You are on page 1of 6

203

IBM/ICM method for interactions between moving bodies and free surfaces

Deng Jian *, Xie Fang-fang, Shao Xue-ming, Yu Zhao-sheng


School of Aeronautics and Astronautics, Zhejiang University, Hangzhou, China
*Email: zjudengjian@zju.edu.cn

ABSTRACT: The IBM/ICM method is developed to transformations and mapping techniques is possible
simulate the interactions between moving bodies and free but requires a highly accurate way of calculating the
surfaces. This method is a combination of Immersed Boundary transformation Jacobians, and even for simple
Method (IBM) and Interface Capture Method (ICM), inheriting geometries, generating a good-quality body-conformal
the IBM’s capability of using non-body conformal grids to grid can be an iterative process requiring significant
represent the effect of a moving body in the flow by only devotion of the time. As the geometry becomes more
adding a pseudo body-force in the right side of Navier-Stokes complicated, the task of generating an acceptable grid
equations, as well as the capability of ICM, i.e. VOF, to capture becomes increasingly difficult. The unstructured grid
the interface between two immiscible fluids. In the current approach is inherently better suited for complex
paper, the Lagrange-multiplier based IBM method is employed geometries, but here, grid quality can deteriorate with
and the volume of fraction (VOF) is introduced as an indicator increasing complexity in the geometry. As an
function to mark the different fluids. The numerical method is alternative solution, the non-boundary-fitted method
validated by the case of high-speed impact by a horizontal has become increasingly popular for the solution of
circular cylinder, which exhibits a wide range of dynamical the fluid flow problems in a complex geometry or
response characteristics, i.e. the formation of a cavity and with moving boundaries, for example the immersed
induced pressure waves in the late stages of the cavity collapse, boundary method (IBM). The main advantages of this
depending primarily on the projectile’s impact velocity. method are memory and CPU savings and easy grid
generation. Grid complexity and quality are not
KEY WORDS: Immersed Boundary Method; Interface significant affected by the complexity of the geometry.
Capture Method; IBM/ICM Method. Even moving-boundary problems can be handled with
the immersed-boundary method without regenerating
1 INTRODUCTION grids in time, unlike the unstructured grid and
body-fitted grid methods.
Treating complex geometries has been one of the
The IBM was firstly developed by Peskin[1] to model
main challenges in computational fluid dynamics
the moving boundaries in heart valve simulations.
because most engineering problems have complex
Then, it was improved and developed by Goldstein et
geometries. So far, two traditional techniques have
al. [2], Mohd-Yusof [3], Fadlun et al. [4] and Mittal and
been developed: using the unstructured grid for finite
Iaccarino[5]. It is convenient and economical to use the
volume method or finite element method and the
IBM to simulate the flow field with moving bodies,
coordinate transformations and mapping techniques
since the whole simulations can be carried on an
for finite difference method. The use of coordinate
unchanged Cartesian grid each time step and only a
204

body-force term is added to the momentum equation. liquid-gas flow field.


There exist various IBM based methods in the
literature that differ in the way the body-force is 2 NUMERICAL MODEL
calculated. For example, for the elastic-force based
2.1 Fictitious domain formulation
methods[1, 6], the boundary moves with the local fluid
and the body force is explicitly calculated from the The momentum equation for the fluid flow is
displacement or deformation of the boundary (large du
ρf = ∇ ⋅σ (1)
stiffness constant as an approximation to the dt
rigid-boundary problems). For the direct-forcing where ρ f is the fluid density, denoting the mixture
based methods[7-11], the moving body or boundary is density for two-phase flow, u the fluid velocity, and
modeled by a series of Lagrangian points, on which σ the fluid stress. For Newtonian fluid
the body force is defined, and a discrete δ-function is σ = −( p + ρgh)I + 2μD , here p being the fluid
used to interpolate the fluid velocity from the Eulerian pressure, μ the viscosity and D the rate-of-strain
nodes to the Lagrangian points for the calculation of tensor.
the body-force and then to distribute the body-force The motion of the rigid body is governed by
from the Lagrangian points to the Eulerian nodes for Newton’s equation of motion as follows:
the solution of the fluid momentum equation. For the
dU
Lagrange-multiplier based fictitious domain M = F H + Mg ,
dt
[13-14] (2)
method , the body-force is introduced as a d( J ⋅ ω s )
= TH
Lagrange multiplier in a weak formulation, and solved dt
from the boundary condition in an iterative way. There where M, J, U, and ω s are the body mass, moment
also exits another kind of IBMs[15-18], for which the of inertia tensor, translational velocity and angular
body-force is directly defined at the Eulerian nodes velocity, respectively. g is the gravitational
located in the immediate vicinity of the boundary in
acceleration, and ρ s is the solid density. F H and
either the real fluid region or the fictitious domain,
T H are the hydrodynamic force and torque on the
and the target velocities at the forcing points for the
body, respectively, defined by
calculation of the body-force is obtained by
interpolation of the prescribed velocity at the FH = ∫ n ⋅ σ ds ,
boundary and other fluid velocities at the surrounding ∂P
(3)
Eulerian nodes. These methods do not really introduce TH = ∫ r × (n ⋅ σ)ds
a body-force, instead, it accounts for the boundary ∂P

condition by modifying the expressions of differential where n is the unit outward normal on the body
operators for the Eulerian nodes in the immediate surface and r is the position vector with respect to the
vicinity of the boundary. body mass center. The gravity term is generally
The IBM/ICM In this paper, following the main idea considered as a hydrostatic modification to the
proposed in our previous paper[19], the Lagrangian pressure, and this additional term can be expressed as
points are distributed in the moving body to introduce ρgh = ρg ⋅ rs (4)
a pseudo body-force, which enforce the pseudo fluid
inside the body to satisfy the solid body moving where r is the position vector with respect to an any
behavior, as well as the non-slip condition on the chosen point in the flow field.
surface of the body. The difference between this paper The interior of the body is assumed to be filled with
and our previous papers is that an arbitrary cell based the fluid, liquid or gas, and a pseudo body-force is
finite-volume solver are employed, and moreover, the introduced over the body inner domain to enforce the
IBM is extended to treat with moving body in fictitious fluid to satisfy the rigid-body motion
205

constraint. The following equations are introduced for ρ = αρ l + (1 − α ) ρ g


the interior of the body: (10)
μ = αμ l + (1 − α ) μ g
du
ρf = ∇ ⋅ σ + λ in P(t ),
dt (5) By introducing the indicator function, the surface
u = U + ω s × r in P(t ) tension at the interface of two fluids can be resolved,
where λ is the pseudo body-force. Integrating and the following term is added to Eq.(1) representing
(5) and r × (5) over the body interior P(t) and the source of momentum due to surface tension.
substituting into Eqs.(3) yields, respectively,
∫S (t )
σκ ′n′δ ( x − x′)dS (11)

ρ f dU It only acts at the interface as indicated by the


F H = − ∫ λdp + M ,
P ρs dt integral of a three-dimensional δ -function over the
(6)
ρ d( J ⋅ ω s ) entire surface of the interface S(t). The other variables
T H = − ∫ r × λdp + f
P ρs dt in this term denote the surface tension coefficient σ
as well the curvature κ and the normal vector n of
Substituting (6) into (3) gives, respectively,
the interface.
⎡ ⎛ ρ f ⎞ du ⎤
⎢⎜ 1 − ⎟ − g ⎥ M = − ∫ λdp, 2.3 The PISO algorithm for transient flows
⎣⎝ ρ s ⎠ dt ⎦
P
(7)
⎛ ρ f ⎞ d( J ⋅ ωs ) The basic PISO solver for N-S equations without
⎜1 − ⎟ = − ∫ r × λdp
ρ s ⎠ dt pseudo body-force involved is described as follows:
⎝ P

(i) The momentum equation is solved first. The exact


For detailed description and non-dimensionlized form
pressure gradient source term is not known at this
of the equations above, one can refer to [19].
stage, where the pressure from the previous time-step
2.2 Interface-capturing methodology is used instead. This stage is called the momentum
predictor. The solution of the momentum equation
The interface-capturing methodology employs the
gives an approximation of the new velocity field.
volume of fraction (VOF) as an indicator function to
(ii) A pressure poisson equation is constructed and
mark the different fluids. The interface is not defined
solved with the predicted velocities. The solution of
as a sharp boundary and a transition region exist
the pressure equation gives the first estimate of the
where the fluid is treated as some mixture of the two
new pressure field. This stage is called the pressure
fluids on each side of the interface. In reality, the
solution.
transition region is a discontinuous step. The indicator
(iii) The velocity field should be corrected as a
function is defined as:
consequence of the new pressure distribution. Velocity
⎧1 in liquid correction is done in an explicit manner in this paper.

α = ⎨0 < α < 1 in transitional region (8) This is the velocity correction stage.
⎪0 in gas
⎩ 2.4 Solution of the body subproblem
It is associated with each fluid and is hence
The difficulty of the IBM comes from the fact that the
propagated with them as a Lagrangian invariant.
fluid velocity is defined at the Eulerian nodes,
Therefore, it obeys a transport equation of the form:
whereas the solid body needs to be described with the
∂α
+ (u ⋅ ∇)α = 0 (9) Lagrangian nodes that normally does not coincide
∂t with the Eulerian nodes, and consequently it is not
Using the indicator function, the local density ρ and
straightforward how to solve the equations involving
the local viscosity μ of the fluid are expressed as:
both Eulerian and Lagrangian quantities.
In this paper, a systematic interpolation method is
206

constructed to transfer the quantity between the


Eulerian and Lagrangian frames accurately, i.e. the
velocity is transferred from the Eulerian nodes to
Lagrangian points and the pseudo body-force is
transferred from the Lagrangian points to the Eulerian
nodes. The detailed description about the interpolation
strategy will be discussed in our future papers, and the
illustration about this two systems are shown in Fig.1,
and as one can see the Lagrangian points are also well
organized as a finite volume grid system. Fig.2 Lift and drag coefficients time history at Re=100 for the
flow around a circular cylinder

Fig.3 Vortex shedding from a circular cylinder at Re=100

3.2 Water entry of a horizontal circular cylinder

In this section, the classical problems of water entry of


a horizontal circular cylinder are solved using the
Fig.1 Illustration of Eulerian cell and Lagrangian point IBM/ICM method presented previously. The results
are compared with model tests. Fig.4 presents one
3 NUMERICAL RESULTS
case of our simulations, the free surface deformation
3.1 Uniform flow around a circular cylinder during water entry of a half-buoyant cylinder is
captured. The Froude number is 0.646 in this case.
The uniform flow around a circular cylinder is
The mesh size around the cylinder is 0.05. A initial
simulated to demonstrate the accuracy of this method
time step dt=0.01 and is adjusted according to the
to predict the hydrodynamics forces on an obstacle
Courant number in the following steps. The falling
immersed in a single phase flow. For this validation,
velocity of the cylinder is determined by the gravity
Re=100. Lift and drag coefficients are shown in Fig.2,
and the hydraulic force, as exhibited in Eq.7.
and the vorticity contours are shown in Fig.3. The
As one can see, after the cylinder reaches the free
average drag coefficient is 1.449, slightly higher than
surface, two jets are thrown up on both sides of the
the result by Yu and Shao [19], where a value 1.394 was
cylinder and move away from the cylinder leaving the
found, and very close to Peskin’s result of 1.447[6].
upper part of cylinder dry, even when h/R>2, where h
Fourier analysis of the fluctuating lift yields a Strouhal
is the submergence of the lowest point on the cylinder
number of 0.176. This value matches well with our
relative to the mean free surface. Obviously the
previous numerical result of 0.174 [19].
prediction of the viscous flow separation can be
influenced by the grid size. In the future work,
relatively small grids will be used at the areas of flow
separation and the tiny liquid drops can be captured.
207

of China (Nos. 10802075) and the National Key Basic


Research Program of China (No. 2009CB724303).

REFERENCES
[1] Peskin C S. Numerical analysis of blood flow in the heart.
Journal of Computational Physics, 1977, 25: 220-252.
[2] Goldstein R H, Sirovich L. Modeling a no-slip flow
boundary with an external force field. Journal of
Computational Physics, 1993,105: 354-399.
[3] Mohd-Yusof J. Combined immersed-boundary /b-spline
methods for simulations of flow in complex geometries.
Annual Research Briefs, Center for turbulence research,
1997: 317-327.
[4] Fadlun E A, Verzicco R, Orlandi P, et al. Combined
immersed - boundary finite difference methods for three-

Fig.4 Free surface deformation during water entry of a dimensional complex flow simulations. Journal of

neutrally buoyant cylinder. IBM/ICM simulations (left) and Computational Physics, 2000,161: 35-60.

experiments [20] (right) [5] Mittal R, Iaccarino G. Immersed boundary methods.


Annual Review of Fluid Mechanics, 2005, 37: 239-261.
4 CONCLUSIONS
[6] Lai M C, Peskin C. An immersed boundary method with
A new extension of the immersed-boundary method is
formal second-order accuracy and reduced numerical
implemented, allowing us to simulate the interaction
viscosity. Journal of Computational Physics, 2000,160:
between moving body and gas-liquid interface. This
705-719.
new method is called IBM/ICM due to the fact that it
[7] Uhlmann M. An immersed boundary method with direct
is a combination of IBM and interface capture method.
forcing for the simulation of particulate flows. Journal of
This method and the codes employed in this paper is
Computational Physics, 2005, 209: 448-476.
far from maturity, and expected to be improved in our
[8] Yang J, Balaras E. An embedded-boundary formulation for
future work. The future work will include:
large-eddy simulation of turbulent flows interacting with
(1) A highly-efficient searching algorithm should be
moving boundaries. Journal of Computational Physics,
found to construct the links between the Eulerian cells
2006, 215: 12-40.
and Lagrangian points, especially for moving body
[9] Kim D, Choi H. Immersed-boundary method for flow
problems.
around an arbitrarily moving body. Journal of
(2) To reduce the numerical errors nearby the interface
Computational Physics, 2006, 212: 662-680.
between the solid body and the fluid, especially for
[10] Deng J, Shao X M, Ren A L. A new modification of the
the flux across the interface. Since the non-zero flux
immersed-boundary method for simulation flows with
across the interface will intrude the liquid into the
complex moving boundaries. International Journal for
body (if assumed to be filled with pseudo gas).
Numerical Method in Fluids, 2006,52: 1195-1213.
(3) To introduce a more suitable interface capture
[11] Deng J, Ren A L, Zou J F, et al. Three-dimensional flow
method into our IBM/ICM solver, for example the
around two circular cylinders in tandem arrangement.
coupled level set/volume-of-fluid (CLSVOF) method
Fluid Dynamics Research, 2006,38: 386-404.
to resolve the tiny drops.
[12] Glowinski R, Pan T W, Hesla T I, et al. A distributed
ACKNOWLEDGEMENTS Lagrange multiplier/fictitious domain method for
particulate flows. Int. J. Multiphase Flow, 1999, 25:
The authors gratefully acknowledge the financial
755-794.
support from the National Natural Science Foundation
208

[13] Glowinski R, Pan T W, Hesla T I, et al. A fictitious domain 454-493.


approach to the direct numerical simulation of [17] Guy R D, Hartenstine D A. On the accuracy of direct
incompressible viscous flow past moving rigid bodies: forcing immersed boundary methods with projection
application to particulate flow. Journal of Computational methods. Journal of Computational Physics, 2010, 229:
Physics, 2001, 169: 363-426. 2479-2496.
[14] Yu Z S, Shao X M, Wachs A. A fictitious domain method [18] Pan D, Shen T T. Computation of incompressible flows
for particulate flows with heat transfer. Journal of with immersed bodies by a simple ghost cell method.
Computational Physics, 2006, 217: 424-452. International Journal for Numerical Methods in Fluids,
[15] Le D V, Khoo B C, Peraire J. An immersed interface 2009, 60: 1378-1401.
method for viscous incompressible flows involving rigid [19] Yu Z S, Shao X M. A direct-forcing fictitious domain
and flexible boundaries. Journal of Computational Physics, method for particulate flows. Journal of Computational
2006, 220: 109-138. Physics, 2007, 227: 292-314.
[16] Xu S, Wang Z J. An immersed interface method for [20] Greenhow M., Lin W M. Nonlinear free-surface effects:
simulating the interaction of a fluid with moving experiments and theory. Report, MIT, Department of
boundaries. Journal of Computational Physics, 2006, 216: Ocean Engineering, 1983: 83-119.

You might also like