See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264048650

FWH integral in the frequency domain for arbitrary flow Mach numbers

Conference Paper · September 2010

CITATIONS READS
0 494

2 authors:

Ulf Michel Christian Weckmueller

276 PUBLICATIONS 2,242 CITATIONS

German Aerospace Center (DLR)
16 PUBLICATIONS 160 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Ulf Michel on 21 July 2014.

The user has requested enhancement of the downloaded file.

 FWH integral in the frequency domain for arbitrary
flow Mach numbers
Ulf Michel ∗and Christian Weckmüller †

A far-field solution for the Ffowcs-Williams and Hawkings integral is derived based on
the convective wave equation, which includes interference effects.

Nomenclature
c0 sound speed in ambiance M0 = U0 /c0 , Mach number
Df = 1 − M0 cos θe , Doppler factor U0 uniform flow speed in ambiance
f frequency Ui = (U0 , 0, 0)
xi observer position vi velocity of fluid
yi position of source element ρ density
r = |yi − xi | ρ0 density in ambiance
re emission (wave-normal) distance β = (1 − M02 )1/2
S data surface enclosing volume V θe emission (wave-normal) angle
t time
tr retarded time

I. Introduction
The numerical simulation of the sound radiation of aerodynamic sources is generally performed in two
steps. In a first step the unsteady flow field is simulated inside a limited computational domain V with a
suitable CFD-method. The sound radiation into arbitrary positions outside V is then computed in a second
step by solving an integral over the outer surface S of V . The numerical results of the unsteady flow field
on the surface S obtained in the first step are used as input for the second step. Two different integrals can
be used in the second step, the Kirchhoff integral or the Ffwocs-Williams & Hawkings integral.

I.A. Kirchhoff integral

The Kirchhoff integral for the unsteady sound pressure in a stationary field position xi , for a stationary
surface S, and for a zero mean flow velocity outside S is given by1
Z  
1 ρ0 ∂vn 1 ∂r 1 ∂r ∂p
p(xi , t) = − 2 p− dS(yi ) . (1)
4π r ∂t r ∂n c0 r ∂n ∂t tr
S

The brackets indicate evaluation of the enclosed terms at the retarded time tr = t−r/c0 , where r = |yi −xi |
is the distance between the position yi of the surface element and the observer position xi outside S. c0 is
the speed of sound and ρ0 the mean density in the ambiance. The sound pressure p(xi , t) in the field point xi
at the time t is determined by the time derivative of the normal component ∂vn /∂t of the acoustic particle
velocity (celerity), the sound pressure p and its time derivative ∂p/∂t on the surface S. It can be concluded
that the sound field outside the surface can be described by a distribution of sources on the surface. The
first term in the brackets describes monopole sources, the second and third terms are dipole sources, whose
directivities are determined by ∂r/∂n. The second term is a near field term because it decays with the
∗ Institute of Fluid Mechanics and Engineering Acoustics, Computational Fluid Dynamics and Aeroacoustics Group, TU

Berlin, Berlin, Germany

† German Aerospace Center, Institute of Propulsion Technology, Engine Acoustics Dept., Berlin

1 of 6

American Institute of Aeronautics and Astronautics

square of the distance r between the surface element and the far field point. p and v have to satisfy the
wave equation on the surface S, which means that the surface must be sufficiently far away from the actual
aerodynamic sources inside V .
By inserting the linearized Euler equation
∂vn 1 ∂p
=− (2)
∂t ρ0 ∂n
equation (1) can be rewritten as
Z  
1 1 ∂p 1 ∂r 1 ∂r ∂p
p(xi , t) = − + 2 p+ dS(yi ) , (3)
4π r ∂n r ∂n c0 r ∂n ∂t
S

which means that the sound pressure in any observer position xi outside the surface S can be described in
terms of the pressure fluctuation p and its surface-normal and time derivatives on the surface S. However,
use of equation (1) is preferable over equation (3) because the latter contains a spatial derivative, which may
not be easily available from the CFD simulation performed in the first step. An additional reason for using
equation (1) is, that time derivatives are easy to treat when the integral is transformed into the frequency
domain.
The Kirchhoff integral can also be derived for a uniform mean velocity outside the surface. The corre-
sponding flow Mach number (flight Mach number) is M0 . The derivation is based on the convective wave
equation. The result for the aeroacoustic application was presented by Morino.2 A uniformly moving Kirch-
hoff surface was applied to jet noise predictions by Lyrintzis & Makbadi.3 Arbitrarily moving surfaces were
discussed by Lyrintzis.4 The integral in terms of the fluctuating pressure for a uniform external motion
Ui = (U0 , 0, 0) is given by
Z    
1 1 ∂p 1 ∂rg 1 ∂r ∂yg,1 ∂p
p(xi , t) = − + 2 p+ − M dSg (yi ) . (4)
4π rg ∂ng rg ∂ng c0 rg β 2 ∂ng ∂ng ∂t
Sg

The index g indicates Prandtl-Glauert coordinates. ng is the outward pointing vector on the surface element
dSg . p
rg = (x1 − y1 )2 + β 2 [(x2 − y2 )2 + (x3 − y3 )2 ] (5)
β 2 = 1 − M02 (6)
yg,1 = y1 , yg,2 = βy2 , yg,3 = βy3 (7)
The brackets around the integrand indicate evaluation at the retarded time tr , which is defined by
rg − M (x1 − y1 )
tr = t − . (8)
cβ 2

I.B. Ffowcs-Williams & Hawkings integral

One disadvantage of the Kirchhoff integral is that the (linear) wave equation must be satisfied on and outside
the integration surface S. This also requires that the mean flow velocity on the surface is identical to the
uniform mean flow velocity of the ambiance. In order to meet this requirement the integration surface
has to be defined relatively far away from the sources with the result that the computational domain of
the unsteady flow simulation has to be relatively large. This is why another integral solution, the Ffowcs-
Williams & Hawkings (FW-H) integral is preferred in many cases over the Kirchhoff integral, because the
FW-H integral is less restrictive with respect to a linearity of the unsteady flow field and allows the surface
to be located closer to the sources.
If it is assumed that all sources are enclosed by the surface S, the FW-H integral in terms of the pressure
fluctuation p0 is given for observer positions in the acoustic fa field outside S by
1 ∂2
Z  
qq
4πp0 (xi , t) = 2 2 dV (yi )
c0 ∂t r |1 − Mr |
V (t)
Z   Z  
1 ∂ Fr ∂ ρ0 Un
+ dS(yi ) + dS(yi ) (9)
c0 ∂t r |1 − Mr | ∂t r |1 − Mr |
S(t) S(t)

2 of 6

American Institute of Aeronautics and Astronautics

qq is the quadrupole source term.

p0
   
ρ0
qq (yi , θe , t) = ρ0 vr2 1+ − 1− p0 , (10)
ρ0 c20 ρ

where vr is the component of the fluctuating velocity in the direction of the observer. Fr is the component
of the surface force per unit area in the direction of the observer. Un is the component of the velocity vector
of the surface that is normal to the surface element dS.
Equation (9) simplifies for a stationary surface S with Ui = 0.

1 ∂2
Z h i Z   Z  
0 qq 1 ∂ Fr ∂ ρ0 Un
4πp (xi , t) = 2 2 dV (yi ) + dS(yi ) + dS(yi ) (11)
c0 ∂t r c0 ∂t r ∂t r
V S S

II. Solutions in emission coordinates

II.A. Emission coordinates
It was shown by Michalke & Michel7 that the solution of the Lighthill equation for a quiescent ambiance can
be transformed into the solution for an ambiance with uniform motion by replacing the geometric angle θ by
the emission angle θe , the geometric distance r by the emission distance re , and by replacing the velocities
in the source terms by the velocity relative to the flight speed Uf .
The emission coordinates can be derived from the Green function for the convective wave equation for
an unbounded field.7 The geometrical situation is illustrated in figure 1 for a situation in which the source
position yi and the observer position xi are stationary.

Figure 1. Relation between the source position yi and the observer position xi in a uniform flow (wind-tunnel coordinate
system). re and θe are the emission (or wave-normal) coordinates, r and θ are the geometric or observer coordinates.

The emission or wave-normal distance re is the distance the sound wave has to propagate relative to the
moving fluid from the source element dV (yi ) to the observer position xi . During the propagation in the
direction of the emission (or wave-normal) angle θe over a distance of re the sound wave drifts downstream
by the distance re M0 , where M0 = Uf /c is the Mach number of the external flow (the flight Mach number
of the aircraft). θe = 0 points upstream into the flight direction.
If the coordinate system is chosen such that the velocity vector in the ambiance points into the x1 -
direction, Ui = (Uf , 0, 0), the emission (or wave-normal) distance re and the emission angle θe are related to
the geometric distance r and the geometric angle θ between source and observer as follows.
r
re = q (12)
1 − M02 sin2 θ − M0 cos θ

and q 
2 2
cos θe = cos θ 1 − M0 sin θ − M0 cos θ + M0 . (13)

The inverse relations are q

r = re 1 − 2M0 cos θe + M02 (14)

3 of 6

American Institute of Aeronautics and Astronautics

and
cos θe −M0
cos θ = p . (15)
1 − 2M0 cos θe + M02

II.A.1. Geometric relations in cartesian coordinates

Generally the geometric relations are required in cartesian coordinates. The separation vector between source
at position yi and observer in position xi is defined by

∆xi = xi − yi . (16)

The wave-normal distance re is given by

p
∆x21 + (1 − M02 )(∆x22 + ∆x23 ) − M0 ∆x1
re = . (17)
1 − M02

In addition, the definition of the Doppler factor Df will be needed, which is

Df = 1 − M0 cos θe . (18)

The product re Df of the wave-normal distance and the Doppler factor is given by
q
re Df = rg = ∆x21 + (1 − M02 )(∆x22 + ∆x23 ) . (19)

where rg is the Prandtl-Glauert coordinate defined by equation (5).

II.B. Green function for convective wave equation

The convective wave equation is given by
2
∂2p

1 ∂ ∂
2 + U i p+ =q (20)
c0 ∂t ∂xi ∂x2i

with the source term q. The integral solution of (20) is given by the Green function, which is the solution
for the point source
q = f (t)δ(xi − yi ) (21)
The Green function in terms of the emission coordinates is (see, e.g. Michalke and Michel7 )
1
p(xi , t) = f (t − re /c0 ) (22)
4πre Df

with the Doppler factor Df defined by equation (18).

II.C. Kirchhoff integral in emission coordinates

Z  
1 ρ0 ∂ve,n 1 ∂re 1 ∂re ∂p
p(xi , t) = − 2 p− dS(yi ) . (23)
4π re Df ∂t re Df ∂n c0 re Df ∂n ∂t
S

For comparison:
Z    
1 1 ∂p 1 ∂rg 1 ∂r ∂y0,1 ∂p
p(xi , t) = − + 2 p+ −M dS0 (yi ) . (24)
4π rg ∂n0 rg ∂n0 crg β 2 ∂n0 ∂n0 ∂t
S0

4 of 6

American Institute of Aeronautics and Astronautics

II.D. FW-H integral in emission coordinates
Similarly to equation (4) the FW-H equation can be derived for a uniform motion outside the surface S,
while the surface S, yi , and xi remain stationary.8 The derivation is based on the convective wave equation
(20).9 Without loss of generality the system of coordinates is chosen such that x1 and y1 are aligned with U1 ,
thus Ui = (U0 , 0, 0). The time-domain integral solution for the acoustic far field, using the Green function
for homogeneous flow, equation (22), is:

ρ0 Ui u0j Γ1j + (ρ0 u0i + ρ0 Ui ) (c0 − Uj Γ1j ) + p0 Γ1i − τij Γ1j

Z  
∂
4πp0 (xi , t) = ni dS(yi ) (25)
∂t c0 re Df
S

Γ1i is the spatial derivative of the emission radius re :

 
∂re /∂x1
Γ1i =  ∂re /∂x2  (26)
 

∂re /∂x3

By introducing emission coordinates into equation (11) we arrive at

Z " #
0 1 ∂2 qqe
4πp (xi , t) = 2 2 dV (yi )
c0 ∂t re Df2
V
Z   Z  
1 ∂ Fre ∂ ρ0 Une
+ dS(yi ) + dS(yi ) (27)
c0 ∂t re Df ∂t re Df
S S

Fourier transform into frequency domain

ω2
Z
q̂qe ikre
p̂(xi , ω) = − 2 e dV (yi )
4πc0 re Df2
V
Z Z
ω F̂re ikre ω ρ0 Ûne ikre
− e dS(yi ) − e dS(yi ) (28)
4πc0 re Df 4π re Df
S S

Abbreviation for the surface integral

Z
p̂(xi , ω) = Â(yi , xi , ω)dS(yi ) (29)
S

If the data surface is axially symmetric, the integrand  can be expanded into an azimuthal Fourier
series. X
Â(y1 , θ) = Âm (y1 )eimψ (30)
m

The integral (29) can then be described as sum of azimuthal components.

X
p̂(xi , ω) = p̂m (xi , ω) (31)
m
Z
p̂m (xi , ω) = Âm (y1 , xi , ω)ds (32)
S

The surface integral (29) is now replaced by a line integral.

FW-H equation derived for the convective wave equation is applied to calculate the sound radiation from
a permeable data surface to the far field. In this formulation, the observer remains stationary with respect
to the permeable data surface for arbitrary flight Mach numbers. The main advantage over the standard
approach based on a moving data surface in a stationary fluid is the following one: periodic fluctuations
remain periodic with the same frequency at the observer position. Thus a frequency domain solution can

5 of 6

American Institute of Aeronautics and Astronautics

 be easily derived which is suitable for the study of open rotor noise. For axisymmetric configurations, the
solution can be decomposed into circumferential modes in order to drastically reduce the computational
speed and also the quantity of data to be save. The same idea is found in 2.5D CAA10 therefore the two
methods can be directly coupled to calculate acoustic radiation in far field.
Since in the moving medium case, there is no relative speed between the data surface and the observer,
no Doppler frequency shift occurs. Therefore the solution can be reformulated in the frequency domain. The
frequency formulation is typically of one order of magnitude faster than the time-domain formulation, when
applied to tonal problems, and of higher accuracy since the time derivation as well as the interpolation to
emission time are exact in the frequency domain.11 Another modification proposed here is to decompose the
surface data into circumferential modal components with the objective to transform the surface integral into
a line integral. This enables the use of standard integration techniques and also helps to reduce the quantity
of CFD data to save.

III. Conclusion
References
1 Ehrenfried,
K., Strömungsakustik , Mensch & Buch Verlag Berlin, 2004.
2 Morino,
L., “Mathematical foundations of integral methods,” Computational Methods in Potential Aerodynamics, edited
by L. Morino, Springer-Verlag,New York, 1985, pp. 271–291.
3 Lyrintzis, A. S. and Mankbadi, R. R., “On the prediction of far-field jet noise using Kirchhoff’s formulation,” University of

Minnesota Supercomputer Institute Research Report UMSI 94/241, December 1994, Paper number AIAA 1995-0508, presented
at the 33rd AIAA Aerospace Science Meeting, Reno, NV, Jan. 1995.
4 Lyrintzis, A. S., “The use of Kirchhoff’s method in rotorcraft aeroacoustics,” University of Minnesota Supercomputer

Institute Research Report UMSI 94/198, October 1994, Presented at the AGARD 75th Fluid Dynamics Panel Meeting and
Symposium onAerodynamics and Aeroacoustics of Rotorcraft, Berlin, Germany, Oct. 1994.
5 Lyrintzis, A. S., “Surface integral methods in computational aeroacoustics - From the (CFD) near-field to the (Acoustic)

far-field,” International Journal of Aeroacoustics, Vol. 2, 2003, pp. 95–128.

6 Di Francescantonio, P., “A new boundary integral formulation for the prediction of sound radiation,” J. Sound Vib.,

Vol. 202, No. 4, 1997, pp. 491–509.

7 Michalke, A. and Michel, U., “Prediction of Jet-Noise in Flight from Static Tests,” J. Sound Vib., Vol. 67, 1979, pp. 341–

367.
8 Weckmüller, C., Wellner, J., Guerin, S., Michel, U., and Schnell, R., “Ffowcs Williams & Hawkings formulation for

the convective wave equation and permeable data surface,” AIAA-2010-3710, June 2010, 16th AIAA/CEAS Aeroacoustics
Conference, 7-9 June 2010, Stockholm, Sweden.
9 Wellner, J., Herleitung und Anwendung der konvektiven Kirchhoff-Ffowcs-Williams-Hawkings-Gleichung, Diploma the-

sis, Technische Universität Berlin, Müller-Breslau-Str. 8, 10623 Berlin, Germany, May 2009.
10 Li, X. D., Schemel, C., Michel, U., and Thiele, F., “On the azimuthal mode propagation in axisymmetric flow ducts,”

AIAA J., Vol. 42, 2004, pp. 2009–2027.

11 Shur, M., Spalart, P., and Strelets, M., “Noise prediction for increasingly complex jets. Part I: Method and tests,”

International Journal of Aeroacoustics, Vol. 4, 2005, pp. 213–246.

6 of 6

American Institute of Aeronautics and Astronautics

View publication stats