Sie sind auf Seite 1von 18

Aeroacoustics: Lighthill's analogy and its

limitations

Research Seminar report

Information Technology
Alpen-Adria-University Klagenfurt
Faculty for Technical sciences

Univ.-Prof. Dr. techn. Dr.-Ing. habil Manfred Kaltenbacher


Applied Mechatronics
Klagenfurt

Abstract

A huge amount of noise is heard everyday in our daily life. It is generated by turbulent
ows. To study the ow induced sound, the physics involved is very complicated. The
use of numerical simulation is a good way to analyze this type of problems. To study
the computation of ow induced noise dierent methods are being developed which are
based on Lighthill's analogy. In this paper, the formulation of Lighththill's analogy and
its limitations are presented. In the limitations, how the acoustic analogy theory is not
applicable to real world problems is discussed. The requirements for the sound simulation
are also presented.

Contents

Introduction

1.1

1.2
2

Domain decomposition .
1.1.1 Sound source . . .
1.1.2 Propagation region
1.1.3 Far-eld . . . . .
Aero-Acoustical Analogy

. . .
. . .
. .
. . .
. . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

1
3
3
3
3

Lighthill's Analogy

2.1
2.2

6
7

Derivation of Lighthill's equation . . . . . . . . . . . . . . . . . . . . . . . .


Approximation of Lighthill's stress tensor . . . . . . . . . . . . . . . . . . .

Limitations in Lighthill's analogy

Bibliography

12

ii

List of Figures

1.1
1.2
1.3

Essential ingredients for numerical scheme[12] . . . . . . . . . . . . . . . . .


Domain decomposition[22] . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Process Overview[4] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

iii

2
2
4

Chapter 1
Introduction

Aeroacoustics is the study of the generation and transmission of sound by uid ow. Computational aeroacoustics includes dierent numerical schemes and algorithms to analyze
the ow induced sound. In order to provide a reliable simulation we need to consider the
following characteristics

Good far-eld prediction


Reasonable computing time
Reproduction of sound characteristics
Conservation of the dispersion relation
Propagation without dissipation
The essential ingredients necessary for the numerical scheme (see gure 1.1) are:

Flow : ow generates sound waves (pressure uctuations)


Source: ow noise is generated in a particular region, e.g. wake or boundary layer
Acoustic medium (air, water) to transport or propagate the sound waves

1.1

Domain decomposition

The domain decomposition is as shown in the gure 1.2. The requirements of the numerical
simulation for the the correct reproduction of sound characteristics are very expensive. This
burden is reduced by the decomposition of the computational domain with respect to the
dierent modelling requirements. This approach allows an accuracy in simulation with an
optimized numerical approach for each of the dierent regions, despite of the fact that the
problem of sound propagation to the far-eld is usually large.

CHAPTER 1.

INTRODUCTION

Figure 1.1: Essential ingredients for numerical scheme[12]

Figure 1.2: Domain decomposition[22]

CHAPTER 1.

INTRODUCTION

1.1.1 Sound source


The ow which is responsible for various noise sources (e.g. around the fan blades, combustion chambers, turbines...) is simulated using the unsteady Navier-Stokes equations with
some turbulence-handling techniques. This type of computational uid dynamics (CFD)
simulations are very expensive, however used only for the relatively conned source regions.

1.1.2 Propagation region


For the calculation of the sound propagation, the data from this source simulation is given
as input. The sound propagationis inuenced by the steady mean ow and its interaction
with soft or hard walls. Using disturbance regions, these eects can be modelled in a
linearised manner. The acoustic scales are much larger, due to their propagation at the
speed of sound relative to the medium. This approach allows the simulation of sound
propagation reducing number of grid points in a very impressive manner by neglecting the
inuence of slowly-propagating hydrodynamic disturbances. This makes the simulation of
sound propagation with an optimized CAA method much more ecient, and therefore less
expensive than the CFD methods used in the source region.

1.1.3 Far-eld
When the far-eld noise levels are considered as more important, then the sound eld in
the intake and in the vicinity of the engine is not sucient for most applications. By
assuming uniform mean ow and neglecting wall eects on the sound propagation, the
sound radiation to the far-eld is modeled by the aeroacoustic analogy of Ffowcs Williams
& Hawkings. The solution by such an integral method is less expensive than CAA and
produces no amplitude decay. The basis for correct prediction at a given observer position
is nonetheless a correct sound eld in the intake and in the near eld of the engine. Due
to the decomposition of the domain, the total error is an aggregation of all the errors
produced by each of the methods. Nevertheless, the capabilities and advantages of each
methods have been utilized to full capacity.
1.2

Aero-Acoustical Analogy

The Aero-Acoustical Analogy is an ecient hybrid method for the computation of owinduced noise :

Generation of ow-induced noise : ow computations by means of CFD


Denition of aero-acoustical sources
Propagation of ow-induced noise : computation of sound pressure levels by means
of computational acoustics
The most common approach for computational aero-acoustics is the Lighthill's analogy. In
this analogy, rearrangement of Navier-Stokes equations for incompressible ow is done and
thereby making a connection between uid mechanics and acoustics. The following gure
gives a general strategy of all the methodologies used for the computation of sound.

CHAPTER 1.

INTRODUCTION

Figure 1.3: Process Overview[4]


The process in Fig. 1.3 shows the minimization of the numerical errors associated with
developing the acoustic sources from the ow computation. In the next steps the Lighthill's
equation is derived and the limitations of this analogy are discussed.

Chapter 2
Lighthill's Analogy

The compressible Navier-Stokes equations are reformulated to derive a wave equation with
a source term, which includes a pressure and density contribution by Lighthill. According to
this analogy, the noise due to an unsteady ow is equivalent to the noise, that is generated
by equivalent sources radiating in a medium at rest. These sources are given by Lighthill
stress tensor.
Curle extended Lighthill's integral formulation to handle solid boundaries upon aerodynamic sound. Therefore Curle solved the Lighthill equation when hard surfaces are taken
into consideration. According to Curle, the noise due to a ow passing by a body is equivalent to the noise that is generated by dipole sources on the surface of the body and
quadrupole sources in the volume. These dipole sources are given by the hydro-dynamic
pressure.[5]
Ffowcs-Williams and Hawkings made an approach, extending the Lighthill and Curle integrals to take into account the interaction of the ow with rotating surfaces.[21] They
reformulated the Navier-Stokes equations introducing mathematical surfaces, that coincide
with the surfaces of the moving solid, and imposing boundary conditions on it. According
to their model the noise, as a result of ow interacting with the rotating surface, is due to
three kinds of sources:

Monopoles: thickness noise


Dipoles: loading noise
Quadrupoles: vortex noise
Therefore the situation of an interacting ow with a rotating surface is equivalent to an
acoustic medium at rest containing three source distributions:
2 0
2 ij
2 0
2

a
=
s
t2
xi xj
x2i

(2.1)

where0 denotes the uctuating density of the uid,ij is the stress tensor introduced
by Lighthill, as is the speed of sound

CHAPTER 2.

2.1

LIGHTHILL'S ANALOGY

Derivation of Lighthill's equation

To obtain Lighthill's equation, the rearrangement of Navier-Stokes equations is done. Then


a connection is made between uid mechanics and acoustics. The derivation of Lighthill's
equation is as follows. The rst equation of interest is conservation of mass,

vi
+
=0
t
xi

(2.2)

where is the density and v is the i th component of ow velocity vector v .


Next equation is conservation of momentum,

ij
vi
p

vi
=

=
(pij + ij )
+ vj
t
xj
xi
xj
xj

where p is the total pressure within the ow,ij


Kronecker symbol.
For Newtonian uids [ ] can be expressed as


vj
vi
ij =
+
+
xj
xi

(2.3)

is the viscous stress tensor, is the

vk
2
ij
3
xk

(2.4)

where is the dynamic viscosity


Now multiplying 2.2 by v and add it to 2.3 and in addition substituting p by p p0 , we
have

vj
vi
i
vi

+ vi
+ vj
+ vi
=
((p p0 )ij ij )
t
t
xj
xj
xj
 
Now we introduce momentum ux tensor I
ijI = vi vj + (p p0 )ij ij

(2.5)

2.5 can be rewritten as

ijI
vi
=
t
xj

(2.6)

For linear acoustics the following relation holds

iju = (p p0 )ij
2.6 can be rewritten as

vi
+ (p p0 ) = 0
t
t
For linear acoustics the following properties are fullled
p0 = p p0
0 = 0

(2.7)

CHAPTER 2.

LIGHTHILL'S ANALOGY

p0
= c2
0
where p0 and 0 are the acoustic pressure and acoustic density respectively and c is the
speed of sound.
Substituting p p0 by c2 0 in 2.7 and applying /xi to this equation we obtain


2
(vi ) + 2 (c2 0 ) = 0
t xi
xi
Using conservation of mass we can rewrite 2.8 as


2 2 0
2
2 0
1 2
2 (c ) =

(c2 0 ) = 0
t2
c2 t2 x2i
xi

(2.8)

(2.9)

Since there is no ow and no other excitation elds, no acoustic sound eld will be generated, since the solution of 2.9 is 0 = 0 = 0

Lij = ijI ij0


Lij = vi vj + ((p p0 ) c2 ( 0 ))ij ij

(2.10)

where Lij is the Lighthill's turbulence stress tensor.


Equation 2.6 is rewritten as momentum equation for ideal, linear acoustic medium subjected to external stress according to 2.10

(ijI ij0 )
vi ij
+
=
t
xi
xi
Lij
vi
2
+
(c ( 0 )) =
t
xi
xi
Eliminating the momentum density vi as in linear case from 2.11, results in


2 Lij
2
1 2
2

(c
(

))
=

0
c2 t2 x2i
xi xj

(2.11)

(2.12)

This equation is called Lighthill's inhomogenous wave equation.


2.2

Approximation of Lighthill's stress tensor

In order to compute the noise radiation, the important aspect of aeroacoustic analogy is the
approximation of Lighthill's stress tensor. The leftside of equation gives the propagation
terms and the rightside of the equation gives the noise source terms. These source terms are
assumed to be vanished outside the turbulent region. In this outer region, the velocity vi
consists small motion characteristics of sound only and this velocity appears quadratically
in the calculation of the tensor as vi vj . The eects of viscosity and heat conduction cause
only a slow damping due to the conversion of acoustic energy into heat. Therefore, eij is

CHAPTER 2.

LIGHTHILL'S ANALOGY

neglected. For isotropic media, the term ((p p0 ) c2 ( 0 )) is very small and can be
neglected. Therefore, the resulting approximate of Lighthill's tensor is given by

Lij vi vj

Chapter 3
Limitations in Lighthill's analogy

The most prevalent theory of aeroacoutics is the Lighthill's aeroacoustic analogy. In this
chapter some of the weaknesses in the formulation of this acoustic analogies is presented.
In Lighthill's integral formulation, Curle and other integral formulations, Green functions
are used to solve the propagation equation. Curle formulation is an extension made to
Lighthill's general theory of aerodynamic sound, so as to incorporate the inuence of solid
boundaries upon the sound eld. This inuence is twofold, namely

Reexion and diraction of the sound waves at the solid boundaries, and
A resultant dipole eld at the solid boundaries which are the limits of Lighthill's
quadrupole distribution. [5]
Ffowcs Williams & Hawkings extended the Lighthill-Curle theory of aerodynamic sound
by including arbitrary convective motion. When both the bounding surfaces and the
turbulence are compact relative to the radiated length scales, the turbulence is acoustically
equivalent to a volume distribution of moving quadrupoles and the surfaces to dipole and
monopole distributions[21]. The main hindrance of all these integral explicit formulations
is that they use free eld Green functions for solving the propagation equation. This makes
the resolution valid only for external propagation problems, where the free eld assumption
is acceptable. In most of the industrial problems, this assumption is not valid because the
presence of ducts, casings, walls etc perturbs a lot the acoustic waves.
One of the most important characteristic in acoustic analogy approach is that the
formulation possess some in-built non-uniqueness. One of them is the choice of variable.
In Lighthill's analogy density is chosen as the variable. Some researchers who followed
this analogy have chosen pressure p, while some others preferred some other quantities.
Another non-uniqueness is the choice of propagation operator. Lighthill's analogy was
made such that, the sound is generated in a static quiescent environment. In this analogy,
a simple wave propagation operator is chosen as the propagation part of the governing
equation. In this analogy, the base ow is considered as uniform. In the presence of
non-uniform mean ow, refraction eects on the propagation of sound due to velocity or
density gradients, which cannot be neglected. Realizing the importance, Lilley proposed
a new method with the integration of the refraction eect on the wave propagation and
thereby deriving Lilley's equation. Lilley's equation is as follows:
Lo =
(3.1)
9

CHAPTER 3.

Do Do2
2
U 2
Dt Dt2 xi c xi + 2 xi x1 c xi
U x 1 is the convective derivative based

where Lo
Do
Dt

10

LIMITATIONS IN LIGHTHILL'S ANALOGY

= t
+
1
ln pp0 &

on the mean ow velocity and

c2 = RTo
When the acoustic analogy is carried out completely by considering mean refraction
and nonlinear steepening eects, the governing equation takes of the form
Q E F
G
E
F
G
+
+
+
=
+
+
t
x
y
z
x
y
z

(3.2)

where

Q=

F =

, E =

u
2 + p

(E + p)

E =

F =

G=

u
u2 + p
u
u
(E + p)u

, G =

2 + p
(E + p)

0
xx
xy
xz
uxx + xy + xz qx

0
yx
yy
yz
uyx + yy + yz qy

0
zx
zy
zz
uzx + zy + zz qz

(qx ,qy ,qz ) are the components of the heat ux vector. The left side of 3.2 is the Euler
equation and the terms on the right side are the viscous terms. Now the acoustic analogy
arguement ends up with only viscous noise source terms. This is denitely erroneos and
thus the theory will fail to identify any meaningful acoustic sources.
Several examples are provided by Tam to show clearly the limitations and failures of
the acoustic analogy theories when used to identify the correct noise sources. The examples

CHAPTER 3.

11

LIMITATIONS IN LIGHTHILL'S ANALOGY

involve initial value problem and boundary value problem. By these examples quadrupoles,
the source terms of the Lighthill's equation are ctitious noise source terms. By boundary
value problem the real source of noise is the thermal uctuation at the surface of the
sphere[19]. They show that the quadrupoles are not the true sources of sound. Another
shows that the weak solutions of Euler or Navier-Stokes equation not necessarily satify
the Lighthill's equation. The acoustic analogy theory is derived by dierentiating the
set of rst order dierential equations (continuity and momentum equations) to forma
a second order nonhomogenous wave equation. There are some fundamental dierences
in the mathematical structure of Euler and Lighthill's equations. The solution of rst
order partial dierential equation may not be identical to that of higher order equation.
A normal shock propagation solution of Euler's equation cannot be retained by solving
the Lighthill's equation with quadrupole terms and sources. When the propagation of a
normal shock wave into a stationary gas in one-dimension is considered, by well-known
Rankine-hugoniot relations we have


2Ms2 + 1
p2 = p1
(3.3)
+1


( + 1)Ms2
2 = 1
(3.4)
( 1)Ms2 + 2


2(Ms2 1)
u2 = cs
(3.5)
( + 1)Ms2
where cs denotes the shock speed, 1 ,p1 & a0 denote the gas density, pressure and sound
speed ahead of the shock and 2 ,p2 & u2 denote the gas density, pressure and sound speed
behind the shock respectively, Ms = cs/a0 is the shock Mach number
The general solution of 2.11 in the region x > a0 t satisfying the initial condition, t = 0

(x, 0) = f (x)

t (x, 0)

= g(x)

(3.6)

For the normal shock propagation, f = 1 and g = 0. The solution is obtained as




1 , x > cs t
(x, t) =
(3.7)
2 , a0 t < x < cs t
where



1
2p1 (Ms2 1) 2a20 p1 (Ms2 1)
41 (Ms2 1)
1
2 = 1 + 2

+1
( 1)Ms2 + 2 ( 1)Ms2 + 2 ( 1)Ms2
a0 c2s
From 3.4 and 3.7, it is clear that acoustic analogy is unable to reproduce the weak solution of the Euler's solution. Another example shows that the acoustic analogy theories
could provide misleading physics. When the instability wave noise generation problem is
computed through acoustic analogy approach, there are no real noise source terms. By
following this approach one inevitably conclude that the quadrupoles are the only noise
sources and could never discover other forms of noise sources including instability waves or
turbulence structures. Thus this example not only makes clear that the acoustic analogy

CHAPTER 3.

LIMITATIONS IN LIGHTHILL'S ANALOGY

12

do not contain correct physics but also can give the misleading information about the true
noise source terms of jet noise.
Outside the source zone, the wave operator of Lighthill's analogy is restricted to constant ow conditions. There is no variation of density, speed of sound and Mach number
in this analogy. Once an acoustic wave passes the dierent mean ow conditions, they
are identied as strong sources with opposite sign by the analogy. Part of the acoustic
wave is removed by one source and a new wave is radiated to x the dierent wave speed.
This leads to very large volumes with strong sources. Several modications to Lighthill's
original theory have been proposed to account for the sound-ow interaction. To improve Lighthill's analogy dierent quantities inside the wave operator as well as dierent
wave operators are considered by following analogies. All of them obtain modied source
terms, which sometimes allow a clear sight on the real sources. The acoustic analogies of
Lilley[13], Pierce[15], Howe[9] and Mhring[14] are only some examples for aeroacoustic
analogies based on Lighthill's ideas. All acoustic analogies require a volume integration
over a source term. The major diculty with the acoustic analogy, however, is that the
sound source is not compact in supersonic ow. Errors could be encountered in calculating
the sound eld, unless the computational domain could be extended in the downstream
direction beyond the location where the sound source has completely decayed. Furthermore, an accurate account of the retarded time-eect requires keeping a long record of
the time-history of the converged solutions of the sound source, which again represents
a storage problem. For realistic problems, the required storage can reach the order of 1
terabyte of data.

Bibliography

[1] R.J. Astley. Numerical methods for noise propagation in moving ows, with application to turbofan engines. Acoustical science and technology, 30(4):227239, 2009.
[2] M. Cabana, V. Fortun, and P. Jordan. Identifying the radiating core of lighthill's
source term. Theoretical and Computational Fluid Dynamics, 22(2):87106, 2008.
[3] M. Cabana, V. Fortun, P. Jordan, et al. A look inside the lighthill source term. &
Proceedings, 2006.
[4] S. Caro, R. Sandboge, J. Iyer, and Y. Nishio. Presentation of a caa formulation based
on lighthill's analogy for fan noise. In Fan Noise 2007 Conference, pages 1719, 2007.
[5] N. Curle and N. Curle. The inuence of solid boundaries upon aerodynamic sound.
Proceedings of the Royal Society of London. Series A. Mathematical and Physical
Sciences, 231(1187):505514, 1955.
[6] M. Escobar. Finite element simulation of ow-induced noise using lighthill's acoustic
analogy. Dr.-Ing. Thesis, 2007.
[7] M.E. Goldstein.
An exact form of lilley's equation with a velocity
quadrupole/temperature dipole source term. Journal of Fluid Mechanics, 443(1):231
236, 2001.
[8] JC Hardin and DS Pope. An acoustic/viscous splitting technique for computational
aeroacoustics. Theoretical and Computational Fluid Dynamics, 6(5):323340, 1994.
[9] MS Howe. Contributions to the theory of aerodynamic sound, with application to
excess jet noise and the theory of the ute. Journal of Fluid Mechanics, 71(04):625
673, 1975.
[10] M. Kaltenbacher. Numerical simulation of mechatronic sensors and actuators.
Springer Verlag, 2007.
[11] M. Kaltenbacher, M. Escobar, S. Becker, and I. Ali. Numerical simulation of owinduced noise using les/sas and lighthill's acoustic analogy. International journal for
numerical methods in uids, 63(9):11031122, 2010.
[12] LMS Virtual Lab. Flow-Induced Noise Simulation using LMS Virtual.Lab AeroAcoustics.
https://www.dropbox.com/s/cj8ips08dnihz62/Flow%20Induced%20Noise%
20Simulation.pdf/.
13

BIBLIOGRAPHY

14

[13] GM Lilley. On the noise from air jets. Aeronautical Research Council, Noise Research
Committee, 1948.
[14] W. Mohring. On vortex sound at low mach number. J. Fluid Mech, 85(4):685691,
1978.
[15] A.D. Pierce. Wave equation for sound in uids with unsteady inhomogeneous ow.
The Journal of the Acoustical Society of America, 87:2292, 1990.
[16] S. Seidl. Numerical simulation of noise generation and propagation in turbo machinery.
[17] W.Z. Shen and J.N. Srensen. Aeroacoustic modelling of low-speed ows. Theoretical
and computational uid dynamics, 13(4):271289, 1999.
[18] S.A. Slimon, MC Soteriou, and DW Davis. Computational aeroacoustics simulations
using the expansion about incompressible ow approach. AIAA journal, 37(4):409
416, 1999.
[19] C. Tam. Further consideration of the limitations and validity of the acoustic analogy
theory. 2002.
[20] C.K.W. Tam. On the failure of the acoustic analogy theory to identify the correct
noise sources. In Proc. 7th AIAA/CEAS Aeroac. Conf., Maastricht, 2830 May 2001,
2001.
[21] J.E.F. Williams and D.L. Hawkings. Sound generation by turbulence and surfaces in
arbitrary motion. Philosophical Transactions of the Royal Society of London. Series
A, Mathematical and Physical Sciences, 264(1151):321342, 1969.
[22] C. Richter L. Panek X. Li, N. Schnwald.
The CAA Methods in use
at the HFI. http://www.cfd.tu-berlin.de/index.php?sec=research&subsec=
acoustics&subsubsec=overview&lang=english/.

Das könnte Ihnen auch gefallen