Wang 2020

JID: YJSVI
ARTICLE IN PRESS [m3Gsc;October 23, 2020;13:59]

Journal of Sound and Vibration xxx (xxxx) xxx
Contents lists available at ScienceDirect
Journal of Sound and Vibration

journal homepage: www.elsevier.com/locate/jsv
Investigation of acoustic propagation and source localization

in a hot jet flow
Lican Wang, Rongqian Chen∗, Yancheng You, Ruofan Qiu
School of Aerospace Engineering, Xiamen University, Xiamen 361105, China
a r t i c l e i n f o a b s t r a c t
Article history: A jet flow typically occurs in open-jet wind tunnels and turbofan engines, where there
Received 25 April 2020 are velocity and temperature gradients between the source region and the outside envi-
Revised 24 August 2020
ronment. These gradients form a thermal shear layer that determines acoustic propagation
Accepted 19 October 2020
and source localization. Moreover, they induce Kelvin–Helmholtz instabilities in numerical
Available online xxx
simulations. To stabilize simulations, a gradient-term suppression and filtering method is
MSC: proposed and validated using time-domain linearized Euler equations. For sound propaga-
00-01 tion, the influences of the steep gradients of the velocity and temperature together with
99-00 the spreading angle and source frequency are investigated through a self-similar shear
layer with superposition of compressibility and heat. Those refraction influences strongly
Keywords: change the beamforming-related phase rather than the directivity-related amplitude. Next,
Aeroacoustics
the computed database is processed by delay-and-sum beamforming to localize the source
Hot jet
Refraction
position. In the beamforming part, refraction and amplitude corrections based on Amiet’s
Beamforming theory with an extension for temperature differences are numerically evaluated and later
applied when analyzing the localization error. This localization error is basically propor-
tional to the jet Mach number, temperature ratio, and source frequency and is inversely
proportional to the spreading rate. Finally, the localization error is fitted as a linear func-
tion of the product of the jet Mach number and the Strouhal number.
© 2020 Elsevier Ltd. All rights reserved.
1. Introduction
Aerodynamic noise, already important in military and commercial aviation, is becoming increasingly restricted with the
growth of air traffic and the stringent noise regulations of the Federal Aviation Administration in the U.S. Thus, an in-
depth and accurate understanding of sound generation and propagation through complex non-uniform media is necessary
when making inverse measurements of the source characteristics or flow properties. A typical example is the propagation of
sound across a local shear flow, which is a fundamental feature of compressible engine and airframe applications [1–3] and
incompressible open-jet wind tunnels [4–10], as shown in Fig. 1. The shear layer, formed between two free streams with
different velocities and temperatures, distorts the acoustic wave in terms of convection [5], refraction [8], scattering [11],
and spectral broadening [12]. After passing through the sheared region, the original source characteristics of the amplitude
and phase will have changed when recorded by far-field detectors. Therefore, detailed investigations of the effects of a shear
layer on sound propagation and successive source localization are important for an acoustic understanding and making
reliable measurements.
∗
Corresponding author.
E-mail address: rqchen@xmu.edu.cn (R. Chen).
https://doi.org/10.1016/j.jsv.2020.115801
0022-460X/© 2020 Elsevier Ltd. All rights reserved.
Please cite this article as: L. Wang, R. Chen, Y. You et al., Investigation of acoustic propagation and source localization in a
hot jet flow, Journal of Sound and Vibration, https://doi.org/10.1016/j.jsv.2020.115801
JID: YJSVI
L. Wang, R. Chen, Y. You et al. Journal of Sound and Vibration xxx (xxxx) xxx
Fig. 1. Sketches of the shear layer in (a) an open-jet facility and (b) a turbofan engine. For the wind tunnel (a), the sound wave emitted from the jet
propagates through the incompressible and isothermal shear flow toward the microphone array located in an anechoic chamber. For the engine (b), the
noise from the fan and turbine travels across the compressible and non-isothermal shear layers formed by mixing streams from the engine exhaust, bypass
fan, and flight.
Extensive theoretical and experimental studies have evaluated the influence of the shear layer. Two theoretical treatments
have been proposed, based on the ratio of the wavelength λ and the thickness L of the shear layer [13]. The first is the
Wentzel–Kramers–Brillouin (WKB) approximation, which applies when L/λ 1. The other is the Born approximation for
L/λ 1. The WKB limit allows us to use geometric acoustics and ray tracing, which evaluates the sound trajectories by
simplifying the sheared region as a mean flow [5], a ripple interface [6,7], an infinite thin vortex sheet [8], a convex interface
[9,10], and so on [4]. Given the trajectory, the propagation time from the supposed source to the receiver can be calculated
via Fermat’s principle and then delayed for the beamforming process used in experimental source localization. One very
suitable example is Amiet’s theory [8], who assumed a thin shear layer and corrected the sound ray in two dimensions.
Although Amiet’s theorem has been wildly applied and validated [14], the neglected thickness and spreading angle of the
shear layer may cause problems if the thickness scale is close to or larger than the wavelength of the sound source. Note
that the theoretical and experimental requirements are generally not fulfilled for a non-parallel flow with steep temperature
differences and large Mach number range.
In addition to theories and experiments, numerical solvers based on the linearized Euler equations (LEE) are useful for
parametric studies and supporting multi-scale databases of convection, refraction, and scattering. However, the mean flow
gradients of temperature and velocity in shear layers separately induce Rayleigh–Taylor and Kelvin–Helmholtz instabilities
[15,16]. Because the LEE neglects non-linearity and viscosity, these instabilities become amplified by the mean flow gra-
dient. Their exponential growth, if not treated properly, produces spurious reflections and non-physical amplification [17].
Through extensive efforts [2,18–22], it has been found that the vortical wave in a modal decomposition [18] is the dominant
factor influencing the consistency of the acoustic solution outside of the shear layer. As the vortical wave is the curl of the
perturbed velocity in momentum equations, two time-domain approaches for dealing with the unstable amplification due
to the flow gradient have been proposed: gradient-term suppression (GTS) and gradient-term filtering (GTF). Basically, GTS
keeps the irrotational terms and directly removes or partly weakens those linked to the coupling of velocity perturbations
with the flow gradient, in either conservative or non-conservative forms, such as GTS-1 [23], GTS-2 and GTS-3 [2], LPCE
[24], APE [25], and so on [3,26,27]. The exponential growth of the instabilities is effectively reduced if the Strouhal number
is below 10 [28]. However, for some configurations with strong refraction, GTS may fail to provide the correct results due to
the propagation neglected in the terms removed. Instead of removing the gradient components, the treatment of the GTF to
this matter is filtering out rotational velocity perturbations and includes the irrotational parts in the acoustic transportation.
Zhang et al. [1] proposed a filtering procedure known as GTF-1 using extra Poisson equations before each iteration. GTF-2
is an enhancement of this approach in which the filtered terms have been reformulated [2]. The Poisson equation can be
replaced with different operators [29], but care should be taken with the swirling shear flow, since the potential disturbance
has a rotational part as well. Current work has combined the ideas of GTS and GTF as an alternative strategy called gradient-
term suppression and filtering (GTSF), in which GTS-1 is adopted to attain the acoustic-related velocity perturbations in the
filtering of the mean gradient terms in the LEE.
With regard to shear layer refraction, Redonnet et al. [30] investigated the effects of typical isentropic jet structures on
sound propagation at Mach number 0.17. Bennaceur et al. [12] simulated the scattering of sound by a cold and planar shear
layer at convective Mach number Mc ≈ 0.12. Casalino et al. [31]analytically and numerically studied the low-frequency sound
refraction by a thermal parallel jet and a cold mixing layer. Using GTS-1, Jiao et al. [32] extensively examined the directivity
influences of 2D shear layers with constant thickness, spreading, and curves as well as 3D planar and rectangular shear lay-
ers. The results were compared with those from Amiet’s isothermal theory for the velocity range 0–60 m/s. Moratilla-Vega
et al. [33] coupled APE with a flow solver to predict the mean flow refraction of isolated jet noise. Other enhancements in-
cluded a frequency-domain strategy [17,21]. In addition to sound wave propagation, beamforming has recently been applied
2
JID: YJSVI
to localize the source position from simulated data [34,35]. The pioneering numerical work of Padois et al. [5] using GTS-1
gave the variation of the sound location beneath a parallel shear layer at constant temperature. Using the same method and
analogous conditions, Wang et al. [36] further extended the approach to spreading for a Mach number smaller than 0.4. GTS
methods and other simplified models have been practically employed in specific applications, but they are still challenging
for extreme situations with a high Mach number and high temperature due to the elimination of the mean flow gradient.
So far, there has been little research on the physical mechanisms of sound refraction in these situations [21,37,38].
This paper develops a GTSF method for refraction by a strong and heated shear layer, and applies it to relevant appli-
cations involving sound propagation and source localization. The GTSF method is examined with two benchmark examples:
sound refraction by a (i) weak and cold or (ii) strong and hot sheared mean flow. Using this practical approach, the influ-
ence of velocity and temperature gradients together with spreading angle and source frequency on sound propagation are
investigated. A spreading shear layer is realized by a self-similar Görtler velocity profile with corrections for compressibility
and temperature. Simulated data were used in source localization via delay-and-sum beamforming (DAS). In the beamform-
ing procedure, Amiet’s refraction and amplitude corrections, also considering the temperature disparity, were examined in
detail and solved for the time delay and wave attenuation. A parametric study of source localization was conducted with
DAS beamforming and Amiet’s corrections, which quantitatively examined the dependence of the localized error on different
factors. The localized error was finally scaled with the jet Mach number and the Strouhal number.
The rest of this work is arranged as follows. The GTSF method is proposed and validated in Section 2, and the DAS
beamforming technique is introduced in Section 3. In Section 4, GTSF is applied to sound propagation to provide numerical
data for source localization, which is discussed in Section 5. The conclusions are given in Section 6.
2. GTSF method for acoustic propagation
2.1. GTSF method
Like GTF-1 [1] and GTF-2 [2], GTSF is a two-step process. In the first step, the GTS equations [2,20–23] are used to
calculate an approximately rotational-free velocity, and in the second step, this velocity is re-injected into the gradient
terms of LEE, so that refraction effects are retained whereas the rotational instabilities are filtered out. In this work, GTS-1
[23] is adopted for filtering.
Based on the assumption of an ideal gas, the conservative LEE [36,39] of mass, momentum, and energy can be written
as
∂ρ ∂ρS
+ ∇ · ρ V 0 + ρ0 V = , (1)
∂t ∂t
∂ ( ρ0 V )
+ ∇ · ρ0 V0 V + ∇ p + ρ0 V ·∇ V0 + ρ V0 ·∇ V0 = 0, (2)
∂t
∂ p ∂ pS
+ ∇ · p V0 + γ p0 V + (γ − 1 ) p ∇ ·V0 − V · ∇ p0 = , (3)
∂t ∂t
where ∂ ρS /∂ t and ∂ pS /∂ t are the source terms. ρ ,V = ui + vj + wk, and p are the density, velocity vector, and pressure,
respectively, with the time-averaged part subscripted with 0 and the perturbed wave amplitude superscripted with . The
specific heat ratio γ = 1.4.
The terms for the mean flow gradient in the LEE are removed by GTS-1 [23]. Thus, the momentum Eq. (2) and the
energy Eq. (3) can be simplified as:
∂ ρ0 Vf
+ ∇ · ρ0 V0 Vf + ∇ p = 0, (4)
∂t
∂ p ∂ pS
+ ∇ · p V0 + γ p0 Vf = , (5)
∂t ∂t
where Vf denotes the irrotational velocity perturbations. Here, the governing equations of GTS-1 are Eqs. (1), (4), and (5).
Note that the vortical wave ωf = ∇ × Vf in the curl of Eq. (4) is not entirely removed when ρ0 varies in space. Specifically,
the continuity Eq. (1) is equivalent to the energy Eq. (5) under the isentropic assumption.
Substituting the velocity perturbation of GTS-1 into the mean flow gradient terms of LEE gives
∂ρ ∂ρS
+ ∇ · ρ V 0 + ρ0 V = ,
∂t ∂t
∂ ( ρ0 V )
+ ∇ · ρ0 V0 V + ∇ p + ρ0 Vf ·∇ V0 + ρ V0 ·∇ V0 = 0,
∂t
∂ p ∂ pS
+ ∇ · p V0 + γ p0 V + (γ − 1 ) p ∇ ·V0 − Vf · ∇ p0 = , (6)
∂t ∂t
3
JID: YJSVI
where the presence and absence of subscript fdenote the perturbations from GTS-1 and LEE, respectively. Combining the
momentum and energy equations in Eqs. (4)–(6), we get a second-order form of the GTSF:
⎡ ⎤
gradient terms of momentum equation
⎥
∂ 2 p ∂ p γ p0 ⎢
⎢∇ · ρ0 V0 V + ∇ p + ρ0 V f · ∇ V0 + ρ (V0 · ∇ )V0 ⎥
+ V · ∇ − ∇ ·
∂t2 ∂ t
0
ρ0 ⎣ ⎦
O (M ) O (M )
2 O ( )
M 3 O ( )
M 4 O ( M 5
)
gradient terms of energy equation
⎡ ⎤
(γ − 1 ) ⎢ ⎥ 2
⎢∇ · ρ0 V0 V f + ∇ p ⎥ · ∇ p0 = ∂ pS ,
+ (7)
ρ0 ⎣ ⎦
∂t2
O (M )
4 O ( M 3
) O (M )
with a Mach number dependency [13,40] scaled by a parallel shear flow (∇ · V0 = 0)
ρ0 ∼ ρ∞ , u0 ∼ u∞ , p0 ∼ ρ∞ u2∞ , t ∼ l/c∞ ,
u ∼ Mu(1 ) , ( p , pS ) ∼ ρ∞ c∞ u , (ρ , ρS ) ∼ ρ∞ /c∞ u ,
where M is the Mach number, free-stream variables for the shear layer are denoted with subscript ∞, and the velocity is
expanded as u = u0 + Mu(1 ) + · · · .
3 u (1 ) /l 2 · O Mn . Note that the gradient terms associated with the vortical mode
In Eq. (7), each term is in the form ρ∞ c∞ ( )
in Eq. (7) have a Mach number dependency of order n ≥ 3, and the first term and the source term have the same leading
order O(M ). In the analysis of Seo et al. [40], the high-order gradient terms related to refraction are removed to give the
governing equations for low Mach number flow, whereas those terms are retained by GTSF because the refraction effects due
to the velocity, temperature, and pressure gradients become progressively more significant at higher Mach number. With the
assumption of homogeneous mean pressure, the gradient terms of the momentum equation are mainly responsible for the
perturbed vorticity. This is where GTS and GTF employ their suppression and filtering operations. If there is a uniform mean
flow, the gradient terms are zero and GTS-1 and GTSF are equivalent. In GTSF, GTS-1 operates independently in parallel. It
transfer information about filtered perturbations to Eq. (6) for the same time point. Although more resources are required
for parallel computing, GTSF is coefficient independent and practical for resolving sound refraction due to velocity and
temperature gradients. Moreover, many of the improvements [21] for LEE are naturally feasible for GTSF.
2.2. Numerical methods and validation
The time domain computational aeroacoustic (CAA) in-house code of XTER lab was established based on the framework
developed by Chen et al. [36,41], which has been fully verified and compared with recognized benchmarks. For the spatial
resolution of GTSF, a dispersion-relation-preserving scheme [42] with a 7-point 4th-order stencil was used and combined
with an 11-point 10th-order filter to eliminate high-frequency errors that might be induced by the non-uniform grids. An
optimized 5-step low-dissipation and low-dispersion explicit Runge–Kutta scheme [43] was applied for the time advance-
ment to resolve the interactions between the sound and the shear layer. We employed well-refined grids in the physical
domain of the numerical simulation to ensure we have a high-fidelity representation of the fluid structure and acoustic na-
ture. We utilized stretched grids in the boundary domain to reduce spurious reflections. For the central-difference scheme
of the filtering, in the stretched zone, six ghost points with a cell-to-cell ratio of 2.0 grow exponentially to build a damping
layer. With the damping layer and spatial filter, the boundary is nearly non-reflective. In all simulations, there were at least
10 points per wavelength, which has considered the wavelength variations due to jet convection.
We first validate GTSF by considering the propagation of sound through a cold shear layer [25] for a vorticity of thickness
δw = 50 and Mach number M jet = 0.5. The shear layer is weak so allows a stable solution of the LEE with a computational
domain of (−100 < x, y < 100 ) and equidistant grid spacing of 0.4. The steady distribution of the mean flow and the source
terms are
⎧
⎨ρ0 = 1,
⎪
u0 = M jet tan (2y/δw ),
(8)
⎪v0 = 0,
⎩
p0 = 1/γ ,
and
∂ ρ
−ln(2 )/9· (x2 +y2 )
∂t = e
S
cos (0.5t ),
(9)
∂ pS −ln(2 )/9· (x2 +y2 )
∂ t = 2e cos (0.5t ).
4
JID: YJSVI
Table 1
Error comparison for cold shear layer refraction.

ρrms urms vrms prms
GTS-1 14.63% 16.43% 13.97% 14.63%

GTSF 1.28% 1.51% 1.19% 1.28%
Fig. 2. Perturbed pressure distributions extracted from (a) y = 15 and (b) y = 50 at the 15th source period.
To evaluate the accuracy quantitatively, the L2 numerical error is expressed as

N
nodes 2

i=1 ( ) cal ( ) re f
• − •

e L2 = × 100%, (10)
N
nodes
2
(• ) re f

i=1
where (• ) refers to the relevant density, velocity, or pressure perturbations, and the subscripts cal and re f denote calcu-
lation and reference results, respectively. The percentage eL2 represents the accuracy of each solver. The root-mean-squared
(rms) value of the LEE is used for reference in Table 1. It can be seen that the rms errors of GTSF are much smaller than
those of GTS-1.
The second benchmark was designed as an aeroacoustic issue in the 4th CAA Workshop [19]. It has a strong and heated
shear layer with M jet = 0.756 and T jet /Tair = 2.0. Fig. 2compares numerical snapshots for GTSF and GTS-1 along y = 15 and
50 at the 15th source period with the analytical solutions [20]. For GTS-1, the pressure distribution downstream of the
sound source is deeply affected by the elimination of the mean flow gradient, whereas GTSF clearly provides an acceptable
prediction, both near to and far from the jet. Moreover, the solution of GTSF is not influenced by the numerical Kelvin–
Helmholtz instability in comparison with that of LEE, as documented in [2].
The above results briefly demonstrate the accuracy of GTSF in acoustic resolution and its ability to suppress numerical
Kelvin–Helmholtz instabilities.
5
JID: YJSVI
Fig. 3. Sketch map of sound propagation through a jet shear layer.
3. DAS beamforming method for source localization
Beamforming is an efficient tool for imaging aeroacoustic sources based on far-field signals recorded by a microphone
array in experiments or by observers in a numerical simulation. Despite the better capability of DAMAS [44] and CLEANSC
[45] at low frequencies, time-domain DAS beamforming [39] is good enough for the current work. A DAS beamformer steers
all the microphones toward one particular point in a user-defined scanning line, where the sound source is supposed to be
located. Then, the time delay tk , weight coefficient ωk , amplitude correction pA /pO , and steering vector χ are applied to
the signals pk (t ) from each microphone. The resulting signals are summed to obtain the signals pj (t ) at j-th scanning point.
These processes are shown as follow:

K
( pA /pO )k χK/2
pj (t ) = ωk p (t − tk ), (11)
( pA /pO )K/2 χk k
k=1
where pj and pk are the pressures at the j-th scanning point ( j = 1, . . . , J ) and the k-th microphone (k = 1, . . . , K ), respec-
tively. ωk = 1/K is the weight coefficient. tk = tk − tK/2 is the retarded time (compared to the central microphone) between
when a sound ray is emitted from the j-th scanning point and reaches the k-th microphone after being convected by the
jet flow and refracted by the shear layer. pA /pO is the ratio of the corrected pressure pA in a uniformly moving flow and the
measured pressure pO after refraction by the shear layer. Following [44,46], the steering vector χK/2 /χk is used to normalize
the amplitude attenuation of the wave at the k-th microphone compared to an observer at the center of the array. The
maximum pj (t ) in Eq. (11) is the magnitude of the j-th scanning point. After the DAS beamformer processes all the points
on the scanning line, the position of the scanning point with the largest magnitude represents the localized sound source.
Since it is difficult to fully consider the impact of actual environment on the time delay and amplitude correction, etc., the
localized source location may shift from the numerically defined position. This shift is also called source drift or localization
error.
3.1. Refraction correction for a hot jet
The propagation time t in Eq. (11) is determined by the refraction path, which is corrected through the theory of Amiet
[8,47] with an extension for the temperature gradient. It assumes that there is an infinitely thin shear layer flanked by
homogeneous velocity, density, and pressure. Fig. 3 is a sketch of the sound transmitted through a hot jet (u jet , v jet ) to cold
air (uair , vair ). Subscripts S,B, and O denote parameters for the source, interface point, and observer. In the absence of a
shear layer, the emission of a sound ray from source S(xS , yS , pS ) toward B(xB , yB , pB ) linearly reaches observer A(xA , yA , pA )
without deflection. With a shear layer, the sound from the source is deflected from A to O(xO , yO , pO ) with yA = yO . This
refraction follows Snell’s law, which is based on the principle of velocity continuity across the interface:
c jet cair
u jet + = uair + , (12)
cos θS cos θO
where θ denotes the sound emission angle in a stagnant flow. According to wave convection, the emission angle θS in a jet
flow is
sin θS
θS = tan−1 . (13)
cos θS + u jet /c jet
We restrict our attention to the configuration with one jet flow and one medium at rest, where uair = 0, sound speed
ratio μ = c jet /cair , and jet Mach number M jet = u jet /cair . Reformulating Eqs. (12) and (13), the relation between observers
A and O is
μζ
tan θA = tan θS = , (14)
(μ2 − M2jet ) cos θO + M jet
6
JID: YJSVI
with

ζ= (1 − M jet cos θO )2 − μ2 cos2 θO . (15)
For θO = 0 or π in Eq. (14), the critical emission angle for total internal reflection is
⎧ √
⎪
⎨tan−1 μ μ2(−M
1−M jet )2 −μ2
2 +M , θO = 0 ,
θS =
jet
jet
√ (16)
⎪
⎩tan−1 μ (1+M jet )2 −μ2
, θO = π .
M2jet −μ2 +M jet
For θS = 0 or π in Eq. (14), the critical refraction angle for the zone of silence, where no sound ray can directly reach
through refraction, is

cos−1 1
M jet +μ
, θS = 0,
θO = (17)
cos−1 1
M jet −μ
, θS = π .
From the above, we can see that the limiting values of the emission angle and refraction angle depend only on M jet
and μ. For a subsonic isothermal flow with μ = 1 [8,32], the total internal reflection in Eq. (16) and the silent zone in
Eq. (17) are in the upstream and downstream directions, respectively. Taking into account temperature, the total reflection
occurs with μ < 1 − M jet downstream and μ < 1 + M jet upstream, whereas the zone of silence occurs with μ > 1 − M jet
downstream and μ > 1 + M jet upstream. Total reflection is not included in Amiet’s refraction prediction and the zone of
silence needs to be avoided during programming, so that ζ in Eq. 15 is a real number and the refraction path can be
obtained by the geometric relations of the observers at S,B and O. It has been theoretically shown that Snell’s law, Eq. (14),
which is used for the refraction, also satisfies the minimum travel time of Fermat’s principle [10,14]. Thus in two dimensions,
the subsonic propagation time is
t = tSB + tBO , (18)

with

M jet (xS − xB ) (xS − xB )2 + (yS − yB )2 M jet (xS − xB )2
2
1
tSB = + + , (19)
cair μ2 − M2jet μ2 − M2jet (μ2 − M2jet )2
and

( xS − xB )2 + ( yS − yB )2
tBO = , (20)
cair
where tSB and tBO represent the transmission time inside and outside the jet, respectively.
There are several strategies for solving the least propagation time in Eq. (18). The first way calculates the refraction path
with Eq. (14). Instead of using an iterative solution, spatial interpolation and time reversal [10,48] are adopted to reduce the
computational cost. An amount of rays are emitted simultaneously from the observer O and back-radiated to the scanning
region, and the propagation time t for each scanning point is then linearly interpolated. Another method, known as one
receiver and one source, directly finds the smallest value of t in Eq. (18) from a matrix of points xB at the shear interface.
This approach requires more computer time but is easier to code [5].
3.2. Amplitude correction for a hot jet
The amplitude correction pA /pO in Eq. (11) represents the transmission loss and magnitude variation induced by the
shear layer. The ratio of the perturbed pressures for observers A and O is
pA pB+ pB− pA

= · · , (21)
pO pO pB+ pB−
where pB+ /pO is calculated based on the conservation of acoustical energy within a 2D ray tube [8,47]. Based on the
Blokhintzev invariant [4], the pressure ratio is calculated by Eq. (14) in conjunction with the geometry of a ray path:

pB+ h2 ζ3
= 1+ −1 , (22)
pO h1 μ sin3 θO
where h1 = yB − yS and h2 = yO − yS .
pB− /pB+ is the transmission ratio of the perturbed pressures before ( pB− ) and after ( pB+ ) the ray passes through the
shear layer. Following Ribner [7] and Miles [6], the shear layer is assumed to be a rippled surface for which the pressure
7
JID: YJSVI
and displacement are matched on both sides. The planar pressure wave is linked to the velocity potential [32] and derived
as
pB− 1
= ζ + μ sin θO (1 − M jet cos θO )2 . (23)
pB+ 2μζ
The decay of pA /pB− is equal to h1 /h2 in 3D, whereas in 2D, it follows the Green’s function [49–51] for a monopole
source radiating in a moving flow:

i 1 x2 + (1 − M2 )y2 −i(M/1−M2 )kx−iωt
G(x, y, t ) = √ H (1 ) k e , (24)
4 c2 1 − M 2 0 1 − M2
jet
where k = ω/c jet and M = M jet /μ. In reality, the sound source is a singularity for the Green’s function. This makes it some-
what difficult to derive the amplitude attenuation from the near field to the far field. Therefore, in this work, we focus
on the far field, which is where a microphone array is generally positioned. Thus, the Hankel function for the far field
[52]asymptotically behaves as
1
!− 12
(1 )
H0 (X ) ≈ πX e±i(X −π /4−1/8X ) . (25)
2
By combining Eqs. (24) and (25), the decay of a sound wave in a convecting medium is related to
1
χ ≈ √ , (26)
R

with R = (x − xS )2 + (1 − M2 )(y − yS )2. Therefore,
pA χA
=
pB− χB−
is obtained and the product of
( pA /pO )k χK/2
· and pk (t − tk )
( pA /pO )K/2 χk
in Eq. (11) is equivalent to
pOK/2
pk (t − tk ),
pO
k
which represents a normalization of the measured pressure from the k-th microphone to the central one with consideration
of shear layer induced amplitude variation.
4. Sound propagation in a hot jet flow
Noise caused by automobiles, aircraft, helicopters, wind turbines, fans, and so on is annoying at low Mach number, which
is generally investigated in an open-jet aeroacoustic wind tunnel with an incompressible and cold shear layer. In a jet engine
of a modern subsonic commercial aircraft, compressibility and thermal effects dominate the development of a shear layer
at a relatively high Mach number. To give insights into the refraction by the shear layer, GTSF is used throughout this paper
to consider the effects of a gradient region of the velocity and temperature. The sheared region is idealized with a moving
fluid and a fluid at rest. Previous research into shear flow fields was realized with computational fluid dynamics [21,32] or
incompressible theory analysis, such as the tangent relation of Candel et al. [5,37,53] and the velocity profiles of Görtler
[54]. In Appendix A, we extend Görtler’s model with a compressible and thermal correction as well as the Crocco–Busemann
relation [55] for the variations of the Mach number and temperature.
4.1. Influence of velocity and temperature gradients
A thin but non-zero-thickness shear layer is idealized to evaluate the sound refraction that depends on velocity and
temperature gradients. The flow field is obtained from Appendix A with the spreading rate σ0 = 106 (a recognized constant
for characterizing the shear layer), whereas the velocity/shear boundary is defined as 5% and 95% of the jet velocity. Three
subsonic jet speeds u∗jet = 69.44, 138.88, and 277.75 m/s and three jet temperatures T jet ∗ = 30 0, 450, and 60 0 K were tested.
Superscript ∗ represents a dimensional quantity. The governing equations of the GTSF are made dimensionless using far-field
characteristics, including sound speed cair ∗ = 347.2 m/s, length scale L∗ = 1 m, density ρ ∗ = 1.2 kg m−3 , temperature T ∗ =
air air air
300 K, and pressure ρair ∗ c∗2 . We use two forms of the Mach number. The local convective Mach number M is normalized
air
with the jet velocity c∗jet . The jet Mach number M jet is normalized with cair
∗ . In other words, M
jet = 0.2, 0.4, and 0.8 is relative
8
JID: YJSVI
Fig. 4. Perturbed pressure fields of sound propagation through a thin shear layer at time t = 3 for (a) a low-speed isothermal jet (M jet = 0.2 and T jet /Tair =
1.0), (b) a high-speed subsonic cold jet (M jet = 0.8 and T jet /Tair = 1.0), and (c) a high-speed subsonic hot jet (M jet = 0.8 and T jet /Tair = 2.0). (d), (e), and
(f) The corresponding rms distributions. The solid lines denote the positions of the shear layers. The white dotted lines denote the sound rays along the
critical emission angle. The solid white dots are the interaction points between the ray path and the thin shear layer.
to the sound speed in the far field. The non-dimensional computational domain ranges from (−0.7, −0.5 ) to (0.7, 0.9 ) with
an equal spacing of x = y = 0.0028. The shear layer originates from (−0.7, 0 ) and is centered at y = 0.
The monopole source terms are
∂ρS ∂ pS
= 0, = f (x, y ) sin(ωt ), (27)
∂t ∂t
and
" #
ln(2 )
f (x, y, z ) = ε exp − ( x − xS )2 + ( y − yS )2 , (28)
(3) 2
with angular frequency ω = 92.4, frequency f = 14.7, source location (xS , yS ) = (0, −0.3 ), amplitude ε = 0.01, and =
0.005. In this section, the thickness between two basically parallel velocity boundaries (within one mesh cell) is L = 0.00252
and the wave number of the sound source is k = ω/cair = 92.4. Thus, we get kL = 0.233 < 1 [56], so that the shear layer can
be treated as thin enough that we can ignore its spreading angle and mean thickness compared to a spreading layer with
L ≈ O ( 0.1 ).
Fig. 4 shows examples of the instantaneous sound fields and the rms pressure for (a) M jet = 0.2 and T jet /Tair = 1.0, (b)
M jet = 0.8 and T jet /Tair = 1.0, and (c) M jet = 0.8 and T jet /Tair = 2.0. As shown in Fig. 4(a), GTSF produces a stable solution
near the thin velocity/shear interface. Though the gradient is steep, the acoustic convection, refraction, and reflection are re-
solved well. Within the jet, the sound wave harmonically radiates upstream and downstream at speeds of 1 ∓ M jet , respec-
tively. Due to the convection, the acoustic wavelength upstream of the source decreases and conversely for the downstream
direction. When the incident wavefront of the sound encounters the sheared border between the jet and the outside air,
because of the mean flow gradient, some waves are refracted to the ambient air whereas the rest are reflected back to the
jet. The reflected waves interact with themselves and the direct incident waves from the source according to the ray tracing
results of [32,37]. Thus, there are patterns induced by constructive and destructive interference below the shear layer in
Fig. 4(a). Comparing Fig. 4(b) and Fig. 4(a), increasing the jet velocity from M jet = 0.2 to 0.8 enhances the flow convection,
which decreases the upstream wavelength, accelerates its attenuation, and subsequently weakens the acoustic energy in the
reflected region. With the change in the temperature ratio from 1.0 to 2.0, the sound field in Fig. 4(c) is significantly differ-
9
JID: YJSVI
Fig. 5. Spreading angle θsp versus spreading rate σ0 = [9.0, 11.0, 13.0] for different temperature ratios T jet /Tair = [1.0, 1.5, 2.0].
ent from that in Fig. 4(b). For a fixed jet velocity, the superposition of the higher temperature increases the sound speed in
the jet flow of Fig. 4(c) and indirectly increases the wavelength and reduces the local convective Mach number. The change
in the wavelength has little influence on the sound traversing the thin interface, but the reduction in convection accordingly
decreases the flow gradients and the related refraction angle, which is consistent with the results for GTS-1 in the literature
[37]. Finally, the interference patterns are clearer in the reflected region.
The critical emission angles θS for total internal reflection were calculated with Eq. (16) and are shown in Figs. 4(a)–4(c)
as 138.9◦ , 73.6◦ , and 109.6◦ , respectively. The white dotted lines denote the sound rays along the critical emission angle and
the solid white dots are the interaction points between the ray paths and the thin shear layers, which are indicated by the
black solid lines. If there is total internal reflection, the acoustic energy reflected in the upstream direction together with
the resultant interference between the different waves become stronger. There are interference patterns near to the left of
the intersection point, which is consistent with Eq. (16). Meanwhile, since more energy is reflected into jet, a zone of weak
energy is visible in the upstream refracted region. Comparing Figs. 4(a)–4(c), it can be seen that the critical emission angles
decrease with the increase in jet speed and increase with the increase in temperature.
Unlike Figs. 4(a)–4(c), which have both amplitude and phase information, the rms features in Figs. 4(d)–4(f) show only
the magnitude, as the acoustic phase has been filtered out. It can be seen from Fig. 4(d) that the sound distribution for
M jet = 0.2 is relatively uniform. There are several weak interference patterns above the source and beneath the shear layer.
For a higher Mach number, as shown in Fig. 4(e), a Mach cone preliminarily formed in the nearly transonic jet flow, which
was attenuated due to the higher temperature of Fig. 4(f). Note that there are reflections in Fig. 4(e). They are more com-
plicated but weaker than those in Fig. 4(f) [37].
4.2. Influence of spreading angle
Different from the thin layer considered in Section 4.1 and the parallel case in [5], actual shear layers are thick and can
expand along the flow direction. This expansion results in a spreading angle, which is defined in Eq. (A.9) as the angle of
the velocity boundaries with 5% and 95% of the jet speed. Compared to a parallel shear flow, the spreading changes the dis-
tributions of the thickness and velocity, which can decrease the upstream reflections and increase downstream refractions.
Interested readers can refer to our previous investigation [36]. Experimentally, the spreading angles of jets injected by dif-
ferent nozzles differ from each other due to the shape, material, and configuration of the nozzle. Therefore, we investigate
the impact of spreading angle on the amplitude and phase of far-field signals.
Before discussing the dependence of spreading angle on acoustic propagation, it is more appropriate to analyze the influ-
ence of compressibility and temperature corrections on the flow field. From Eq. (A.5) in the appendix, the compressibility
correction has factor:

1
= 1 + 0.5468 −1
1 + 35.53Mc8.614
where Eq. (A.6) gives that Mc = M jet /(μ + 1 ). For M jet = 0.8 and μ = 1, we get Mc = 0.4 and = 0.993. The latter is close
to 1, which means the compressibility has a small impact. Certainly, note that the compressibility will cause a significantly
reduced growth rate in a supersonic environment [57,58]. With regard to thermal correction, Fig. 5shows the variation of
spreading angle θsp with spreading rate σ0 and temperature ratio T jet /Tair . The spreading rate is commonly used to charac-
terize the growth rate of a shear layer. For fixed temperature, the spreading angle decreases as the spreading rate increases:
σ0 = [9.0, 11.0, 13.0]. For an increase in the jet temperature, the spreading angle increases over its whole range.
Fig. 6 shows the instantaneous and rms pressure distributions for a spreading flow with a fixed frequency f = 14.7, and
(a) M jet = 0.2 and T jet /Tair = 1.0, (b) M jet = 0.8 and T jet /Tair = 1.0, and (c) M jet = 0.8 and T jet /Tair = 2.0. The propagation
features of the sound waves in Figs. 6(a)–6(c) are similar to those in Figs. 4(a)–4(c), as the sound undergoes convection,
refraction, and reflection. Comparing Fig. 6(b) to Fig. 6(a), we can see that the acoustic power becomes much weaker in the
upstream reflected and refracted region when the jet Mach number increases from M jet = 0.2 to 0.8. Superposing a high
10
JID: YJSVI
Fig. 6. Perturbed pressure fields of sound propagation through a spreading shear layer at time t = 3 for (a) a low-speed isothermal jet (M jet = 0.2 and
T jet /Tair = 1.0), (b) a high-speed subsonic cold jet (M jet = 0.8 and T jet /Tair = 1.0), and (c) a high-speed subsonic hot jet (M jet = 0.8 and T jet /Tair = 2.0). (d),
(e), and (f) The corresponding rms distributions. The velocity boundary of the shear layer is indicated with two solid lines.
temperature, the refraction is lower as the sound wavelength is higher and the mean flow gradient is lower, as already
noted. The refraction is further enhanced as the spreading angle increases from θsp = 12.1◦ in Fig. 6(b) to θsp = 14.8◦ in
Fig. 6(c). Numerically, the Rayleigh–Taylor instabilities at high temperatures are more evident, which may make the LEE
unconditionally unstable [59] and also distorts the wavefront simulated by GTSF. Comparing the area between x = 0.35 and
x = 0.70 below the shear layer in Figs. 4(c) and 6(c), the wavefront is relatively smoother in a spreading structure than in a
thin configuration. This is consistent with [2,59], that is, a spreading shear layer tends to partly stabilize a numerical sim-
ulation for either the LEE or GTSF. The rms features of the pressure distributions in Figs. 6(d)–6(f) are also similar to those
of Figs. 4(d)–4(f). The sound magnitude is large near the source and gradually attenuates outward. There are interference
patterns in the reflected region.
To illustrate the effects of the spreading angle, Fig. 7 shows the differences between the fields of Fig. 6 and those of
Fig. 4 under the same conditions. From the relative pressures in Figs. 7(a)–7(c), it can be observed that the impact of
spreading angle enhances with increasing Mach number and temperature in a jet flow. The impacted area covers the up-
stream reflected region and most of the refracted region. From the relative rms pressures in Figs. 7(d)–7(f), it can be seen
that the differences between shear layers with or without spreading occur mainly in the reflected region and only slightly
in the refracted region. The rms value denotes only the magnitude of the sound wave, whereas the instantaneous value in-
cludes the amplitude and phase. The effects of refraction on the phase patterns, perhaps, outweigh the amplitude directivity
for the spreading configuration [30].
In Fig. 8, we quantitatively compare the relative errors of amplitude and phase calculated via Eq. (10) with those for
spreading rate σ0 = 11.0. The acoustic amplitude and phase for different spreading rates were found for jet Mach number
M jet = [0.2, 0.4, 0.8] and temperature ratio T jet /Tair = [1.0, 1.5, 2.0] using 501 sampling points from x = −0.7 to x = 0.7 with
y fixed at 0.6. On the whole, there is clearly a low level of errors in Fig. 8(a) for the different configurations. The maximum
amplitude errors were 0.023%,0.013%, and 0.31% for σ0 = [9.0, 13.0, 106 ], respectively. The magnitudes for σ0 = 9.0 and 13.0
are almost the same as those for σ0 = 11.0 in comparison to the thin configuration with σ0 = 106 . Regarding the phase er-
ror in Fig. 8(b), the influence of velocity, temperature, and spreading angle was strongly dissimilar to that of the amplitude
changes, since the behavior of an acoustic phase is more sensitive to variations in the flow field. The relative errors shown
in Fig. 8(b) are more than two orders of magnitude higher than those in Fig. 8(a). This comparison quantitatively demon-
strates that the spreading angle affects the refractive acoustic characteristics in terms of phase rather than amplitude [2,60].
Moreover, the results for a thin shear layer deviate from the other cases in both amplitude and phase. We recommend re-
11
JID: YJSVI
Fig. 7. Relative pressure fields of sound propagation through a thin shear layer (σ0 = 106 ) with respect to the spreading configuration (σ0 = 11.0 ) at
time t = 3 for (a) a low-speed isothermal jet (M jet = 0.2 and T jet /Tair = 1.0), (b) a high-speed subsonic cold jet (M jet = 0.8 and T jet /Tair = 1.0), and (c) a
high-speed subsonic hot jet (M jet = 0.8 and T jet /Tair = 2.0). (d), (e), and (f) Corresponding relative rms distributions.
placing a very thin shear layer with a gradually spreading one (such as σ0 = 11.0), as this is more realistic and would help
to suppress numerical instabilities [2,59].
4.3. Influence of source frequency
To check the effects of frequency on sound transmission, two source frequencies, f = [8.82, 19.8], for the same conditions
used for Fig. 8 were computed for a shear layer with (σ0 = 11.0) or without (σ0 = 106 ) spreading. The relative errors of am-
plitude and phase are shown in Fig. 9. The results for f = 14.7, as shown in Fig. 8, are redrawn for comparison. Once again,
the amplitude error between sound transmission through a thin shear layer and a spreading one is small at the different
frequencies. The maximum percentages are 0.21%,0.31%, and 0.98% for f = [8.82, 14.7, 19.8], respectively. The phase error
outweighs the amplitude error, with maximum values of 52.1%,60.1%, and 73.9% for f = [8.82, 14.7, 19.8]. For an infinitely
thin shear layer, the acoustic refraction, based on Amiet’s theory [5,8–10], is independent of source frequency. However,
a frequency dependence can be seen in Fig. 9. There are refraction differences between the approximately zero-thickness
shear flow and the thicker spreading shear layer.
5. Source localization in a hot jet flow
The steady shear layer distorts the acoustic characteristics in terms of amplitude and phase. The variation of the am-
plitude due to shear layer refraction has been demonstrated to be minor, whereas the phase change is of fundamental
importance to source localization, as the time delay from source to receiver depends on the phase. In source localization
via DAS beamforming, the time delay caused by the shear layer is generally treated using Amiet’s theory based on an in-
finitely thin assumption. Thus, some questions are addressed in the following: (i) What are the differences between Amiet’s
semi-analytical method and a numerical solution based on a thin shear layer? (ii) How do different factors influence the
localization accuracy? (iii) Can these influences be modeled empirically?
5.1. Evaluation of Amiet’s correction method
Amiet’s theory is a widely used correction technique for sound refraction in an incompressible open-jet wind tunnel. It
is based on the assumption of an isothermal and infinitely thin shear layer. Sometimes it is employed as a simple approach
12
JID: YJSVI
Fig. 8. Relative errors of (a) amplitude and (b) phase for different spreading rates, σ0 = [9.0, 13.0, 106 ], in comparison to σ0 = 11.0. The temperature ratio
varies from T jet /Tair = 1.0 to 2.0 and the Mach number M jet = [0.2, 0.4, 0.8]. The virtual array is at x = −0.7 to 0.7 and y = 0.6 and has 501 evenly distributed
points.
Fig. 9. Relative errors of (a) amplitude and (b) phase between thin shear layer σ0 = 106 and spreading shear layer σ0 = 11.0 for different frequencies,
f = [8.82, 14.7, 19.8]. The temperature ratio ranges from T jet /Tair = 1.0 to 2.0 and the Mach number M jet = [0.2, 0.4, 0.8]. The virtual array is at x = −0.7 to
0.7 and y = 0.6 and has 501 evenly distributed points.
to correct mean flow refraction outside the control surface of Kirchhoff and Ffowcs-Williams and Hawkings methods [61],
due to its universality and efficiency for general source types. In this section, we numerically evaluate two parameters
used in a non-isothermal Amiet’s correction: the refraction angle θO of Eq. (14) and the pressure ratio pA /pO of Eq. (21).
The refraction angle can be calculated because the radiation direction is perpendicular to the wavefront and there is no
scattering above the steady shear layer. A wavefront for p = 0 in the far field is extracted at t = 3 and fitted with a fourth-
order polynomial in MATLAB. The refraction angle can be interpolated from the results obtained by the local normal vector
of the polynomial. Following Jiao [32], the pressure ratio is obtained from the pressures with and without a shear layer.
The observers A and O are in the same positions in the theoretical calculations (Amiet’s method) and in interpolating the
perturbed pressure in the numerical evaluation (GTSF). pO is the amplitude recorded outside the shear layer and pA is the
magnitude in a uniform moving flow. The ratio of pA and pO represents the pressure variation caused by the shear layer
due to refraction of the sound path and divergence of the ray tube.
Fig. 10 shows the dependence of refraction angle, as predicted by Amiet and GTSF, on the Mach number (0.2 to 0.8)
and temperature ratio (T jet /Tair = [1.0, 2.0]). For the isothermal configuration in Figs. 10(a) and 10(b), the magnitude of the
refraction angle gradually increases from downstream (x = 0.4 ) to upstream (x = −0.4 ), which indicates that the sound is
13
JID: YJSVI
Fig. 10. Refraction angle θO in Eq. (14) as a function of jet Mach number, observer position, and temperature ratio. The refraction angle is calculated
by Amiet’s method (left) and GTSF (right). The temperature ratio is T jet /Tair = 1.0 for (a) and (b), and T jet /Tair = 2.0 for (c) and (d). The virtual array is at
x = −0.4 to 0.4 and y = 0.6 and has 287 evenly distributed points.
fully deflected to the left without interference. This is consistent with Fig. 4, in which the wavefront in a quiet medium is
refracted upstream. For the same x, the refraction angle increases with Mach number due to the stronger velocity gradient
on both sides of the shear layer. As for a fixed jet velocity and source frequency, the superposition of a higher temperature
increases the sound speed in the jet flow, as shown in Figs. 10(c) and 10(d). Indirectly, it reduces the local convective Mach
number. Since the effect of convection is alleviated by the temperature, the refraction angle decreases accordingly. The
refraction angles for both Amiet’s method and GTSF are consistent, which confirms the potential ability of Amiet’s method
for determining refraction in a heated flow.
Fig. 11 depicts the amplitude correction for jet Mach number M jet = [0.2, 0.4, 0.8] and temperature ratio T jet /Tair =
[1.0, 2.0]. In Figs. 11(a) and 11(b), the trend of corrected pressure ratio is consistent with the isothermally experimental
and theoretical results [5,62]. The influence of the cold shear flow on the wave amplitude is imperceptible at low Mach
number. It becomes increasingly important at high Mach number and is a maximum upstream. This tendency is an indirect
result of shear layer refraction [4]. A large Mach number increases the transmission loss of the shear layer in Eq. (23), en-
hances the energy decay of the ray tube in Eq. (22), and causes a longer refraction trajectory, which increases the pressure
difference between the original position O and the corrected position A. Moreover, compared to the incident direction, the
refracted ray is bent upstream toward the point O in Fig. 3, which is ahead of point A. Thus, the magnitude of pO is smaller
than that of pA at x < 0 and the ratio pA /pO is greater in the upstream region. With heating, the magnitude of the ampli-
tude ratio, as shown in Figs. 11(c)–11(d), decreases due to the reduction of the local convective Mach number. This implies
that a higher temperature tends to counterbalance the velocity effect over the amplitude correction. Besides, the results for
Amiet’s method and GTSF have similar trends.
In summary, GTSF provides novel insights into Amiet’s traditional correction method, relating to the refraction angle,
amplitude correction, and temperature. The refraction angle is accurately predicted by Amiet’s method via Eq. (14) for a
14
JID: YJSVI
Fig. 11. Pressure ratio pA /pO in Eq (21) as a function of jet Mach number, observer position, and temperature ratio. The pressure ratio is calculated by
Amiet’s method (left) and GTSF (right). The temperature ratio is T jet /Tair = 1.0 for (a) and (b), and T jet /Tair = 2.0 for (c) and (d). The virtual array is at
x = −0.4 to 0.4 and y = 0.6 and has 287 evenly distributed points.
specific jet temperature. It is applicable for the fast refractions in the region above the source. The amplitude correction
becomes increasingly significant at high Mach numbers and upstream. Here, Eq. (21) can provide the general trends, but
is limited to configurations with one moving medium and one stationary medium. For refraction by two moving streams, it
has to be modified with the static-to-flight approach or other approaches, like a numerical simulation.
5.2. Influence of velocity and temperature gradients
As well as the numerical evaluation, it is necessary to assess the applicability of Amiet’s method for source localization. If
not specified hereafter, the virtual array is spaced as x = 0.0028 and centered at (x, y ) = (0.0, 0.6 ) and has 287 equidistant
grid points. The sampling frequency of the signal is equivalent to the numerical time step. The sampling starts at t = 1.2 and
ends at t = 5.0. The scanning region is at y = −0.3 and from x = −0.5 to x = 0.5, and is divided into 2500 possible sources.
We consider the localization error, which is defined as the distance between the predicted position and the actual source
position in the source term, Eq. (28). This error is xshi f t , which is normalized by the wavelength λ = cair / f in the far field.
If needed, the wavelength can be transformed to a Helmholtz number by multiplying by 2π [11].
Fig. 12 shows the localization errors for M jet = 0.2 and 0.8 by considering or neglecting the velocity or temperature gra-
dient in Amiet’s model. Here, a thin shear layer with σ0 = 106 is adopted to exclude the spreading effect. If the influence of
flow is not considered, the estimated source drifts downstream and is positive, as shown in Fig. 12(a). Moreover, the error
level is slightly reduced by the high gas temperature due to the attenuated convection. When only the velocity gradient
is corrected in the beamforming process, the isothermal case has a smaller shift and the other cases are erroneous. The
convection in a hot flow is overestimated by Amiet’s model, which uses an isothermal transmission relation. Thisoveresti-
mation explains why for T jet /Tair = 1.5 the shift becomes negative, and further negative for T jet /Tair = 2.0. By considering all
15
JID: YJSVI
Fig. 12. Normalized shift of the source ( f = 14.7) in a thin shear layer for (a) M jet = 0.2 and (b) M jet = 0.8 as a function of temperature ratio T jet /Tair . The
blue line (squares) is without the velocity and temperature gradients. The black line (triangles) is with only the velocity gradient. The red line (circles) is
with the velocity and temperature gradients. The virtual array is at x = −0.4 to 0.4 and y = 0.6 and has 287 evenly distributed points. (For interpretation
of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 13. Normalized shift of source ( f = 14.7) in different shear layers (σ0 = [9.0, 11.0, 13.0, 106 ]) with (a) M jet = 0.2 and (b) M jet = 0.8 as a function of
temperature ratio T jet /Tair with consideration of complete flow characteristics of velocity and temperature gradients. The virtual array is at x = 0.42 to 0.7
and y = 0.6 and has 100 evenly distributed points.
the flow characteristics, the temperature effect is successfully taken into account in Amiet’s correction. Increasing the Mach
number from M jet = 0.2 to 0.8, the effects of the velocity and temperature gradients on the source shift increase, as shown
in Fig. 12(b), because the shift is approximately proportional to the Mach number [5]. By using a combination of the velocity
and temperature gradients, the localization has an acceptable error for both cold and hot jets.
5.3. Influence of spreading angle
Fig. 13 compares the localization errors of a sound source for different shear layers with (σ0 = [9.0, 11.0, 13.0]) or with-
out (σ0 = 106 ) spreading for f = 14.7,M jet = [0.2, 0.8], and T jet /Tair = [1.0, 1.5, 2.0]. The center of the microphone array is
in a downstream position (0.56, 0.60 ) and has 100 adjacent grid points to record data for processing with DAS beam-
forming and Amiet’s correction. From Fig. 13(a), it can be seen that the thin shear layer, which basically satisfies Amiet’s
zero-thickness assumption, has localization errors close to 0 for different temperature ratios. However, with spreading, the
increasing thickness no longer satisfies this assumption. The located source position shifts downstream and the localization
error has a positive magnitude. As the jet temperature becomes higher, the spreading angle together with the shear layer
thickness increases according to Fig. 5, which further results in a relatively large localization error. Increasing the Mach
number from M jet = 0.2 in Fig. 13(a) to M jet = 0.8 in Fig. 13(b), the localization error in the thin configuration is still small,
while with spreading, it is larger. The maximum source shift xshi f t = 1.1, which is the same order of magnitude as with
the velocity and temperature gradients mentioned in Section 5.2. This highlights the importance of spreading angle in the
localization of a sound source downstream of a high-speed hot jet, where the shear layer has a significant thickness [63].
5.4. Influence of source frequency
Fig. 14 shows the localization errors of a sound source for different frequencies in a spreading shear layer (σ0 = 11.0)
for f = [8.82, 14.7, 19.8],M jet = [0.2, 0.8], and T jet /Tair = [1.0, 1.5, 2.0]. The virtual array is at x = −0.4 to 0.4 and y = 0.6 and
16
JID: YJSVI
Fig. 14. Normalized shift of source f = [8.82, 14.7, 19.8] in a spreading shear layer (σ0 = 11.0) for (a) M jet = 0.2 and (b) M jet = 0.8 as a function of tem-
perature ratio T jet /Tair with consideration of complete flow characteristics of velocity and temperature gradients. The virtual array is at x = −0.4 to 0.4 and
y = 0.6 and has 287 evenly distributed points.
Fig. 15. (a) Sketch of the definition of mean thickness L. (b) Normalized source shift for the GTSF numerical database as a function of the product of the
jet Mach number M jet and Strouhal number St. The solid circles denote the calculated data, which are fitted to a line.
has 287 evenly distributed points. Fig. 14(a) shows the low-speed configuration. The localization error of the source for
different frequencies increases slightly with the temperature ratio. For the same temperature but higher frequency, there is
a larger localization error. Fig. 14(b) shows the results for a high-speed jet with M jet = 0.8. The localization error is larger in
comparison with Fig. 14(a). The difference for the localization error between the two frequencies is amplified by the Mach
number. So, the wavelength of the sound also affects the source shift and it needs to be considered if accurate acoustic
characteristics are required.
Other tests on the dependence of the localization error on the pressure amplitude ratio pA /pO and the steering vector
χK/2 /χk in Eq. (11) were done but the results are not shown here. They contribute only to a maximum improvement of
the source location of about 0.02 for M jet = 0.8 and T jet /Tair = 1.0. Since the maximum of the absolute source shift for the
velocity and temperature gradient in Fig. 12is 3.74 and for the spreading angle and source frequency in Figs. 13 and 14 is
1.56, it is clear that different factors affect the refraction in addition to the velocity and temperature gradients. The most
important of these other factors are the spreading angle and source frequency, followed by the amplitude correction and
the steering vector. The amplitude correction can generally be ignored as pA /pO = 1.0 for a low-speed jet flow, whereas it is
also of less importance for the localization of the source in a high-speed application. Therefore, in the next section, the jet
Mach number as a function of velocity gradient, the shear layer thickness as a function of temperature and spreading angle
as well as the source frequency are considered in an empirical scaling.
5.5. Evaluation of empirical scaling of localization error
Following an earlier isothermal study [36], the normalized localization error or source shift xshi f t depends on two scaling
factors: jet Mach number M jet and the Strouhal number St. The latter is defined as St = f Lmean /cair = Lmean /λ with Lmean
the average thickness. A similar scaling was also used in Eqs. (2.88) and (6.3) of [54] to construct a dimensionless pa-
rameter for spectral broadening. As briefly shown in Fig. 15(a), the thickness is defined by Eq. (A.8) as the distance be-
tween the boundaries with 5% and 95% of the jet velocity. Moreover, Lmean = (L1 + LK )/2 is half of the thickness projected
from the 1-st and K-th microphones to the shear layer. This mean thickness depends on the spreading rate and temper-
ature ratio if we filter the spreading behavior of the velocity and temperature. Fig. 15(b) shows the beamforming results
processed through the 108 configurations calculated in the above sections for jet Mach number M jet = [0.2, 0.4, 0.8], tem-
17
JID: YJSVI
perature ratio T jet /Tair = [1.0, 1.5, 2.0], spreading rate σ0 = [9.0, 11.0, 13.0, 106 ], frequency f = [8.82, 14.7, 19.8], and array
position y = [0.4, 0.6, 0.8] with 287 surrounding mesh points.
Fig. 15 (b) shows that the localization error depends nearly linearly on the product of the scaling factors M jet and St. This
confirms other research and the above discussion. In a numerical study on localizing a 10-kHz source below a cold parallel
shear layer [5], the Mach number and mean flow thickness were confirmed to be the dominant coefficients influencing
the localization accuracy. In the experimental result of [64], the position shift after Amiet’s correction was found to be
proportional to the jet Mach number in the open test section of a wind tunnel. Another experiment over a range of jet
velocities (33.3–69.4 m/s) was conducted in a 3/4 open-jet wind tunnel by Wang et al. [63], who observed a larger source
drift in the middle test section than in the front section near the nozzle exit.
In Fig. 15(b), the discrete points near the origin indicate small errors of sound localization, which represent configura-
tions of a thin shear layer with Lmean close to 0. This trend is consistent with the results in Section 5.2, where there is
a pretty small error when considering the velocity and temperature gradients of a thin shear flow. When taking into ac-
count spreading, the mean thickness Lmean is non-zero and M jet · St increases as the temperature ratio, jet Mach number,
and source frequency increase or as the spreading rate decreases. As discussed in Section 5.3, the spreading angle and mean
thickness increase as the temperature ratio increases or the spreading rate decreases, which leads to a larger source shift in
Figs. 13(a) and 13(b). Comparing the two figures, the source shift also increased on changing the jet Mach number from 0.2
to 0.8. In Section 5.4, the position drift is positively correlated to the source frequency, which determines the wavelength
in St at a constant far-field sound speed cair . Therefore, the relation between the source shift xshi f t and the product M jet · St
is consistent with the aforementioned results. St = Lmean /λ was previously used to assess the degree to which sound wave
propagation is affected by the size of flow inhomogeneities [13]. It is included in the current scaling relation to show that
the effect of the mean shear layer thickness on source localization is considerable for Lmean /λ 1 and small for Lmean /λ 1.
The broad range of discrete points in Fig. 15(b) was linearly fitted by setting the intercept to zero. The fitting equation is
xshi f t = 0.2368M jet · St. Note that the fitted data were from a 2D numerical database, which might not be general enough to
cover all situations. The localization results are also sensitive to the beamforming technique [10], 3D effects [9], turbulence
[4]as well as many unknowns and uncertainties in numerical and real environments. The linear fit highlights the importance
of the non-dimensional jet Mach number and Strouhal number on source localization. It may be convenient for roughly
forecasting the possible localization error in applications where a numerical simulation would be limited by computing
resources. For example, in a large wind tunnel [63], the downstream thickness of a M jet = 0.2 shear flow is of the order of
0.5 m, so for a 10-kHz sound source, the shift would be 0.7 m.
6. Conclusions
There are numerical Kelvin-Helmholtz instabilities in linearized simulations of sound propagation through a steady shear
layer. These result in self-excited perturbations and cause the solution to diverge. To investigate sound refraction in a sheared
environment, a GTSF method to suppress the development of instabilities is proposed and tested. Using GTSF, sound prop-
agation and source localization in a hot jet flow were numerically investigated for jet Mach numbers M jet = [0.2, 0.4, 0.8]
∗ = [300, 450, 600](K).
and temperatures T jet
For sound propagation in a thin shear layer, a high jet velocity promotes acoustic refraction, whereas a high temperature
weakens the refraction by increasing the sound wavelength and decreasing the mean flow gradient. With a spreading rate,
the decrease in spreading angle due to compressibility is negligible for the configurations considered, but increases with
temperature, which can enhance refraction. Compared to the configuration of spreading rate σ0 = 11.0, the variation of
spreading angle is found to exert a minimal effect on far-field directivity, whereas it significantly changes the acoustic phase
in the refracted region. We recommend replacing a thin shear layer with a spreading one, as this is more realistic and would
give smooth wavefronts in environments with a high Mach number and steep temperature gradient.
For source localization, Amiet’s corrections were extended to non-isothermal cases and compared with GTSF in terms of
direction and amplitude. Amiet’s theory can provide a relatively accurate refraction prediction and a right amplitude trend
above the sound source. When the gas temperature in jet increases, the corrections of both direction and amplitude are
weakened. Next, numerical beamforming via Amiet’s correction was used to examine the influence of velocity and tem-
perature gradients, spreading angle, and source frequency on the localization error. Amiet’s theory can accurately correct
the refraction induced by the velocity and temperature gradients of a thin shear layer. For a spreading shear layer, the lo-
calization error increases as the temperature ratio, jet Mach number, and source frequency increase or the spreading rate
decreases. Finally, it was empirically shown that the localization error has a mainly linear dependence on the product of
two scaling factors, the jet Mach number and the Strouhal number.
In brief, this work was mainly performed in 2D using a steady shear layer as the background flow for numerical sound
propagation and source localization. Ultimately, detailed studies of 3D applications are required. The spectral broadening
induced by a volume of turbulence is also worth investigating.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have
appeared to influence the work reported in this paper.
18
JID: YJSVI
CRediT authorship contribution statement
Lican Wang: Conceptualization, Methodology, Software, Investigation, Validation, Visualization, Writing - original draft.
Rongqian Chen: Conceptualization, Methodology, Software, Writing - review & editing, Funding acquisition. Yancheng You:
Writing - review & editing, Funding acquisition, Supervision. Ruofan Qiu: Writing - review & editing.
Acknowledgements
The authors are warmly grateful for financial support from the National Natural Science Foundation of China under Grant
No. 11602209 and from the Key Laboratory of Aerodynamic Noise Control under Grant No. ANCL20190203. The authors
would like to kindly thank Prof. Wang Xunnian and Prof. Zhang Shuhai for their suggestions for solving engineering prob-
lems using numerical methods. The authors want to acknowledge Dr. Wang Zhaohuan for invaluable discussions and his
constructive advice on deriving the monopole steering vector.
Appendix A. Self-similar shear layer with compressibility and temperature correction
We developed a unified and efficient way to model self-similar shear flow with compressibility and heat correction. The
Görtler velocity profile [54] was adopted for incompressible situations and the extended correction coefficient was obtained
from [57]. This was fitted and compared with an experimental database. For simplicity, we assumed that the planar shear
layer quickly achieved velocity self-similarity and linearly grew downstream of the initial mixing region.
For constant turbulent viscosity and pressure, an incompressible Görtler’s distribution with a compressibility modification
profile can be derived for ambient air above the jet as
u0 −uair
= 12 [1 + erf (ξ + ξ0 )],
u jet −uair ! (A.1)
v0 −vair
ξ0 erf (ξ + ξ0 ) + − ( ξ + ξ0 )
− 2ξσ0 ,
2
u jet −uair
= 21σ √1
πe
where subscripts jet and air denote the flow parameters inside the jet and outside the shear layer, respectively. Other
variables include:
σ (y − ysh )
ξ =− , (A.2)
x − xsh
σ0
σ= (A.3)
λs

1 − uair /u jet ρair /ρ jet
1+ 1+μ
λs = = , (A.4)
2 1 + uair /u jet ρair /ρ jet 2

1
= 1 + 0.5468 −1 , (A.5)
1 + 35.53Mc8.614
u jet − uair M jet
Mc = = , (A.6)
c jet + cair μ+1
where uair = 0 and (xsh , ysh ) is the origin of the shear layer. The experimental spreading rate σ0 = 9 ∼ 15 [54,65]. σ is the
spreading rate corrected for compressibility and temperature. ξ0 = 0.4 [54] is the shift of the shear layer centerline due to
nozzle thickness. Mc is the convective Mach number. is the compressible correction factor [57,58]. It connects the shear-
layer spreading rate with the compressibility. At low speed, → 1 and gradually degrades with larger convective Mach
number Mc . The ratio λs takes into account the influence of velocity and density on the spreading rate. The shear layer
is bordered by velocity ratios u0 and u jet , where u0 /u jet = 5% for a quiet medium and u0 /u jet = 95% for the jet side [54].
Equation (A.1) can be reformulated as:
$ %
1 u
y=− (x − xsh ) erfinv 2 0 − 1 − ξ0 + ysh , (A.7)
σ u jet
where erfinv is the inverse of the error function erf in MATLAB. The distance in the y direction between the two velocity
boundaries is the thickness of the shear layer at position x:
2.3262(x − xsh )
L= , (A.8)
σ
which is similar to Eq. (2.33) in [54]. The spreading angle
L
! 2.3262
!
θsp = tan−1 = tan−1 . (A.9)
x − xsh σ
19
JID: YJSVI
For an ideal gas with constant isobaric mean pressure p0 and Prandtl number P r = 1.0, the density profile can be for-
mulated using the Crocco–Busemann relation [55]:
γ − 1 |V0 | ρair u jet − ρ jet uair
ρ0 = −|V0 |2 + u jet + uair |V0 | − u jet uair + ρ jet − ρair + . (A.10)
2γ p0 u jet − uair u jet − uair
The temperature can be determined from the ideal-gas law: p0 = ρ0 Rg T0 with the gas constant Rg = 287 m2 s−2 K−1 .
References
[1] X. Zhang, X. Chen, J.R. Gill, Gradient term filtering for stable sound propagation with linearized Euler equations, in: 20th AIAA/CEAS Aeroacoustics
Conference, 2014, p. 3306, doi:10.2514/6.2014-3306.
[2] Y. Sun, R. Fattah, S. Zhong, X. Zhang, Stable time-domain CAA simulations with linearised governing equations, Comput. Fluids 167 (2018) 187–195,
doi:10.1016/j.compfluid.2018.03.025.
[3] F.Q. Hu, X. Li, X. Li, M. Jiang, Time domain wave packet method and suppression of instability waves in aeroacoustic computations, J. Fluids Eng. 136
(6) (2014) 060905, doi:10.1115/1.4025866.
[4] J. Zhang, X. Wang, J. Zhang, C.J. Doolan, J.R. Fischer, D. Moreau, A study of shear-layer corrections and a tensioned fabric wall for the localization of
sound sources in wind tunnel, in: 25th AIAA/CEAS Aeroacoustics Conference, 2019, p. 2717, doi:10.2514/6.2019-2717.
[5] T. Padois, C. Prax, V. Valeau, Numerical validation of shear flow corrections for beamforming acoustic source localisation in open wind-tunnels, Appl.
Acoust. 74 (4) (2013) 591–601, doi:10.1016/j.apacoust.2012.09.013.
[6] J.W. Miles, On the reflection of sound at an interface of relative motion, J. Acoust. Soc. Am. 29 (2) (1957) 226–228, doi:10.1121/1.1908836.
[7] H.S. Ribner, Reflection, transmission, and amplification of sound by a moving medium, J. Acoust. Soc. Am. 29 (4) (1957) 435–441, doi:10.1121/1.1908918.
[8] R. Amiet, Refraction of sound by a shear layer, J. Sound Vibr. 58 (4) (1978) 467–482, doi:10.1016/0022- 460X(78)90353- X.
[9] R. Porteous, T. Geyer, D.J. Moreau, C.J. Doolan, A correction method for acoustic source localisation in convex shear layer geometries, Appl. Acoust. 130
(2018) 128–132, doi:10.1016/j.apacoust.2017.09.020.
[10] L. Wang, R. Chen, Y. You, W. Wu, R. Qiu, A unified correction method for the acoustic refraction (UCMAR) caused by a three dimensional shear layer,
Acta Acust. United Acust. 105 (5) (2019) 732–742, doi:10.3813/AAA.919353.
[11] V. Clair, G. Gabard, Spectral broadening of acoustic waves by convected vortices, J. Fluid Mech. 841 (2018) 50–80, doi:10.1017/jfm.2018.94.
[12] I. Bennaceur, D. Mincu, I. Mary, M. Terracol, L. Larchevque, P. Dupont, Numerical simulation of acoustic scattering by a plane turbulent shear layer:
spectral broadening study, Comput. Fluids 138 (2016) 83–98, doi:10.1016/j.compfluid.2016.08.012.
[13] J. Thomas, New model for acoustic waves propagating through a vortical flow, J. Fluid Mech. 823 (2017) 658–674, doi:10.1017/jfm.2017.339.
[14] K. Ahuja, H. Tanna, B. Tester, An experimental study of transmission, reflection and scattering of sound in a free jet flight simulation facility and
comparison with theory, J. Sound Vibr. 75 (1) (1981) 51–85, doi:10.1016/0022- 460X(81)90235- 2.
[15] A. Michalke, On spatially growing disturbances in an inviscid shear layer, J. Fluid Mech. 23 (3) (1965) 521–544, doi:10.1017/S0 0221120650 01520.
[16] R. Astley, R. Sugimoto, P. Mustafi, Computational aero-acoustics for fan duct propagation and radiation. Current status and application to turbofan liner
optimisation, J. Sound Vibr. 330 (16) (2011) 3832–3845, doi:10.1016/j.jsv.2011.03.022.
[17] Y. Özyörük, B. Tester, Application of frequency-domain linearized euler solutions to the prediction of aft fan tones and comparison with experimental
measurements on model scale turbofan exhaust nozzles, J. Sound Vibr. 330 (16) (2011) 3846–3858, doi:10.1016/j.jsv.2011.02.008.
[18] L.S. Kovasznay, Turbulence in supersonic flow, J. Aeronaut. Sci. 20 (10) (1953) 657–674, doi:10.2514/8.2793.
[19] M.D. Dahl, Fourth computational aeroacoustics (CAA) workshop on benchmark problems, Technical Report, 2004.
[20] A. Agarwal, P.J. Morris, R. Mani, Calculation of sound propagation in nonuniform flows: suppression of instability waves, AIAA J. 42 (1) (2004) 80–88,
doi:10.2514/1.619.
[21] K. Hamiche, S. Le Bras, G. Gabard, H. Bériot, Hybrid numerical model for acoustic propagation through sheared flows, J. Sound Vibr. 463 (2019) 114951,
doi:10.1016/j.jsv.2019.114951.
[22] S. Zheng, M. Zhuang, Time-domain acoustic wave solutions in sheared mean flows, 5, 2006, pp. 263–277, doi:10.1260/1475-472X.5.3.263.
[23] C. Bogey, C. Bailly, D. Juvé, Computation of flow noise using source terms in linearized Euler’s equations, AIAA J. 40 (2) (2002) 235–243, doi:10.2514/
2.1665.
[24] J.-H. Seo, Y.J. Moon, Perturbed compressible equations for aeroacoustic noise prediction at low Mach numbers, AIAA J. 43 (8) (2005) 1716–1724,
doi:10.2514/6.2005-2927.
[25] R. Ewert, W. Schröder, Acoustic perturbation equations based on flow decomposition via source filtering, J. Comput. Phys. 188 (2) (2003) 365–398,
doi:10.1016/S0 021-9991(03)0 0168-2.
[26] C. Prax, F. Golanski, L. Nadal, Control of the vorticity mode in the linearized Euler equations for hybrid aeroacoustic prediction, J. Comput. Phys. 227
(12) (2008) 6044–6057, doi:10.1016/j.jcp.2008.02.022.
[27] S. Zhong, X. Zhang, A sound extrapolation method for aeroacoustics far-field prediction in presence of vortical waves, J Fluid Mech 820 (2017) 424C450,
doi:10.1017/jfm.2017.219.
[28] R. Ewert, O. Kornow, J. Delfs, T. Roeber, M. Rose, A CAA based approach to tone haystacking, in: 15th AIAA/CEAS Aeroacoustics Conference, 2009,
p. 3217, doi:10.2514/6.2009-3217.
[29] S. Zhong, X. Zhang, A generalized sound extrapolation method for turbulent flows, Proceedings of The Royal Society A: Mathematical, Physical and
Engineering Sciences 474 (2210) (2018) 20170614.
[30] S. Redonnet, Investigation of the acoustic installation effects of an open-jet anechoic wind tunnel using computational aeroacoustics, 169, 2020,
p. 107469, doi:10.1016/j.apacoust.2020.107469.
[31] D. Casalino, Finite element solutions of a wave equation for sound propagation in sheared flows, AIAA J. 50 (1) (2012) 37–45, doi:10.2514/1.J050772.
[32] J. Jiao, Aeroacoustic wind tunnel correction based on numerical simulation, DLR, 2017 Ph.D. thesis. https://www.researchgate.net/publication/
318491731_Aeroacoustic_Wind_Tunnel_Correction_Based_on_Numerical_Simulation.
[33] M. Moratilla-Vega, K. Lackhove, J. Janicka, H. Xia, G. Page, Jet noise analysis using an efficient les/high-order acoustic coupling method, Comput. Fluids
199 (2020) 104438, doi:10.1016/j.compfluid.2020.104438.
[34] J. Bult, S. Redonnet, Landing gear noise identification using phased array with experimental and computational data, AIAA Journal 55 (11) (2017)
3839–3850, doi:10.2514/1.J055643.
[35] D. Evans, M. Hartmann, J. Delfs, Beamforming for point force surface sources in numerical data, J. Sound Vibr. 458 (2019) 303–319, doi:10.1016/j.jsv.
2019.05.030.
[36] L. Wang, R. Chen, Y. You, C. Zhengwu, Q. Ruofan, Effects of shear layer characteristics on acoustic propagation and source localization, J. Northwest.
Polytech. Univ. 37 (6) (2019) 1148–1157, doi:10.1051/jnwpu/20193761148.
[37] I. Rakotoarisoa, D. Marx, C. Prax, V. Valeau, Array processing for the localisation of noise sources in hot flows, Mech. Syst. Signal Proc. 116 (2019)
160–172, doi:10.1016/j.ymssp.2018.06.038.
[38] V. Suponitsky, N.D. Sandham, A. Agarwal, On the Mach number and temperature dependence of jet noise: results from a simplified numerical model,
J. Sound Vibr. 330 (17) (2011) 4123–4138, doi:10.1016/j.jsv.2011.02.007.
[39] I. Rakotoarisoa, J.R. Fischer, V. Valeau, D. Marx, C. Prax, L.E. Brizzi, Time-domain delay-and-sum beamforming for time-reversal detection of intermit-
tent acoustic sources in flows, J. Acoust. Soc. Am. 136 (5) (2014) 2675–2686, doi:10.1121/1.4897402.
20
JID: YJSVI
[40] J.H. Seo, Y.J. Moon, Linearized perturbed compressible equations for low Mach number aeroacoustics, J. Comput. Phys. 218 (2) (2006) 702–719, doi:10.
1016/j.jcp.20 06.03.0 03.
[41] R. Chen, Hybrid methods for the computational aeroacoustics based on acoustic propagation equations, Nanjing University of Aeronautics and Astro-
nautics, 2012 Ph.D. thesis. http://cdmd.cnki.com.cn/Article/CDMD-10287-1014005421.htm.
[42] C.K. Tam, J.C. Webb, Dispersion-relation-preserving finite difference schemes for computational acoustics, J. Comput. Phys. 107 (2) (1993) 262–281,
doi:10.1006/jcph.1993.1142.
[43] C. Bogey, C. Bailly, A family of low dispersive and low dissipative explicit schemes for flow and noise computations, J. Comput. Phys. 194 (1) (2004)
194–214, doi:10.1016/j.jcp.20 03.09.0 03.
[44] T.F. Brooks, W.M. Humphreys, A deconvolution approach for the mapping of acoustic sources (DAMAS) determined from phased microphone arrays, J.
Sound Vibr. 294 (4) (2006) 856–879, doi:10.1016/j.jsv.2005.12.046.
[45] P. Sijtsma, CLEAN based on spatial source coherence, Int. J. Aeroacoust. 6 (4) (2007) 357–374, doi:10.1260/147547207783359459.
[46] Sarradj, Ennes, Three-dimensional acoustic source mapping with different beamforming steering vector formulations, Adv. Acoust. Vibr. 20121–12.
doi:https://doi.org/10.1155/2012/292695.
[47] W. Dobrzynski, Shear-layer correction after Amiet under consideration of additional temperature gradient. Working diagrams for correction of signals,
Technical Report, 1984 https://ntrs.nasa.gov/search.jsp?R=19850 0 04346.
[48] E. Sarradj, A fast ray casting method for sound refraction at shear layers, Int. J. Aeroacoust. 16 (1) (2016) 65–77, doi:10.1177/1475472X16680463.
[49] C. Bailly, D. Juve, Numerical solution of acoustic propagation problems using linearized Euler equations, AIAA J. 38 (1) (20 0 0) 22–29, doi:10.2514/2.949.
[50] S.M. Candel, C. Crance, Direct fourier synthesis of waves in layered media - Application to acoustic source radiation at the edge of uniform flow, 18th
Aerospace Sciences Meeting, 1980, doi:10.2514/6.1980-37.
[51] Z.-H. Wang, S.W. Rienstra, C.-X. Bi, B. Koren, An accurate and efficient computational method for time-domain aeroacoustic scattering, J. Comput. Phys.
(2020) 109442, doi:10.1016/j.jcp.2020.109442.
[52] S.W. Rienstra, A. Hirschberg, An Introduction to Acoustics, 18, 2004 https://www.win.tue.nl/~sjoerdr/.
[53] S. Candel, Numerical solution of conservation equations arising in linear wave theory - application to aeroacoustics, 83, 1977, pp. 465–493, doi:10.
1017/S0 0221120770 01293.
[54] Z. Sulaiman, Effect of open-jet shear layers on aeroacoustic wind tunnel measurements, Delft University of Technology, 2011 Ph.D. thesis. https:
//www.researchgate.net/publication/265033147_Effect_of_Open-Jet_Shear_Layers_on_Aeroacoustic_Wind_Tunnel_Measurements.
[55] F.M. White, I. Corfield, Viscous Fluid Flow (Chapter 7), 3, McGraw-Hill New York, 2006.
[56] G. Gabard, Boundary layer effects on liners for aircraft engines, J. Sound Vibr. 381 (2016) 30–47, doi:10.1016/j.jsv.2016.06.032.
[57] M.F. Barone, W.L. Oberkampf, F.G. Blottner, Validation case study: prediction of compressible turbulent mixing layer growth rate, AIAA J. 44 (7) (2006)
1488–1497, doi:10.2514/1.19919.
[58] D.A. Yoder, J.R. DeBonis, N.J. Georgiadis, Modeling of turbulent free shear flows, Comput. Fluids 117 (2015) 212–232, doi:10.1016/j.compfluid.2015.05.
009.
[59] J. Manera, B. Schiltz, R. Leneveu, S. Caro, J. Jacqmot, S. Rienstra, Kelvin-Helmholtz instabilities occurring at a nacelle exhaust, in: 14th AIAA/CEAS
Aeroacoustics Conference, 20 08, p. 2883, doi:10.2514/6.20 08-2883.
[60] R. Leneveu, B. Schiltz, J. Manera, S. Caro, Paralled DGM scheme for LEE applied to exhaust and bypass problems, in: 13th AIAA/CEAS Aeroacoustics
Conference, 2007, doi:10.2514/6.2007-3510.
[61] A.R. Pilon, A.S. Lyrintzis, Mean flow refraction corrections for the Kirchhoff method, J. Aircr. 35 (4) (1998) 661–664, doi:10.2514/2.2355.
[62] R. Schlinker, R. Amiet, Experimental assessment of theory for refraction of sound by a shear layer, Technical Report, 1978 https://ntrs.nasa.gov/search.
jsp?R=19780017888.
[63] Y. Wang, J. Yang, Q. Jia, Z. Yang, Z. Shen, An improved correction method for sound source drift in a jet flow and its application to a wind tunnel
measurement, Acta Acust. United Acust. 101 (3) (2015), doi:10.3813/AAA.918859. 642–649(8).
[64] X. Zhang, B. Chen, Q. Lu, Verification of angle refraction correction based on Amiet’s shear layer theory, J. Appl. Acoust. (2014), doi:10.11684/j.issn.
10 0 0-310X.2014.05.0 09.
[65] Z. Ni, J. Zhang, M. Wang, J. Zhang, Characteristics of shear-layer flow field of open-jet wind tunnels, J. Aerosp. Power 35 (2) (2020) 244–251, doi:10.
13224/j.cnki.jasp.2020.02.003.
21

Wang 2020

Uploaded by

Copyright:

Available Formats

Wang 2020

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Wang 2020

Uploaded by

Copyright:

Available Formats

JID: YJSVI

ARTICLE IN PRESS [m3Gsc;October 23, 2020;13:59]

Contents lists available at ScienceDirect

Journal of Sound and Vibration

Investigation of acoustic propagation and source localization

2. GTSF method for acoustic propagation

2.1. GTSF method

with a Mach number dependency [13,40] scaled by a parallel shear ﬂow (∇ · V0 = 0)

2.2. Numerical methods and validation

GTS-1 14.63% 16.43% 13.97% 14.63%

To evaluate the accuracy quantitatively, the L2 numerical error is expressed as

Fig. 3. Sketch map of sound propagation through a jet shear layer.

3. DAS beamforming method for source localization

3.1. Refraction correction for a hot jet

t = tSB + tBO , (18)

3.2. Amplitude correction for a hot jet

pA pB+ pB− pA

4. Sound propagation in a hot jet ﬂow

4.1. Inﬂuence of velocity and temperature gradients

4.2. Inﬂuence of spreading angle

4.3. Inﬂuence of source frequency

5. Source localization in a hot jet ﬂow

5.1. Evaluation of Amiet’s correction method

5.2. Inﬂuence of velocity and temperature gradients

5.3. Inﬂuence of spreading angle

5.4. Inﬂuence of source frequency

5.5. Evaluation of empirical scaling of localization error

Declaration of Competing Interest

CRediT authorship contribution statement

Appendix A. Self-similar shear layer with compressibility and temperature correction

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

pA pB+ pB− pA