Chinese

Journal of
Aeronautics
Chinese Journal of Aeronautics 22(2009) 480-485
www.elsevier.com/locate/cja
Numerical Simulation of Low Reynolds Number Fluid-Structure
Interaction with Immersed Boundary Method
Ming Pingjian, Zhang Wenping
School of Power and Energy Engineering, Harbin Engineering University, Harbin 150001, China
Received 19 September 2008; accepted 21 April 2009
Abstract
This article introduces a numerical scheme on the basis of semi-implicit method for pressure-linked equations (SIMPLE) algo-
rithm to simulate incompressible unsteady flows with fluid-structure interaction. The Navier-Stokes equation is discretized spa-
tially with collocated finite volume method and Eulerian implicit method in time domain. The hybrid method that combines im-
mersed boundary method (IBM) and volume of fluid (VOF) method is used to deal with rigid body motion in fluid domain. The
details of movement of immersed boundary (IB) and calculation of VOF are also described. This method can be easily applied to
any existing finite-volume-based computational fluid dynamics (CFD) solver without complex operation, with which fluid flow
interaction of arbitrarily complex geometry can be realized on a fixed mesh. The method is verified by low Reynolds number
flows passing both stationary and oscillating cylinders. The drag and lift coefficients acquired by the study well accord with other
published results, which indicate the reasonability of the proposed method.
Keywords: fluid-structure interaction; immersed boundary method; volume of fluid; vortex shedding; incompressible flow
1. Introduction
1

Over the past two decades, time-marching algo-
rithms have found wide application in simulating com-
pressible flows to solve Euler and Navier-Stokes (N-S)
equations. However, they are not fit for incompressible
flows due to its inefficiency and poor convergence with
the reality
[1]
. Several methods have been developed to
overcome this problem and the most successful one
seems to be the pressure-based semi-implicit method
for pressure-linked equations (SIMPLE) algorithm pro-
posed by S. V. Partankar, et al. in 1972
[2]
, which has
been extended to high-Mach-number flows by the au-
thors of Refs.[3]-[4].
Recently, numerical method has attracted ever-
growing interest from researchers. They investigate
moving bodies in flow fields and, as a result, a few
methods have been put forward. Of them, the di-
rectly-related one is to use Lagrangian moving mesh to
track the interfaces. Unfortunately, it is very inefficient
or even impossible to simulate big changes, for which a
remeshing procedure is needed to deal with low mesh
quality. Some researchers adopted an immersed bound-

*Corresponding author. Tel.: +86-451-82569458.
E-mail address: pingjianming@hrbeu.edu.cn

1000-9361/$ - see front matter 2009 Elsevier Ltd. All rights reserved.
doi: 10.1016/S1000-9361(08)60129-6
ary method to simulate the rigid bodys motion on a
fixed mesh in the last couple of years. The immersed
boundary method was originally proposed by C. S.
Peski
[5]
to analyze the blood flow in a human heart
model by merely performing the simulation on a simple
Cartesian mesh system, which, in fact, did not conform
to the geometry of the heart (non-body-fitted mesh).
Since then, the term immersed boundary (IB) method
(IBM) has begun to be broadly applied, i.e., the body is
simply immersed in the background mesh domain
without fitting the underlying mesh points to the im-
mersed body as practiced in most of the conventional
computational fluid dynamics (CFD) approaches
[6]
.
The methods of simulating flows with complex im-
mersed boundaries can be briefly classified into two
kinds
[7]
: To represent the effects of the immersed
boundary on the surrounding fluid phase, the immersed
boundary is regarded as a diffuse interface of finite
thickness. The thickness of the boundary is usually of
the order of the local mesh spacing. Among this kind of
methods, most employ body and/or surface forces
and/or mass sources, and also some use the volume-
of-fluid and phase-field approaches. The boundary
is tracked as a sharp interface, either explicitly as
curves or as level sets on fixed meshes. The communi-
cation between the moving boundary and the flow
solver is usually accomplished directly by modifying
the computational stencil near the immersed boundary.
No.5 Ming Pingjian et al. / Chinese Journal of Aeronautics 22(2009) 480-485 481

The second kind is more difficult to use and may re-
quire substantial changes of existing CFD codes.
Therefore, the first kind is adopted in this article to
implement in an in-house code and solve fluid-structure
interaction problems in engineering.
2. Computational Scheme
The conventional method uses structured and/or un-
structured meshes that are fitted to the body (body-
fitted mesh), which is very convenient to create a mesh
for IBM. Fig.1 shows the difference between IB and
conventional body-fitted mesh method.

Fig.1 Different meshing methods.
2.1. Governing equation
The incompressible Navier-Stokes equation can be
expressed in an integral form by
d 0
O

c
=
)
U S (1)
for continuity, and
T
( )
d d d
[ +( ) ] d d
p V
t
V
O O O
O O

c
c
c
+ = +
c
+
) ) )
) )
U
UU S
U U S F
V
V V (2)
for momentum.
In Eqs.(1)-(2), O is the integral domain, cO the
boundary, dS the elemental area vector, dO the ele-
mental domain, U the velocity vector [u v], the
fluid density, the fluid dynamic viscosity, p the static
pressure, and F the body force vector [ f
x
f
y
].
2.2. Numerical scheme
(1) Momentum prediction
Given a certain face flux and pressure, the momen-
tum equation Eq.(2) can be solved for the cell center
velocity vector.
With the finite volume method and Eulerian implicit
scheme, the linear equations can be obtained as fol-
lows:
*
p p j j t p
n
x x x
p
A u A u D Df Au
x
c
' ' = + +
c
_
(3)
*
p p j j t p
n
y y y
p
A v A v D Df Av
y
c
' ' = + +
c
_
(4)
where
p j t x x
A A A = +
_
,
p j t
,
y y
A A A = +
_
A
xj
and A
yj

are neighbor cell influence coefficients in x- and y-
direction, A
t
is the coefficient related to unsteady term,
D the cell volume, p
*
the initial or intermediate resul-
tant pressure,
p
u' and
j
u' are the velocities in x-direction
of current cell and neighbor cell,
p
v' and
j
v' the
counter parts in y-direction. All these velocities are of
current time step values while
p
n
u and
p
n
v are of previous
time step values.
The immersed boundary effects on the flow are real-
ized by volume of fluid (VOF) value. The body forces
can be calculated according to the method
x
u u
f
t
'' '
=
A
(5)
y
v v
f
t
'' '
=
A
(6)
where u and v are the second correction values,
which can be linearly interpolated with the fluid and
immersed rigid body velocities. This method was also
used by K. C. Ng
[8]
:
1
s
(1 )
n
u u u o o
+
'' ' = + (7)
+1
s
(1 )
n
v v v o o '' ' = + (8)
where o is the solid fraction within a cell. By substi-
tuting u and v into Eqs.(5)-(6), the following can be
acquired
1
s
n
x
u u
f
t
o
+
'
=
A
(9)
1
s
n
y
v v
f
t
o
+
'
=
A
(10)
Then the force acting on rigid body can be found by
summing up forces over all cells.
( d )
x x
F f V =
_
(11)
( d )
y y
F f V =
_
(12)
(2) Face flux calculation
In order to overcome the check-board pressure dis-
tribution, a fourth order damping scheme proposed by
C. M. Rhie, et al.
[9]
is implemented. The cell face flux
can be computed by
482 Ming Pingjian et al. / Chinese Journal of Aeronautics 22(2009) 480-485 No.5

f
m ' ' = U S (13)
where m' is the face flux,
f
' U the cell face velocity
vector [
f
u'
f
v' ] and S the cell face area vector.
(3) Pressure correction
After predicting the momentum, the velocity does
not necessarily satisfy the continuity equation. When
the solution converges, the velocity and pressure must
satisfy the ensuing momentum equation:
p p j j t p
( )
n
x x x
p
A u A u D Df Au
x
c
= + +
c
_
(14)
where u
p
, u
j
and p are converged results.
By subtracting Eq.(3) from Eq.(14) and neglecting
influences of neighbor cells on the third velocity cor-
rection, the following equation can be derived:
p x
D p
u
A x
' c
''' =
c
(15)
Similarly,
p y
D p
v
A y
' c
''' =
c
(16)
where
p p p
u u u '' ''' = + ,
p p p
v v v '' ''' = + and
*
p p p' = + .
The face velocities can be read in the same form as
the cell center Eqs.(15)-(16).
f
f p
f
x
D p
u
A x
| |
' c | |
'' = |
|
|
c
\ .
\ .
(17)
f
p
f
f
y
D p
v
A y
| |
' | | c
'' = |
|
|
c
\ .
\ .
(18)
f
m = '' '' U S (19)
in which m'' is the face flux modification and
f
'' U
the cell face velocity vector [
f
u''
f
v'' ] modification.
( ) 0 m m'' + =
_
(20)
Substituting Eq.(13) and Eq.(19) into Eq.(20) gives
p i
p p
1 1
( )
x x y y
SD p p m
A d A d

( | |
' ' ' + = ( |
|
(
\ .
_ _
(21)
which is a pressure correction equation. The pressure
can be corrected by solving Eq.(21).
(4) Updating flux and velocity
Results:
1
p
p
n
x
D p
u u
A x
+
' c
'' =
c
(22)
1
p
p
n
y
D p
v v
A y
+
' c
'' =
c
(23)
f f 2 1
p
( )
x
D
u u p p
A
' ' ' = (24)
f f 2 1
p
( )
y
D
v v p p
A
' ' ' = (25)
(5) Calculating VOF
As the immersed boundary is tracked by Lagrangian
method, it is necessary to decide which cell the point is
in. A subroutine point_in_polygon has been used to
determine the situation efficiently. As shown in Fig.2, a
line is drawn through the immersed boundary points
(IBPs), for instance a, b, c. And the number of points,
which have crossed the cell faces on right side of the
IBP, is counted. If the number is odd, then the IBP is
outside the cell; otherwise, it is in the cell. In Fig.2,
there are two points on the right side of point a, but
none on the right side of point c. Thus, points a and c
are outside the cell while there is one crossed point on
the right side of the point b, meaning in the cell.

Fig.2 Relative position of point and cell.
After deciding the cells containing IBP, the VOF can
be calculated. Since VOF can be defined as the ratio of
the solid volume in one cell to the cell volume, 1 stands
for solid cell and 0 for fluid cell.
In Fig.3(a), there is one boundary line crossing the
cell, so there must be at least another line cutting this
cell, for example line cd in Fig.3(b).

Fig.3 Schematic diagram showing calculation of VOF of
immersed body.
As shown in Fig.3(b), the IBPs a and d, crossed
points m and n and the cell node A constitute the solid
area polygon 2Amadn, which can be divided into three
triangles (mAn, mnd and mda), and the solid
area equals the sum of the areas of these triangles.
2.3. Numerical algorithm
The computational steps can be summarized as fol-
lows.
Step 1 Start the computation at the time step n +1.
Step 2 Determine the velocities of all IBPs at the
time step n depending on the prescribed motion.
Step 3 Update the new position of each IBP at the
time step n+1 by using individual velocities at the pre-
vious time step n:
1 1
s
d
n n n
x x u t
+ +
= + (26)
No.5 Ming Pingjian et al. / Chinese Journal of Aeronautics 22(2009) 480-485 483

1 1
s
d
n n n
y y v t
+ +
= + (27)
Step 4 With the new position of IBP, calculate the
VOF of each cell at the time step n+1 by using the
method introduced in Section 2.2.
Step 5 Solve the momentum prediction equations
Eqs.(3)-(4) to obtain u' and v'.
Step 6 Update intermediate velocities u and v
with Eqs.(7)-(8).
Step 7 Calculate the fluxes with Eq.(13).
Step 8 Solve the pressure correction equation Eq.(21)
and obtain the corrected pressure.
Step 9 Update the velocities with Eqs.(22)-(23)
and fluxes with Eqs.(24)-(25).
Step 10 Repeat the procedure from Step 5 to Step
9 until a converged solution is gained.
Step 11 Repeat the whole procedure from Step 1
until the end time is reached.
3. Results
3.1. Stationary cylinder
To quantify the accuracy of the proposed immersed
boundary method, several classical benchmark prob-
lems used to validate incompressible flow solvers are
computed. The first case is of low Reynolds number 2D
unsteady flow over a cylinder with the diameter d,
which exhibits vortex shedding at Reynolds numbers
(Re =100). The computational domain is 30d30d,

the
cylinder center 10d away from inlet and 15d away from
both top and bottom boundaries. The number of mesh
points is 301301 uniformly distributed in both stream-
wise and transverse directions.
The Dirichlet boundary condition of u/u
in
= 1 and
v = 0 m/s is applied to the inlet, a slip wall to both the
top and the bottom, and the fixed pressure condition to
the outlet boundary are used. As the force acting on the
rigid body is given by Eqs.(11)-(12), the drag and lift
coefficients can be computed by

2
in
/ 2
x
D
F
C
u d
= (28)
2
in
/ 2
y
L
F
C
u d
= (29)
Table 1 lists the drag and lift coefficients and Strou-
hal number Sr, which well accord with other available
data and attest to the reasonability of the proposed
method.
Table 1 Results of flows passing a cylinder at Re = 100
Re = 100
Method
C
D
C
L
Sr
Ref.[10] 1.33 0.320 0.165
Ref.[11] 1.350.014 0.300 0.175
Ref.[12] 1.380.007 0.322 0.169
Ref.[13] 1.340.011 0.315 0.164
Proposed 1.340.008 0.313 0.165
The time history of the steady vortex shedding is il-
lustrated in Fig.4, which displays the variation of fre-
quency of lift coefficient approximately twice the drag
coefficient and, in addition, there are two different am-
plitudes appearing alternately in drag coefficients.

Fig.4 Time history of drag and lift coefficients with steady
vortex shedding.
Fig.5 shows the streamline distribution at different
time. From it, it is seen that, before 200 s, a pair of
symmetric vortices forms behind the cylinder and
stretches with time and finally vortex is shed. Fig.5
shows a direct look of the wake behind the cylinder.


484 Ming Pingjian et al. / Chinese Journal of Aeronautics 22(2009) 480-485 No.5





Fig.5 Streamline distribution at different time.
3.2 Oscillating cylinder
To verify the suitability of the proposed method for a
moving boundary problem, an in-line oscillating cylin-
der in stationary fluid is considered. The two character-
istic parameters of the flow are the Reynolds number
and Keulegan-Carpenter number, which can be deter-
mined by
max
U d
Re

= (30)
max
U
KC
fd
= (31)
where U
max
is the maximum velocity of cylinder, f the
frequency of oscillation.
In-line oscillatory motion of the cylinder is typical of
a simple harmonic relationship like
m
( ) sin(2 ) x t A ft = t (32)
where x(t) is the position of cylinder center and A
m
the
amplitude of oscillation.
Then the velocity of the cylinder can be obtained by
deriving the position with respect to time
s m
( ) 2 cos(2 ) u t fA ft = t t (33)
So the correlation between maximum velocity and
amplitude of oscillation is
max m
2 U fA = t (34)
The computation domain is 10d 10d, where d =
0.01 m; the fluid density =1 kg/m
3
and the fluid dy-
namic viscosity =1.010
3
Pas. Given Re =100 and
KC = 5, can be obtained U
max
= 0.1 m/s, f = 2 Hz and
A
m
= 0.007 957.
Fig.6 shows the numerical results of time history of
the drag coefficients in a single period. T represents one
period. The consistency of the curve with that in Ref.[6]
indicates the ability of the proposed method to accu-
rately predict the surface shear stress and the pressure
on the cylinder.

Fig.6 Time history of drag coefficients at Re =100 and
KC = 5.
Fig.7 shows the velocity vectors and the VOF con-
tour at different time. With the movement of cylinder,
the wake flow of cylinder is changed. There are two
symmetry vortices behind the cylinder in most stages.
No.5 Ming Pingjian et al. / Chinese Journal of Aeronautics 22(2009) 480-485 485


(a) T/8 (b) T/4

(c) 3T/8 (d) T/2

(e) 5T/8 (f) 3T/4

(g) 7T/8 (h) T
Fig.7 Velocity vector and VOF contour at different stages in
a single period.
4. Conclusions
This article proposes a numerical method for
fluid-structure interaction, which combines the im-
mersed boundary method and the volume of fluid
method in the framework of pressure-finite volume
method. This method can be easily applied to any ex-
isting finite-volume-based CFD solver without com-
plex operation.
To verify the proposed method, flows passing a sta-
tionary and an oscillating cylinder are taken as exam-
ples to be calculated and the results agree well with the
available results in the literature. This evidences the
reasonability and accuracy of the proposed method.
References
[1] Turkel E. Robust low speed precondition for viscous
high lift flows. AIAA-2002-0926, 2002.
[2] Patankar S V, Spalding D B. A calculation procedure for
heat, mass and momentum transfer in three dimension
parabolic flows. International Journal of Heat and Mass
Transfer 1972; 15: 1787-1796.
[3] Shyy W, Chen M H. Pressure-based multigrid algorithm
for flow at all speeds. AIAA Journal 1992; 30(11):
2660-2669.
[4] Date A W. Solution of Navier-Stokes equations on non-
stagered grid of all speeds. Numerical Heat Transfer,
Part B 1998; 33: 451-467.
[5] Peskin C S. Flow patterns around heart valves: a nu-
merical method. Journal of Computational Physics 1972;
10: 252-271.
[6] Dutsch H, Durst F, Becker S. Low Reynolds number
flow around an oscillating circular cylinder at low
Keulegan-Carpenter numbers. Journal of Fluid Me-
chanics 1998; 360: 249-271.
[7] Udaykumar H S, Mittal R, Rampunggoon P, et al. A
sharp interface cartesian grid method for simulating
flows with complex moving boundaries. Journal of
Computational Physics 2001; 174: 345-380.
[8] Ng K C. A collocated finite volume embedding method
for simulation of flow past stationary and moving body.
Computers & Fluids 2009; 38: 347-357.
[9] Rhie C M, Chow W L. Numerical study of the turbulent
flow past an airfoil with trailing edge separation. AIAA
Journal 1983; 21(11): 1525-1532.
[10] Kim J, Kim D, Choi H. An immersed boundary finite
volume method for simulations of flow in complex
geometries. Journal of Computational Physics 2001;
171: 132-150.
[11] Calhoun D. A Cartesian grid method for solving the
two-dimensional stream function-vorticity equations in
irregular region. Journal of Computational Physics 2002;
176: 231-275.
[12] Russell D, Wang Z J. A Cartesian grid method for mod-
eling multiple moving objects in 2D incompressible
viscous flow. Journal of Computational Physics 2003;
191: 177-205.
[13] Choi J I, Oberoi R C, Edwards J R, et al. An immersed
boundary method for complex incompressible flows.
Journal of Computational Physics 2007; 224: 757-784.
Biographies:
Ming Pingjian Born in 1980, he recived Ph.D. degree in
power machine and engineering from Harbin Engineering
UniversityHEUin 2008 and has been doing postdoctoral
research work since then in HEU. His main research interest
is computational fluid dynamics.
E-mail: pingjianming@hrbeu.edu.cn

Zhang Wenping Born in 1956, he recived Ph.D. degree in
marine engineering from Harbin Engineering University in
1992 and now he is a professor there. His research interests
include noise and vibration control and computational acous-
tics.
E-mail: zhangwenping@hrbeu.edu.cn