16th AIAA/CEAS Aeroacoustics Conference AIAA 2010-3939

Brinkman Penalization Method for Computation of Acoustic

Scattering from Complex Geometry

Youngmin Bae1 and Young J. Moon2†

Department of Mechanical Engineering, Korea University, Seoul, 136-713, Korea

In this study, Brinkman penalization method (BPM) is extended for prediction of

acoustic scattering from complex geometries. The main idea of the BPM is to model the solid
obstacle as a porous material with zero porosity and permeability. With the aim of
increasing the spatial accuracy at the immersed boundaries, computation is carried out on
the boundary-fitted Cartesian-like grid with a high-order compact scheme combined with
one-side differencing/filtering technique at the boundaries, while a slip boundary condition
at the wall is imposed by introducing the ‘anisotropic’ penalization terms to the momentum
equations. Several test cases are considered to demonstrate the accuracy, robustness and
feasibility of the BPM. Numerical results are in excellent agreement with the analytic
solutions for single and two cylinder scattering problems. The present BPM is then used to
solve the acoustic scattering from a three-element high-lift wing (30P30N model).

I. Introduction

NUMERICAL simulation of acoustic scattering from complex geometries has received attention in a wide range
of aeroacoustic problems, such as slat noise and flap side-edge noise from a multi-element airfoil in high-lift
configuration, rotor-stator interaction noise in turbo-machinery, etc. There are two main strategies in direct
simulation of acoustic scattering from solid boundaries of complexity, i.e. structured/unstructured body-fitted grid
methods [1-3] and immersed boundary methods [4-6]. In the former, implementation of wall boundary condition is
straightforward, attaining a desired degree of accuracy at the boundaries. However, when complex geometries are
concerned, the structured body-fitted grid method or even the overset grid method [7-9] often meets difficulties
associated with the grid generation as well as with the quality of the grids. A discontinuous Galerkin (DG) method
[10-12] based on the unstructured grids promises success for real complex geometries but computational cost has
always been an issue.
In this regard, an immersed boundary technique can be considered as an alternative because of its simple and
efficient implementation for arbitrarily shaped surfaces with reasonable computational cost. Following the
pioneering work of Peskin [13,14], a number of immersed boundary methods have been proposed to handle the
complex geometries [4,15,16]. Among them, a Brinkman penalization method (BPM) [17], which was originally
developed to model the fluid flow in porous media, appears attractive because of its easiness to handle the solid
obstacle by simply treating as a porous medium of high impedance. In BPM, porosity and permeability in the
penalty terms which are added in the compressible Navier-Stokes equations are set to zero in the solid region to
impose the immersed boundary effect on the fluids. A no-slip boundary condition is therefore enforced naturally at
the solid boundary. There are, however, two inherent limitations with this penalization technique. First, it requires a
large number of grid points in solid region to retain the order of accuracy at the wall, thus making the method
impractical at highly sophisticated geometries. Another drawback is that only the no-slip boundary condition is
satisfied at the solid wall, whereas a slip boundary condition has to be met with the full or linearized Euler equations.
In the present study, we address these numerical issues. With aim of increasing the spatial accuracy at the
embedded boundary, we conform the immersed boundary grids to the actual shape of the surface following the idea
of reshaped cell approach [18,19]. The slip boundary condition at the solid surface is imposed by introducing the
‘anisotropic’ penalization terms in the momentum equations. The validity of the present method is then assessed by
considering the acoustic scattering from i) a single cylinder, ii) two circular cylinders, and iii) three element high-lift
wing with the deployed slat and flap. We also discuss numerical issues related to the implementation of the reshaped
cell approach and to the stiffness due to the penalty terms.
1
Post-doctoral staff, Member of AIAA.
2†
Professor, Senior Member of AIAA, Corresponding Author (yjmoon@korea.ac.kr).
1
American Institute of Aeronautics and Astronautics

Copyright © 2010 by the Authors. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
 II. Computational Methodology
The computational methodologies are given in this section, including the governing equations for the Brinkman
penalization method, the high-order numerical schemes with the one-sided differencing/filtering techniques, the
implicit treatment of the penalty terms, and the boundary-fitted grid generation details.

A. Brinkman penalization method

In many of aeroacoustic problems, a two-dimensional acoustic field around a steady mean flow is governed by
the linearized Euler equations written as,

∂ρ ′ ∂
∂t
=−
∂x j
( ρ ′U j + ρ0u′j )
∂ui′ ∂ ∂p′ ⎛ ρ ′ ⎞ ∂U
∂t
=−
∂x j
( ui′U j ) − − ⎜ u ′j + U j ⎟ i
∂xi ⎝ ρ0 ⎠ ∂x j
, (1)

∂p′ ∂ ⎛ ∂U j ∂P ⎞
∂t
=−
∂x j
(U j p′ + γ Pu ′j ) − ( γ − 1) ⎜ p′
⎜
− u ′j ⎟+S
∂x j ⎟⎠
⎝ ∂x j

where S is the source term, ρ0, Ui, V, P represent the laminar or turbulent flow field, and ρ', ui' and p' are the
fluctuating variables non-dimensionalized by the ambient density ρ0, the speed of sound c0, and the reference
pressure ρ0c02, respectively. In the conventional body-fitted grid method, the acoustic field can be directly solved
with Eq. (1), while a slip boundary condition is explicitly imposed at the solid conformal boundaries.
In the present Brinkman penalization method, the slip boundary condition at the impermeable surface can be
imposed by adding the penalty terms into the momentum equations and modifying the continuity and energy
equations as

∂ρ ′ 1⎡ ∂ ⎤
∂t
=− ⎢
ε ⎣⎢ ∂x j
( ρ ′U j + ρ0u ′j ) ⎥
⎦⎥
∂ui′ ∂ ∂p′ ⎛ ρ ′ ⎞ ∂U
∂t
+ Rij u ′j = −
∂x j
( ui′U j ) − − ⎜ u′j + U j ⎟ i
∂xi ⎝ ρ0 ⎠ ∂x j
, (2)

∂p′ 1⎡ ∂ ⎛ ∂U j ∂P ⎞ ⎤
∂t
=− ⎢
ε ⎣⎢ ∂x j
(U j p′ + γ Pu′j ) + ( γ − 1) ⎜ p′
⎜ ∂x j
− u′j ⎟⎥ + S
∂x j ⎟⎠ ⎦⎥
⎝

where ε is the porosity defined as the ratio of the volume occupied by the fluid to the total volume of porous material,
ranging from 0 to 1. In Eq. (2), Rij is the ‘anisotropic’ permeability tensor that enforces the embedded solid surface
to satisfy the slip boundary condition and is defined as

1 1 1 1
R11 = s y nx , R12 = s y n y , R21 = − sx nx , R22 = − sx n y , (3)
K K K K

where K is the non-dimensionalized permeability of homogeneous porous medium, s=(sx, sy) and n=(nx, ny) are the
unit vectors tangential and normal to the impermeable boundaries, respectively. When s is expressed in terms of n as
(sx,sy)=(-ny,nx) in 2-D space, the permeability tensor becomes,

1 2 1 1
R11 = nx , R12 = R21 = nx n y , R22 = n y2 . (4)
K K K

2
American Institute of Aeronautics and Astronautics
 Fig. 1. Concept of Brinkman penalization method

Note that a no-slip boundary condition at the wall is easily specified by letting Rxx = Ryy = 1/K and Rxy = Ryx = 0.
The main advantage of this penalization technique is that no additional treatment is needed to impose the
boundary condition. Instead, a single set of governing equations with penalty terms is applied to the whole
computational domain, in which different porosity ε and permeability K are assigned for fluids and solids. In this
approach, the only thing needed is to calculate the wall normal vector, n at the solid surface and to define
appropriate porosity and permeability for each region (see Fig. 1). For example, porosity is equal to unity and
permeability becomes infinite in the fluid region so that Eq. (2) returns to the equations for wave propagation. On
the contrary, when porosity and permeability approach to zero for solid surfaces, the inertia terms become negligible
compared to the penalty terms in the momentum equations so that only the normal component of the velocity is
forced to be zero and the slip condition is asymptotically satisfied at the impermeable surfaces. This will be more
closely discussed in the following section.

B. Asymptotic Analysis of Brinkman Penalization Method

In order for asymptotic analysis, we consider a two-dimensional wave propagation in a fluid at rest (U=V=0). In
this case, the governing equations for the Brinkman penalization method (BPM), Eq. (2) can be expressed as

∂ρ ′ 1 ⎡ ∂u′ ∂v′ ⎤ ,
=− ⎢ + (5)
∂t ε ⎣ ∂x ∂y ⎥⎦

∂u ′ ∂p′ nx2 nx n y
=− − u′ − v′ , (6)
∂t ∂x K K
2
∂v′ ∂p′ nx n y ny
=− − u ′ − v′ , (7)
∂t ∂y K K

∂p′ 1 ⎡ ∂u′ ∂v′ ⎤ ,

=− ⎢ + (8)
∂t ε ⎣ ∂x ∂y ⎥⎦

where the hydrodynamic density, ρ0=1 and the relation of γP=ρ0c02=1 are used. Multiplying nx and ny to Eqs. (6) and
(7), respectively, and combining the resultant equations yields,

∂ ( nx u′ + n y v′ ) ⎛ ∂p′ ∂p′ ⎞ 1
+ ⎜ nx + ny ⎟ + ( nx u ′ + n y v ′ ) = 0 . (9)
∂t ⎝ ∂x ∂y ⎠ K

Eq. (9) can be re-written as,

3
American Institute of Aeronautics and Astronautics
 ∂un′ ∂p′ un′
+ + = 0, (10)
∂t ∂n K

where un'=nxu'+nyv' is the wall normal velocity and ∂p'/∂n=nx∂p'/∂x+ny∂p'/∂y is the normal derivative of pressure on
the solid surface.
Similarly, taking the dot product of a unit tangent vector s with Eqs. (6) and (7) yields,

∂us′ ∂p′
+ = 0. (11)
∂t ∂s

In Eqs. (10) and (11), it is clearly shown ‘anisotropic’ penalty terms in momentum equations are only responsible
for the change of fluctuating variables in the wall normal direction. In other words, only the normal component of
fluctuating velocity, un' tends to be zero at the embedded boundaries, because at the solid surface where K≈0, the
first two terms in Eq. (10) become almost negligible compared with the penalty term un'/K. This result strongly
suggests the slip boundary condition on the solid surface should be asymptotically satisfied

∂p′
un′ ≈ 0, ≈ 0. (12)
∂n

So, the present extension of BPM provides a simple procedure for imposing the immersed boundary effect on the
fluid without additional treatment for boundary condition.
Now, in order to verify the physical consistency of the present BPM, the asymptotic analysis is performed for
wave propagation in porous medium. When solid obstacles are modeled as homogeneous and isotropic porous
media (i.e. Rxx=Ryy=1/K and Rxy=Ryx=0), the divergence of the momentum equations yields

∂ 2 u ′ ∂ 2 v′ ∂ 2 p′ ∂ 2 p′ 1 ⎛ ∂u ′ ∂v′ ⎞ ,
+ =− 2 − 2 − ⎜ + ⎟ (13)
∂t ∂x ∂t ∂y ∂x ∂y K ⎝ ∂x ∂y ⎠

By combining Eq. (13) and Eq. (8), one can obtain the wave equation for porous media

∂ 2 p′ 1 ∂ 2 p′ 1 ∂ 2 p′ 1 ∂p′
− − + = 0, (14)
∂t 2 ε ∂x 2 ε ∂y 2 K ∂t

Considering the incident wave with amplitude of order Φ, it is evident that the first term in Eq. (14) is of order
~O(Φ), while other terms are ~O(Φ/ε) or ~O(Φ/K). So, under an assumption that the porosity and permeability
approach to zero in porous medium, Eq. (14) further simplifies to

∂p′ K ⎛ ∂ 2 p′ ∂ 2 p′ ⎞ .
≈ ⎜ 2 + 2 ⎟ (15)
∂t ε ⎝ ∂x ∂y ⎠

It is interesting to note that in porous medium, fluctuating pressure field can be represented by the diffusion equation.
This result is analogous to the asymptotic analysis of Liu and Vasilyev [17], implying that amplitude and phase
errors of the present BPM is essentially consistent with those for the compressible Navier-Stokes equations [17].
Thereby, in this study, the accuracy of the numerical solutions can be easily secured by employing sufficiently small
porosity ε and permeability K.

C. Numerical Methods
Numerical schemes are also an important issue for the efficient and robust use of Brinkman penalization method.
In order to resolve the acoustic waves with minimizing dispersion and dissipation errors, high-order numerical
schemes are most desirable in this study. However, as pointed out by Liu and Vasilyev [17], introduction of penalty
terms into the governing equations results in additional stiffness or discontinuity of fluctuating variables, thus
requiring the use of stiffly stable hyperbolic solvers.
4
American Institute of Aeronautics and Astronautics
 A numerical scheme should be able to capture the steep gradients of flow variables at the fluid/solid interface,
when applied to the simulation of aeroacoustic phenomena. With this respect, a high-order upwind scheme such as
WENO scheme [21] can be a proper scheme for this class but the implementation for the upwind scheme or the
hybrid upwind/central scheme [22] is complicated and costly as well.
In the present study, a high-order central compact scheme with one-side differencing technique is considered.
The central compact scheme [23] widely used in CAA is efficient and high-order accurate but cannot deal with the
steep gradients associated with the abrupt change of permeability, because the scheme has no numerical diffusion.
To resolve this matter, we also employ the compact spatial-filtering schemes [24], in which the numerical diffusion
can be controlled by filtering coefficient and order of accuracy. These high-order compact/filtering scheme
combined with one-side differencing/filtering technique produce sufficient numerical diffusion in the region where
the steep gradients exist, so they are effective and easy to apply to the present governing equations, as compared to
other high-order upwind schemes [21,22].
At interior points, the sixth order compact finite difference scheme for first derivative is given by,

f i +1 − fi −1 f −f
α f i′−1 + f i′+ α fi ′+1 = β1 + β 2 i+2 i−2 (5)
2h 4h

where h is the unit spacing of grid, α=1/3, β1=14/9, and β2=1/9. The spatial derivatives near the immersed
boundaries are calculated in the outward direction from the wall using one-side differencing schemes of reduced
order;

2 f1′+ 4 f 2′ = ( −5 f1 + 4 f 2 + f3 ) / h
. (6)
f1′+ 4 f 2′ + f 3′ = 3 ( f3 − f1 ) / h

Note that the stencils used in spatial differencing do not cross the solid region.
Likewise, tenth order spatial filtering is also applied at interior points,

N /2
a
α f fi −1 + fi + α f fi +1 = ∑ n ( fi + n + fi − n ) (7)
n =0 2

where

193 + 126α f 105 + 302α f −15 + 30α f

a0 = , a1 = , a2 =
256 256 64 , (8)
45 − 90α f −5 + 10α f 1 − 2α f
a3 = , a4 = , a5 =
512 256 512

N=10 is the order of spatial filtering, and αf=0.49 is the filtering coefficient. This high-order spatial filtering is only
dissipative for very high wave numbers, so it is ideal for fluid regions to suppress the high-frequency numerical
errors with minimizing the dissipation of physical waves. However, near the immersed boundaries, a more diffusive
filtering is required because the application of high-order scheme to the penalization method may be suffered by
severe oscillations arising from steep gradients. Thus, approaching the boundary from the interior, the order of the
filter is systematically reduced. Furthermore, at any boundary point, one-side filtering formulation of second-order
accuracy is applied,

f1 + α f f2 = b1 f1 + b2 f 2 + b3 f3 (9)

where

3+α f 1+ α f −1 + α f
b1 = , b2 = , b3 = . (10)
4 2 4
5
American Institute of Aeronautics and Astronautics
Note again that similar to the spatial differencing, the stencils for filtering do not cross the solid obstacles. Details of
dissipation characteristics of the filtering can be found in Reference [24].
In the present study, the time integration is performed by an explicit four-stage Runge-Kutta method with the
implicit treatment of the penalization terms

Q( 0) = Q m
⎛ Δt ⎞
Q( k ) = ℜ ( k ) ⎜ Q( 0) + RHS( k −1) ⎟ (k = 1 ⋅⋅⋅ 4) , (11)
⎝ 5−k ⎠
Q m +1 = Q( 4)

where m is the time-level, Q=[ρ', u', v', p']T, and RHS corresponds to the right-hand side of Eq. (2). For 2D wave
propagation, the matrix ℜ associated with the anisotropic permeability tensor is given by

−1
⎡1 0 0 0⎤
⎢ ⎥
⎢0 1 + R11Δt R12 Δt
0⎥
5−k 5−k
ℜ( k ) = ⎢⎢ ⎥ .
⎥
(11)
R21Δt R Δt
⎢0 1 + 22 0⎥
⎢ 5−k 5−k ⎥
⎢⎣0 0 0 1 ⎥⎦

Thus, the additional stiffness due to the penalty terms in momentum equations can be avoided at the solid surface. It
is also interesting to note that when the permeability tensor Rij becomes zero in fluid region, Eq. (11) returns to the
classical low-storage Runge-Kutta method.
For the non-reflecting boundary condition, the ‘energy transfer and annihilating’ (ETA) boundary condition [25]
is applied at the far-field. The ETA is a general buffer-zone type absorbing boundary condition and is facilitated
with a rapid grid stretching in a buffer-zone and the spatial filtering. For example, grid distribution in the buffer-
zone is expressed as,

xi = xi −1 + CETA ( xi −1 − xi − 2 ) (i = 1 ⋅⋅⋅ nETA ) (11)

where xi is the position of i-th grid point, CETA is the stretching ratio, and nETA is the total number of grids in the
buffer-zone. In the present study, CETA=1.4 and nETA=20 are used for eliminating any reflections of out-going
acoustic waves at the far-field boundaries.

D. Grid Generation
As other immersed boundary techniques, the Brinkman penalization method is mostly applied on the Cartesian
grids. When acoustic scattering from multiple complex geometries is concerned, Cartesian grid approach is
computationally inefficient, because it requires very fine grids in the solid regions to represent the realistic obstacles
and to retain the order of accuracy in the vicinity of solid walls. As we have argued earlier, the use of penalization
technique is therefore prohibitively expensive without the special grid generation technique, compared to the
conventional body-fitted grid method. Thereby, in order to mimic the real shape of the surface with minimal number
of grids inside of the obstacle, either adaptive mesh refinement [26] or reshaped-cell near the boundary [18,19] is
required.
In the present study, we employ the boundary-fitted Cartesian-like grids for the efficient use of the penalization
technique. The main idea of the boundary-fitted grid generation is to reshape the meshes near the solid bodies by
applying an additional conforming/smoothing procedure. A summary of the boundary-fitted grid generation is given
below:

(1) Generate non-uniform Cartesian grids X.

(2) Find the vertices Xs adjacent to the solid surfaces.
(3) Match Xs to the surface grid Фs .
6
American Institute of Aeronautics and Astronautics
 (a)

(b)

Fig. 2. Examples of obstacles immersed in (a) Cartesian grids and (b) boundary-fitted Cartesian-like grids

(4) Compute a displacement function D=(Dx, Dy), which is given by

D = X′− X . (12)

(5) Apply the spatial filtering, Eq. (7) to the displacement function with αf = 0.2 and N=2.
(6) Compute the vertices of the new grid by using Eq. (12).
(7) Repeat (3)~(6) until the solution is converged or the desired number of iteration is obtained.
(8) Generate the resultant grids by fitting Xs to Фs .

Figure 2 compares the solid obstacles immersed in Cartesian grids and boundary-fitted Cartesian-like grids. It is
clearly shown that the proposed approach promises to not only reproduce the realistic geometry but also minimize
the grid distortion. There are many other techniques to generate good quality grids in two and higher dimensions but
the above procedure seems to be sufficient for now, in spite of its conceptual simplicity. A study for more optimized
grid generation method (e.g. combining the advantages of both adaptive mesh refinement and boundary-fitted grid
generation approach) will be pursued in the future study.

III. Results and Discussion

In this section, the validity of the Brinkman penalization method is assessed in an incremental manner. We first
carry out the prediction of acoustic scattering from a single circular cylinder. This will be followed by multiple-body
acoustic scattering problem. The present penalization method is then applied to the acoustic scattering from three
elements high-lift wing (30P30N) for more practical applications.

A. Acoustic Scattering over Single Cylinder

In order to verify the accuracy and robustness of the present Brinkman penalization method, two-dimensional
computations are conducted for acoustic scattering from a single cylinder, which was considered for the Second
Computational Aeroacoustics Workshop on Benchmark Problems [27]. A circular cylinder is located in the middle
of computational domain, which is extended from -10D to 10D in both x and y directions, where D(=1) is the
cylinder diameter. The initial condition is the Gaussian density and pressure perturbations given by
7
American Institute of Aeronautics and Astronautics
Fig. 3. Time history of fluctuating pressures at three observation points for Cartesian (left) and boundary-
fitted Cartesian-like grids (right)

⎡ ⎛ ( x − 4 )2 + y 2 ⎞ ⎤
ρ ′ = p′ = exp ⎢ − ln ( 2 ) ⎜ ⎟ ⎥ , u ′ = v′ = 0 . (13)
⎢⎣ ⎜ 0.04 ⎟⎥
⎝ ⎠⎦

In the present study, we have tested six different computational meshes for studying the grid dependency, i.e.
Cartesian grids and boundary-fitted Cartesian-like grids, in which the cylinder is represented by 100×100 grid points
(Case 1), 50×50 (Case 2), and 25×25 (Case 3) with ε=0.05 and K=1×10-6, while almost identical grid spacing is used
at the fluid region for all cases. In fact, we expect this problem to be most suitable for demonstrating the efficiency
of the present reshaped-cell approach, because the quality of the second and the third acoustic wave resulting from
the interaction between solid obstacle and principle wave is directly related to the accuracy of the present Brinkman
penalization method.

8
American Institute of Aeronautics and Astronautics
Fig. 4. Comparison of the root-mean-squared pressure distribution Prms along the y=0 line: solid lines:
analytical solution, symbols: present BPM

Figure 3 shows the time history of acoustic pressure at three observation points surround the cylinder:
A(r=5,θ=90°), B(r=5,θ=135°), and C(r=5,θ=180°), where r is the radial distance from the cylinder center and θ is
defined as an angle measured from x axis. For all monitoring points at different locations, the numerical results of
Case 3 are found to be in excellent agreement with the analytical solutions [28], and there is no noticeable
discrepancy between the numerical results on Cartesian meshes and boundary-fitted grids.
It is also interesting to note that the proposed Brinkman penalization method presents negligible amplitude and
phase error of the scattered wave, even if the number of grid points inside obstacle is reduced. This might be
considered as a consequence of proper description of immersed boundary. On the contrary, the numerical solutions
obtained from the unpolluted Cartesian grids show the significant amplitude and phase errors for Case 1 and 2,
resulting in the deficit of the scattered wave with respect to the analytical solution. In other words, the actual
location of the solid body is somewhat ambiguous with the Cartesian grids because it occurs between grid lines.
These results indicate that sufficient grids inside the bodies are essential for the original Brinkman penalization
method employing non body-fitted Cartesian grids. However, it appears in Fig. 3 that the present reshaped-cell
method improves the spatial accuracy at the solid wall with minimal number of grids, thus resulting in negligible
wave transmission into the solid body.

B. Acoustic Scattering over Two Cylinders

As the second test, we solve the acoustic scattering from two circular cylinders, presented at the Fourth
Computational Aeroacoustics Workshop [29]. This problem is served as a stringent test of the performance of the
present Brinkman penalization combined with the reshaped-cell method, to handle increasingly complex geometries.
It also provides a demonstration of numerical robustness, stability and convergence of numerical methods in the
presence of multiple scattering bodies.

9
American Institute of Aeronautics and Astronautics
Fig. 5. Root-mean-squared pressure distribution Prms at the cylinder surfaces: solid lines: analytical solution,
symbols: present BPM

We consider two circular cylinders of unequal diameters (D1=1, D2=0.5), with a co-linearly located source
separated from the center of each cylinder. In this computation, the solid obstacles are represented by the sufficiently
small porosity and permeability, i.e. ε=0.05 and K=1×10-6. The location of the cylinders are given as (x/D1,y/D1)=(-
4,0) and (x/D1,y/D1)=(4,0), respectively. The time-dependent acoustic source term, S in Eq. (2) is given by

⎡ ⎛ x2 + y 2 ⎞⎤
S = Γ ( t ) exp ⎢ − ln ( 2 ) ⎜ ⎟ ⎥ sin ( 2π ft ) . (14)
⎣ ⎝ 0.04 ⎠ ⎦

where f=4 is the frequency and Г(t) is the ramping function, which is added to gradually introduce the source into
the computational domain without generating spurious reflections,

⎧ 2 ⎛ π ft ⎞ 16
⎪sin ⎜ 32 ⎟ , t < f
⎪ ⎝ ⎠
Γ (t ) = ⎨ . (15)
⎪1 16
, t≥
⎪⎩ f

The time-accurate computations are conducted for the domain stretched from -9D1 to 9D1 in the x direction and -
4D1 to 4D1 in the y direction, with the total number of grid points 1201×501 (about 0.6 millions). The computational
grids of 201×201 and 101×101 are used inside the left and right cylinders, respectively. The number of time steps
per period of the source is set to 200 (Δt=1.25×10-3), and the simulations are performed for a total of 40,000 time
steps (or 200 periods) in order to obtain the periodic solutions. Also note that the mean value is computed over the
last 2,000 time steps.
Figure 4 shows the root-mean-squared (RMS) pressure distribution along the centerline. It is clearly observed
that the present penalization method well predicts the scattered acoustic fields resulting from the interference
between the propagating and reflecting waves, without spurious oscillations near the immersed boundaries. A good
agreement between the numerical results and analytical solutions [30] is also found in Fig. 5 which compares the
RMS pressure plots along the cylinder surfaces. Therefore, one can easily conjecture that in the present computation,
the errors associated with the representation of the solid obstacles are minimized with the conformal grids at the
immersed boundaries, and both the discontinuity and the stiffness related to the rapid change of permeability at the
fluid/solid interface are well resolved.

C. Acoustic Scattering from High-lift Wing

Finally, we consider the acoustic scattering from more complex geometries, i.e. three-element airfoil with
deployed slat and flap. This problem is expected to be more practical application of the present Brinkman
penalization method. Figure 6 shows the schematic of a high-lift wing. Under the assumption of the fluid at rest, the
10
American Institute of Aeronautics and Astronautics
 Fig. 6. Schematic of the acoustic scattering from 30P30N airfoil and grid details near slat

linearized Euler equations involving the penalty terms, Eq. (2) are solved on the computational domain ranging from
-1c to 2c in the x direction and -1.5c to 1.5c to the y direction, where c is the chord length. The computation is
carried out on the boundary-fitted Cartesian-like grids of 901×501, while the porosity and permeability for solid
bodies are set to 0.05 and 1×10-6, respectively. The turbulent noise source in Eq. (2) is modeled by a single
monopole located in the slat cove,

⎡ ⎛ ( x + 0.012 )2 + ( y − 0.01)2 ⎞ ⎤
S = exp ⎢ − ⎜ ⎟ ⎥ sin ( 2π ft ) . (16)
⎢⎣ ⎜⎝ 0.0032 ⎟⎥
⎠⎦

where f=10.
Figure 7 shows the instantaneous pressure fluctuation field computed from the present Brinkman penalization
method. Due to the boundaries of complexity, it is observed that there are multiple interferences between the
propagating and reflecting waves near the slat and flap, thus providing complex directivity pattern. It is also seen in
Fig. 7 that the slip condition at the wall is successfully imposed by anisotropic penalty terms, i.e. wall normal
velocity becomes zero, while the tangential velocity is maintained. This is another important aspect of the proposed
methodology, implying that present Brinkman penalization method promises to not only represent realistic complex
geometry but also provide the flexibility needed to impose various boundary conditions at the wall. Figure 8
provides the comparison of directivity pattern between the present penalization method based on boundary-fitted
Cartesian-like mesh and conventional body-fitted grid method. It is clearly seen that the present result agrees very
well with the direct simulation.

Fig. 7. Instantaneous pressure fluctuation plot (left) and fluctuating velocity field (right)

11
American Institute of Aeronautics and Astronautics
 Fig. 8. Comparison of directivity patterns at r=0.8c from the center (xc,yc)=(5,0): solid line represents the
present result and symbol is the solution from curvilinear grid

IV. Conclusions
In the present study, Brinkman penalization method is extended for direct simulation of acoustic scattering from
complex geometry. With the aim of increasing the computational efficiency and accuracy of the penalization
technique, several numerical techniques are also presented. By solving the linearized Euler equations that includes
the anisotropic penalization terms in momentum equations, we demonstrate the performance of the proposed
approaches for i) single source scattering from a cylinder, ii) acoustic scattering from two rigid cylinders, and iii)
acoustic scattering from three-element airfoil in a high-lift configuration. The numerical results strongly suggest that
the Brinkman penalization method is very effective, robust and accurate, when the high-order numerical schemes
with one-side spatial differencing/filtering technique and the implicit treatment of the penalty terms are applied on
the boundary-fitted Cartesian-like grids. The present approach will be beneficial for prediction of the aerodynamic
noise around complex moving boundaries, so this will be the subject of future study.

Acknowledgments
The authors would like to acknowledge Mr. Jong Rok Kim at Korea University for providing the two-
dimensional curvilinear grids of 30P30N model.

References
1
Thompson, J. F., Warsi, Z. U. A, and Mastin, C. W., T. H., “Boundary Fitted Coordinate Systems for Numerical Solution of
Partial Differential Equations – a Review,” Journal of Computational Physics, Vol. 47, pp. 1–108, 1982.
2
Mavriplis, D. J., “Accurate Multigrid Solution of the Euler Equations on Unstructured and Adaptive Meshes,” AIAA Journal,
Vol. 28, pp. 213–221, 1990.
3
Wang, Z. J., “Spectral (Finite) Volume Method for Conservation Laws on Unstructured Grids. Basic Formulation,” Journal
of Computational Physics, Vol. 178, pp. 210–251, 2002.
4
Mittal, R. and Iaccarino, G., “Immersed Boundary Methods,” Annual Review of Fluid Mechanics, Vol. 37, pp. 239–261,
2005.
5
Chung, C. and Morris, P. J., “Acoustic Scattering From Two- and Three-Dimensional Bodies,” Journal of Computational
Acoustics, Vol. 6, pp. 357–375, 1998.
6
Laik, O. A. and Morris, P. J., “Direct Simulation of Acoustic Scattering by Two and Three-Dimensional Bodies,” Journal of
Aircraft, Vol. 37, pp. 68–75, 2000.
7
Tang, H. S., “An Overset-Grid Method for 3D Unsteady Incompressible Flows,” Journal of Computational Physics, Vol.
191, pp. 567–600, 2003.
8
Sherer, S. E. and Scott, J. N., “High-Order Compact Finite-Difference Methods on General Overset Grids,” Journal of
Computational Physics, Vol. 210, pp. 459–496, 2005.
9
Desquesnes, G., Terracol, M., Manoha, E., and Sagaut, P., “On the Use of a High Order Overlapping Grid Method for
Coupling in CFD/CAA,” Journal of Computational Physics, Vol. 220, pp. 355–382, 2006.
12
American Institute of Aeronautics and Astronautics
 10
Hu, F. Q., Hussaini, M. Y., and Rasetarinera, P, “An Analysis of the Discontinuous Galerkin Method for Wave Propagation
Problems,” Journal of Computational Physics, Vol. 151, pp. 921–946, 1999.
11
Delorme, P., Mazet, P., Peyret C., and Ventribout, Y., “Computational Aeroacoustics Applications Based on a
Discontinuous Galerkin Method,” Comptes Rendus Mécanique, Vol. 333, pp. 676–682, 2005.
12
Rao, P. P. and Morris, P. J., “Application of a Generalized Quadrature Free Discontinuous Galerkin Method in
Aeroacoustics,” 9th AIAA/CEAS Aeroacoustics Conference and Exhibit, AIAA 2003-3120, 2003.
13
Peskin, C. S., “Flow Patterns around Heart Valves: a Digital Computer Method for Solving the Equations of Motion,” Ph.D.
Thesis, Albert Einstein College of Medicine, 1972.
14
Peskin, C. S., “Flow Patterns around Heart Valves: a Numerical Method,” Journal of Computational Physics, Vol. 10, pp.
252–271, 1972.
15
Lai, M. C. and Peskin, C. S., “An Immersed Boundary Method with Formal Second Order Accuracy and Reduced
Numerical Viscosity,” Journal of Computational Physics, Vol. 160, pp. 705–719, 2000.
16
Cho, Y., Boluriaan, S., and Morris, P. J., “Immersed Boundary Method for Viscous Flow Around Moving Bodies,” 44th
AIAA Aerospace Sciences Meeting and Exhibit, AIAA 2006-1089, 2006.
17
Liu, Q. and Vasilyev, O. V., “A Brinkman Penalization Method for Compressible Flows in Complex Geometries,” Journal
of Computational Physics, Vol. 227, pp. 946–966, 2007.
18
Ye, T., Mittal, R., Udaykumar, H. S., and Shyy, W., “An Accurate Cartesian Grid Method for Viscous Incompressible
Flows with Complex Immersed Boundaries,” Journal of Computational Physics, Vol. 156, pp. 209–240, 1993.
19
Udaykumar, H. S., Mittal, R., Rampunggoon, P., and Khanna, A., “A Sharp Interface Cartesian Grid Method for Simulating
Flows with Complex Moving Boundaries,” Journal of Computational Physics, Vol. 174, pp. 345–380, 2001.
20
Yang, J. H. and Lee, S. L., “Effect of Anisotropy on Transport Phenomena in Anisotropic Porous Media,” International
Journal of Heat and Mass Transfer, Vol. 42, pp. 2673–2681, 1999.
21
Jiang, G. S. and Shu, C. W., “Efficient Implementation of Weighted ENO Schemes,” Journal of Computational Physics,
Vol. 126, pp. 202–228, 1996.
22
Ren, Y., Liu, M., and Zhang, H., “A Characteristic-Wise Hybrid Compact-WENO Scheme for Solving Hyperbolic
Conservation Laws,” Journal of Computational Physics, Vol. 192, pp. 365–386, 2003.
23
Lele, S. K., “Compact Finite Difference Schemes with Spectral-like Resolution,” Journal of Computational Physics, Vol.
103, pp. 16–42, 1992.
24
Gaitonde, D., Shang, J. S., and Young, J. L., “Practical Aspects of Higher-Order Numerical Schemes for Wave Propagation
Phenomena,” International Journal for Numerical Methods in Engineering, Vol. 45, No. 12, 1999, pp. 1849–1869.
25
Edgar, N. B. and Visbal, M. R., “A General Buffer Zone-Type Non-Reflecting Boundary Condition for Computational
Aeroacoustics,” 9th AIAA/CEAS Aeroacoustics Conference and Exhibit, AIAA Paper 2003-3300, 2003.
26
de Tullio, M. D., De Palma, P., Iaccarino, G., Pascazio, G., and Napolitano, M., “An Immersed Boundary Method for
Compressible Flow using Local Grid Refinement,” Journal of Computational Physics, Vol. 225, pp. 2098–2117, 2007.
27
Tam, C. K. W. and Hardin, J. C. (Eds.), Second Computational Aeroacoustics (CAA) Workshop on Benchmark Problems,
NASA Conference Publication 3352, 1997.
28
Tam, C. K. W. and Hu, F. Q., “An Optimized Multi-Dimensional Interpolation Scheme for Computational Aeroacoustics
Applications Using Overset Grids,” 10th AIAA/CEAS Aeroacoustics Conference and Exhibit, AIAA 2004-2812, 2004.
29
Dahl, M.D. (Ed.), Fourth Computational Aeroacoustics (CAA) Workshop on Benchmark Problems, NASA Conference
Publication 2004-212954, 2004.
30
Sherer, S. E., “Scattering of Sound from Axisymetric sources by multiple circular cylinders,” Journal of the Acoustical
Society of America, Vol. 115, pp. 488-496, 2004.

13
American Institute of Aeronautics and Astronautics