. Kim, et al. International Journal of Heat and Fluid Flow 82 (2020) 108560

Ffowcs Williams, J.E., 1965. Sound radiation from turbulent boundary layers formed on Larsson, J., Davidson, L., Olsson, M., Eriksson, L., 2004. Aeroacoustic investigation of an
compliant surfaces. J. Fluid Mech. 22 (2), 347–358. open cavity at low Mach number. AIAA J 42 (12), 2462–2473.
Forestier, N., Jacquin, L., Geffroy, P., 2003. The mixing layer over a deep cavity at high- Lee, S.B., Seena, A., Sung, H.J., 2010. Self-sustained oscillations of turbulent flow in an
subsonic speed. J. Fluids Mech. 475, 101–145. open cavity. J. Aircr. 47 (3), 820–834.
Germano, M., Piomelli, U., Moin, P., Carbot, H., 1991. A dynamic subgrid-scale eddy- Lighthill, M.J., 1952. On sound generated aerodynamically, I. General theory. Proc. R.
viscosity model. Phys. Fluids 3 (7), 1760–1765. Soc. A 211, 564–587.
Gloerfelt, X., Bailly, C., Juvé, D., 2003. Direct computation of the noise radiated by a Marsden, O., Bogey, C., Bailly, C., 2012a. Investigation of flow features around shallow
subsonic cavity flow and application of integral methods. J. Sound Vib. 266, round cavities subject to subsonic grazing flow. J. Phy. Fluids 24, 125107.
119–146. Marsden, O., Bailly, C., Bogey, C., Jondeau, E., 2012b. Investigation of flow features and
Guo, Z., Sung, H.J., 1997. Analysis on the Nusselt number in pulsating flow. Int. J. Heat acoustic radiation of a round cavity. J. Sound Vib. 331, 3521–3543.
Mass Trans. 40 (10), 2486–2489. Meseguer–Garrido, F., de Vicente, J., Valero, E., Theofilis, V., 2014. On linear instability
Hao, J., Wang, M., Ji, M., Wang, K., 2013. Flow noise induced by small gaps in low-Mach- mechanisms in incompressible open cavity flow. J. Fluid Mech. 752, 219–236.
number turbulent boundary layers. Phys. Fluids 25, 110821. Park, T.S., Sung, H.J., 1995. A nonlinear low-Reynolds number k-ε model for turbulent
Heller, H.H., Bliss, D.B., 1975. The physical mechanisms of flow-induced pressure fluc- separated and reattaching flows. 1. Flow field computations. Int. J. Heat Mass Transf
tuations in cavities and concepts for their suppression. AIAA Paper 75–491. 38 (14), 2657–2666.
Howe, M.S., 1997. Edge, cavity, and aperture tones at very low mach numbers. J. Fluid Raman, G., Raghu, S., 2004. Cavity resonance suppression using miniature fluidic oscil-
Mech. 330, 61–84. lators. AIAA J 42 (12), 2608–2612.
Howe, M.S., 1998. Acoustics of Fluid–Structure Interactions. Cambridge University Press, Rockwell, D., Naudascher, E., 1979. Self-sustained oscillations of impinging free shear
pp. 1–85 Chapter 1. layer. Annu. Rev. Fluid Mech. 11, 67–94.
Howe, M.S., 2004. Mechanism of sound generation by low mach number flow over a wall Rockwell, D., Lin, J.C., Oshkai, P., Reiss, M., Pollack, M., 2003. Shallow cavity flow tone
cavity. J Sound Vib 273, 103–123. experiments: onset of locked-on states. J. Fluids Struc. 17, 381–414.
Jeon, W., Baek, S., Kim, C., 2003. Analysis of the aeroacoustic characteristics of the Rossiter, J., 1964. Wind-tunnel experiments on the flow over rectangular cavities at
centrifugal fan in a vacuum cleaner. J. Sound Vib. 268, 1025–1035. subsonic and transonic speeds. Aeronauti. Res. Counc. Rep. Memo. 3488.
Jung, S.Y., Sung, H.J., 2016. Flow structure and flow-induced noise in an axisymmetric Rowley, C.W., Colonius, T., Basu, A.J., 2002. On self-sustained oscillations in two-di-
cavity with lids. J. Mech. Sci. Tech. 30 (7), 3229–3241. mensional compressible flow over rectangular cavity. J. Fluid Mech. 455, 315–346.
Kang, W., Lee, S.B., Sung, H.J., 2008. Self-sustained oscillations of turbulent flows over an Sarno, R.L., Franke, M.E., 1994. Suppression of flow-induced pressure oscillations in
open cavity. Exp. Fluids 45, 693–702. cavities. J. Aircr. 31 (1), 91–96.
Keller, J.J., Escudier, M.P., 1983. Flow-excited resonances in covered cavities. J. Sound Sarohia, V., Massier, P.F., 1977. Control of cavity noise. J. Aircr. 14 (9), 833–837.
Vib. 86 (2), 199–226. Suponitsky, V., Avital, E., Gaster, M., 2005. On three-dimensionality and control of in-
Kim, J., Sung, H.J., 2006. Wall pressure fluctuations and flow-induced noise in a turbu- compressible cavity flow. Phys. Fluids 17, 104103.
lent boundary layer over a bump. J. Fluid Mech. 558, 79–102. Tam, C.K.W., Block, P.J.W., 1978. On the tones and pressure oscillations induced by flow
Kim, S., Huang, W.-.X., Sung, H.J., 2010. Constructive and destructive interaction modesd over rectangular cavities. J. Fluid Mech. 89 (2), 373–399.
between two tandem flexible flags in viscous flow. J. Fluid Mech. 661, 511–521. Ukeiley, L.S., Ponton, M.K., Seiner, J.M., Jansen, B., 2004. Suppression of pressure loads
Kim, S.J., Sung, H.J., 2018. Design of the solenoid valve of an antilock braking system in cavity flows. AIAA J 42 (1), 70–79.
with reduced flow noise. ASME J. Fluid Eng. 140 (3), 031105. van Leer, B., 1979. Towards the ultimate conservative difference scheme. J. Comput. Phy.
Kim, S.J., Sung, H.J., Wallin, S., Johansson, A.V., 2019. Design of the centrifugal fan of a 32, 101–136.
belt-driven starter generator with reduced flow noise. Int. J. Heat Fluid Flow 76, Wang, M., Lele, S.K., Moin, P., 1996. Sound radiation during local laminar breakdown in
72–84. a low-Mach-number boundary layer. J. Fluid Mech. 319, 197–218.
Lamp, A., Chokani, N., 1996. Active control of compressible cavity flows by using a smart Zhang, X., Chen, X.X., Rona, A., 1999. Attenuation of cavity flow oscillation through
jet. AIAA Paper 96–0446. leading edge flow control. J. Sound Vib. 221 (1), 23–47.

13

Вам также может понравиться