Integral Metho ds in Computational
Aeroacoustics
-From the CFD Near-Field to the Acoustic
Far-Field
y
Anastasios S. Lyrintzis
Scho ol of Aeronautics and Astronautics
Purdue University
W. Lafayette, IN 47907-2023
Abstract
A review of recent advances in the use of integral metho ds in Computational AeroAcoustics
CAA for the extension of near- eld CFD results to the acoustic far- eld is given. These
integral formulations i.e. Kirchho 's metho d, p ermeable p orous surface Ffowcs-Williams
Hawkings FW-H equation allow the radiating sound to b e evaluated based on quantities on
an arbitrary control surface if the wave equation is assumed outside. Thus only surface integrals
are needed for the calculation of the far- eld sound, instead of the volume integrals required
by the traditional acoustic analogy metho d i.e. Lighthill, rigid b o dy FW-H equation. A
numerical CFD metho d is used for the evaluation of the ow- eld solution in the near eld and
presented at the CEAS Workshop "From CFD to CAA" Athens Greece, Nov. 2002.
y
Professor, e-mail: [email protected]. 1
thus on the control surface. Di usion and disp ersion errors asso ciated with wave propagation
in the far- eld are avoided. The surface integrals and the rst derivatives needed can b e easily
evaluated from the near- eld CFD data. Both metho ds can b e extended in order to include
refraction e ects outside the control surface. The metho ds have b een applied to helicopter
noise, jet noise, prop eller noise, ducted fan noise, etc. A simple set of p ortable Kirchho /FW-
H subroutines can b e develop ed to calculate the far- eld noise from inputs supplied byany
aero dynamic near/mid- eld CFD co de.
1 Background
For an airplane or a helicopter, aero dynamic noise generated from uids is usually very imp or-
tant. There are many kinds of aero dynamic noise including turbine jet noise, impulsive noise
due to unsteady ow around wings and rotors, broadband noise due to in ow turbulence and
1
b oundary layer separated ow, etc. e.g. Lighthill . Accurate prediction of noise mechanisms
is essential in order to b e able to control or mo dify them to comply with noise regulations, i.e.
Federal Aviation Regulations FAR part 36, and achieve noise reductions. Both theoretical and
exp erimental studies are b eing conducted to understand the basic noise mechanisms. Flight-
test or wind-tunnel test programs can b e used, but in either case diculties are encounted
such as high exp ense, safety risks, and atmospheric variability,aswell as re ection problems
for wind tunnel tests. As the available computational p ower increases numerical techniques
are b ecoming more and more app ealing. Although complete noise mo dels have not yet b een
develop ed, numerical simulations with a prop er mo del are increasingly b eing employed for the
prediction of aero dynamic noise b ecause they are low cost and ecient. This research has led
to the emergence of a new eld: Computational AeroAcoustics CAA.
CAA is concerned with the prediction of the aero dynamic sound source and the transmission
of the generated sound starting from the time-dep endentgoverning equations. The full, time-
dep endent, compressible Navier-Stokes equations describ e these phenomena. Although recent
advances in Computational Fluid Dynamics CFD and in computer technology have made
rst-principles CAA plausible, direct extension of current CFD technology to CAA requires 2
addressing several technical diculties in the prediction of b oth the sound generation and its
2 3 4
transmission. A review of aerospace application of CAA metho ds was given by Long et al.
Aero dynamically generated sound is governed by a nonlinear pro cess. One class of problems
is turbulence generated noise e.g. jet noise. An accurate turbulence mo del is usually needed
in this case. A second class of problems involves impulsive noise due to moving surfaces e.g.
helicopter rotor noise, prop eller noise, fan noise etc.. In these cases an Euler/Navier Stokes
mo del or even a full p otential mo del is adequate, b ecause turbulence is not imp ortant.
Once the sound source is predicted, several approaches can b e used to describ e its prop-
agation. The obvious strategy is to extend the computational domain for the full, nonlinear
Navier-Stokes equations far enough to encompass the lo cation where the sound is to b e calcu-
lated. However, if the ob jective is to calculate the far- eld sound, this direct approach requires
prohibitive computer storage and leads to unrealistic turnaround time. The impracticalityof
5
straight CFD calculations for sup ersonic jet aeroacoustics was p ointed out by Mankbadi et al.
Furthermore, b ecause the acoustic uctuations are usually quite small ab out three orders of
magnitude less than the ow uctuations, the use of nonlinear equations whether Navier-
6
Stokes or Euler could result in errors, as p ointed out by Stoker and Smith. One usually has
no choice but to separate the computation into two domains, one describing the nonlinear gen-
eration of sound, the other describing the propagation of sound. There are several alternatives
to describing the sound propagation once the source has b een identi ed.
7
Traditional Acoustic Analogy The rst of these approaches is the acoustic analogy. In
the acoustic analogy, the governing Navier-Stokes equations are rearranged to b e in wave-typ e
form. There is some question as to which terms should b e identi ed as part of the sound
source and retained in the right-hand side of the equation and which terms should b e in the
8
left-hand side as part of the op erator e.g., Lilley . The far- eld sound pressure is then given in
terms of a volume integral over the domain containing the sound source. Several mo di cations
to Lighthill's original theory have b een prop osed to account for the sound- owinteraction
or other e ects. The ma jor diculty with the acoustic analogy,however, is that the sound 3
source is not compact in sup ersonic ows. Errors could b e encountered in calculating the
sound eld, unless the computational domain could b e extended in the downstream direction
beyond the lo cation where the sound source has completely decayed. Furthermore, an accurate
account of the retarded time-e ect requires keeping a long record of the time-history of the
converged solution of the sound source, which again represents a storage problem. The Ffowcs
9
Williams and Hawkings FW-H equation was intro duced to extend acoustic analogy in the
case of solid surfaces. However, when acoustic sources i.e., quadrup oles are present in the
ow eld a volume integration is needed. This volume integration of the quadrup ole source
term is dicult to compute and is usually neglected in most acoustic analogy co des e.g.
10
WOPWOP . Recently, there have b een some successful attempts in evaluating this term
11;12
e.g. WOPWOP+ .
Linearized Euler Equations LEE The second alternative is to use LEE in order to
13 14
extend the CFD solutions to the far- eld e.g. Lim et al. , Viswanathan and Sankar , Shih et
15
al. . The LEE equations employ a division of the ow eld into a time-averaged ow and a
time-dep endent disturbance which is assumed to b e small. The hybrid zonal approach consists
of the near- eld evaluation using an accurate CFD co de e.g. for jet noise the co de is usually
based on Large Eddy Simulations: LES and the extension of the solution to the mid- eld
using LEE. Considerable CPU savings can b e realized, since the LEE calculations are much
cheap er than the CFD calculations. This approachisvery promising, b ecause it accounts for a
variable sound velo city outside the near- eld where usually an LES mo del is applied. However,
dissipation and disp ersion errors still exist and an accurate description of propagating far- eld
waves is compromised b ecause of this. On the other hand, this metho d may b e appropriate
for the an intermediate region in some problems, outside from the reactive near- eld where the
sp eed of sound is still not constant, b efore moving to another integral metho d for the far- eld.
Kirchho Metho d Another alternative is the Kirchho metho d which assumes that the
sound transmission is governed by the simple wave equation. Kirchho 's metho d consists of 4
the calculation of the nonlinear near- and mid- eld, usually numerically, with the far- eld
solutions found from a linear Kirchho formulation evaluated on a control surface surrounding
the nonlinear- eld. The control surface is assumed to enclose all the nonlinear ow e ects and
noise sources. The sound pressure can b e obtained in terms of a surface integral of the surface
pressure and its normal and time derivatives. This approach has the p otential to overcome
some of the diculties asso ciated with the traditional acoustic analogy approach. The metho d
is simple and accurate and accounts for the nonlinear quadrup ole noise in the far- eld. Full
di raction and fo cusing e ects are included while eliminating the propagation of the reactive
near- eld.
This idea of matching b etween a nonlinear aero dynamic near- eld and a linear acoustic
16
far- eld was rst prop osed byHawkings . The use of Kirchho 's metho d has increased sub-
stantially the last 10 years, b ecause of the development of reliable CFD metho ds that can b e
used for the evaluation of the near- eld. The separation of the problem into linear and nonlin-
ear regions allows the use of the most appropriate numerical metho dology for each. Wehave
b een referring to this technique as the \Kirchho metho d." It has b een used to study various
aeroacoustic problems, such as prop eller noise, high-sp eed compressibility noise, blade-vortex
interactions, jet noise, ducted fan noise, etc. An earlier review on the use of Kirchho 's metho d
17
was given by Lyrintzis.
Porous FW-H equation A nal alternative is the use of p ermeable p orous surface FW-H
equation. The usual practice is to assume that the FW-H integration surface corresp onds to
a solid b o dy and is imp enetrable. However, if the surface is assumed to b e p orous, a general
equation can b e derived as shown in the original reference 9 and in reference 18. The p orous
surface can b e used as a control surface in a similar fashion as the Kirchho metho d explained
ab ove. Thus the pressure signal in the far- eld can b e found based on quantities on the control
surface provided by a CFD co de.
19
Farassat in a recent review article reviewed all the available FW-H and Kirchho equa-
tions for application to noise evaluation from rotating blades. The current article fo cuses only 5
on control surface metho ds i.e. Kirchho , p orous FW-H and discusses issues with their ap-
plication in various typ es of aero coustic problems including rotor noise, jet noise, ducted fan
noise, airfoil noise etc.. At rst the main formulations will b e reviewed, advantages and dis-
advantages of each metho d will b e discussed. Then we will present several algorithmic issues
and various application examples.
2 Kirchho 's Metho d Formulations
Kirchho 's metho d is an innovative approach to noise problems which takes advantage of
the mathematical similarities b etween the aeroacoustic and electro dynamic equations. The
considerable b o dy of theoretical knowledge regarding electro dynamic eld solutions can b e
utilized to arrive at the solution of dicult noise problems. Kirchho 's formula was rst
20
published in 1882 . It is an integral representation i.e. surface integral around a control
surface of the solution to the wave equation. Kirchho 's formula, although primarily used
in the theory of di raction of light and in other electromagnetic problems, it has also many
applications in studies of acoustic wave propagation.
21
The classical Kirchho formulation is limited to a stationary surface. Morgans derived a
formula for a moving control surface using Green's functions. Generalized functions can also b e
used for the derivation of an extended Kirchho formulation. A eld function is de ned to b e
identical to the real ow quantity outside a control surface S and zero inside. The discontinuities
of the eld function across the control surface S are taken as acoustic sources, represented
9
by generalized functions. Ffowcs- Williams and Hawkings derived an extended Kirchho
formulation for sound generation from a vibrating surface in arbitrary motion. However, in
their formulation the partial derivatives were taken with resp ect to the observation co ordinates
22
and time and that is dicult to use in numerical computations. Farassat and Myers derived
a Kirchho formulation for a moving, deformable, piecewise smo oth surface. The same partial
derivatives were taken with resp ect to the source co ordinates and time. Thus their formulation
is easier to use in numerical computations and their relatively simple derivation shows the 6
power of generalized function analysis.
23 27
It should b e noted that Morino and his co-workers have develop ed several formulations
for b oundary element metho ds using the Green's function approach, which are equivalentto
Kirchho formulations. Morino's formulations were derived with aero dynamic applications
in mind, so the observer is in the moving co ordinate system. However, they can b e used
for aeroacoustics, for example when b oth the control surface and the observer move with a
constant sp eed e.g., wind tunnel exp eriments, as mentioned in reference 17. Their latest
27
formulation app ears to provide an integrated b oundary element framework for Aero dynamics
and Aeroacoustics.
2.1 Farassat's Formulation
Farassat's Kirchho formulation gives the far- eld signal, due to sources contained within the
Kirchho surface. Assume the linear, homogeneous wave equation,
2 2
1 @ @
2
2 = =0 1
2 2
a @t @x @x
i i
is valid for some acoustic variable , and sound sp eed a , in the entire region outside of a closed
and b ounded smo oth surface, S .
The signal, in the stationary co ordinate system, is evaluated with a surface integral over
the control surface, S , of the dep endentvariable, its normal derivative, and its time derivative
gure 1. S is allowed to move in an arbitrary fashion. The dep endentvariable is normally
taken to b e the disturbance pressure, but can b e any quantity which satis es the linear wave
equation.
" "
Z Z
E E
1 2
4~x; t= dS + dS 2
2
r 1 M r 1 M
S S
r r
ret ret
where
i h
@
M 1
2
n
~ _ _