Simulation of Multiphase Flow using

Lattice Boltzmann Method

Submitted by: Saba Mushtaq
MSPE-17
Supervised by: Mrs. Romana Basit
Department of Chemical Engineering
Pakistan Institute Of Engineering And Applied Sciences

Nilore, Islamabad Pakistan
Table of Contents
Table of Contents ........................................................................................................................................... 2
Abstract .......................................................................................................................................................... 3
1 Introduction ................................................................................................................................................. 4
2 Literature Survey ........................................................................................................................................ 5
3 Lattice Boltzmann Method ......................................................................................................................... 7
4 Single Component Multiphase LBM ........................................................................................................ 10
5 Multicomponent Multiphase LBM ........................................................................................................... 13
6 Boundary Conditions ................................................................................................................................ 14
7 Results And Discussions ........................................................................................................................... 17
References .................................................................................................................................................... 28
Table of Figures
2
Figure 1: D2Q9 model ..................................................................................................................................8
Figure 2: Streaming of particles ....................................................................................................................9
Figure 3: Effect of density on interaction potentials....................................................................................11
Figure 4: Effect of pressure reduction on density for G = -48, -72, -92.4, -120, -144................................12
Figure 5: Comparison of velocity profile by LBM and Analytical.............................................................18
Figure 6: Velocity profiles for different values of ...................................................................................19
Figure 7: Density variation with number of iteration for = 0.5,1.75,1.85,1.99........................................19
Figure 8: Obstacle in flow path at x=10......................................................................................................20
Figure 9: Velocity contours for Re = 50......................................................................................................21
Figure 10: Pressure contours for Re = 50 with obstacle present at x=10....................................................21
Figure 11: Stream traces (a) Re =10, (b) Re=20, (c) Re=50, (d) Re=100, (e) Re=150 (f) Re=162............22
Figure 12: Time series plot of vapor phase separation. Results for various time steps a) = 1, b) = 100, c) =
500,d) = 1000, e) = 2000, f) = 4000, g) = 8000, h) =15000, i) 25600 ts.....................................................24
Figure 13: Density profile of a liquid droplet at time=1000 (ts).................................................................25
Figure 14: Effect of interaction strength |G| on phase separation...............................................................26
Figure 15: Pressure difference inside and outside of droplets as function of 1/r.........................................27

Abstract
The objective of this report is to demonstrate the capability of Lattice Boltzmann method
to simulate single and multiphase flow. Simulations have been performed for two
dimensional Poisuelle flow using D2Q9 model. The results obtained are compared with
analytical results and they show good agreement. Then, boundary layer separation has
been studied for various Reynolds number. Phase transition has also been demonstrated
using Shan and Chen single component multiphase model. The results were used to
compute the surface tension using Laplace law. The value of surface tension computed is
14.322 lu mu/ts
2
. This is very close to literature value (14.332 lu mu/ts
2
). The error is
only 0.06977%. After comparison with literature, it has been concluded that lattice
Boltzmann method is a valid numerical scheme emerging as an innovative computational
fluid dynamics technique.
1 Introduction
The lattice Boltzmann method (LBM) is an innovative computational fluid dynamics
(CFD) technique for simulating fluid flows and heat transfer. It has emerged as a
powerful competitor for traditional CFD methods, especially for simulations of
multiphase flows and flows involving complex boundaries. It has been used extensively
to simulate flow through porous media and blood flow simulation. Compared with the
conventional CFD approach, the LBM is easy for programming, intrinsically parallel, and
it is easy to incorporate complicated boundary conditions such as those in porous media.
It possesses several distinct features than the conventional CFD techniques. The
convection operator is linear in phase space unlike Navier Stokes (NS) equations. Also,
4
pressure is obtained through an equation of state rather than solving poisson equation.
LBM and Lattice gas cellular automata (LGCA) are based on mesoscopic kinetic
equation relating the microscopic dynamics of particles to macroscopic properties.
Historically, LBM evolves from LGCA. LGCA describes the physical system by particles
present on nodes having only two states (Boolean variables: 0 and 1). The evolution of
lattice gas models proceeds in two steps that take place during each time step. The first
step is propagation or streaming in which the particles move to new sites according to
their previous positions and their velocities. Then particles collide and scatter according
to the collision rules . Momentum and energy remain conserved during both steps. The
major problem in LGCA is the presence of noise which causes a problem in achieving
high accuracy. In LBM, it has been removed by describing the physical system using
continuous distribution function rather than using Boolean variables. Distribution
function is defined as the probability of finding a particular molecule with a certain
position and momentum at certain time. The physical system is again comprises of the
same steps: streaming and collision.
1.1 Research Objectives
In the 3rd and 4th semester, poisuelle and single component multiphase (SCMP) flows
have been simulated using D2Q9 LBM model. Poisuelle flow is the simplest flow that can
be simulated using LBM. Boundary layer separation has also been observed for various
Reynold numbers. SCMP flow has been simulated using the most popular Shan and Chen
model used. Surface tension in this two phase mixture has also been computed. In the
research semester, simulation of multicomponent multiphase flow will be performed and
its surface tension will be computed.
2 Literature Survey
Shan and Chen (1994) has described a model to simulate flows involving multiple
phases and multiple components. However, the focus was on single component which
obey non ideal gas equation and undergo phase transition process. They have also
obtained density profile in the liquid-gas interface as function of the temperature-like
parameter .
Multicomponent multiphase flow is of tremendous importance in petroleum industry.
Bulk-flow simulations of reservoir-flow performance are an important ingredient in the
decision of whether to invest in major recovery projects . Buckles et al (1994) have
carried out the simulations of an oil recovery project using the lattice Boltzmann model.
Multiple components involved in the system are oil and water. They have plotted relative
permeabilities for the wetting fluid (oil) and the nonwetting fluid (water) as a function of
the wetting-fluid saturation for different values of the capillary number. The concluded
that as saturation of the wetting fluid increases (the fraction of pore space it occupies) its
relative permeability increases, that is, the fluid flows more easily. Similarly, as the
saturation of the nonwetting fluid decreases its relative permeability decreases.
The lattice Boltzmann equation (LBE) has been directly derived from the continuous
Boltzmann equation by Xiaoyi and Luo (1997). They have demonstrated that the LBE is
a special discretized form of the Boltzmann equation. A general procedure to derive the
lattice Boltzmann model from the Boltzmann equation have also been described. The
lattice Boltzmann models derived include the two-dimensional 6-bit, 7-bit, and 9-bit, and
three-dimensional 27-bit models
Martys and Douglas (2001) have calculated the equilibrium properties of LB fluid
mixtures to characterize the critical phenomenon occurring in these processes . They have
also studied the phase separation under quiescent and shearing conditions.
Ladd and Verberg (2001) have applied LBM to particle fluid suspensions. They have
summarized the available methods along with applications for simulations of colloidal
suspensions. Finally, they have also derived the equations for LBM that includes
transverse and longitudinal fluctuations in momentum .
Baochang (2007) has directly applied the lattice Boltzmann method for convection
diffusion equation with source term to solve some important nonlinear complex
equations, including nonlinear Schrodinger (NLS) equation, Klein-Gordon equation etc.
by using complex-valued distribution function and relaxation time. He carried out their
detailed numerical simulation and his results agree well with the analytical solution .
Physical behavior of a number of mesoscopic models for multiphase flow have been
analyzed by Sbragaglia et al (2007). They have developed an extended pseudopotential
model which permits to tune the equation of state and surface tension independently of
each other . Their model has also enhanced the flexibility of LBM to high density ratios.
Huang et al (2008) have studied the viscous coupling effects for the simulation of
immiscible two phase flow through porous media using Shan and Chen type multiphase
LBM. They have investigated coupling effects due to capillary number, viscosity ratio
and wetting angle. They have observed that when the wetting phase is less viscous and
covers the solid surface, the relative permeability of the non-wetting phase may be greater
than unity .
6
Gokaltun and McDaniel (2010) have studied the feasibility of multiple relaxation time
LBM for multiphase flow simulations The LBM becomes unstable when simulating high
density ratio and low viscosity multiphase flows. This instability has been avoided by
using multiple relaxation time rather than single relaxation time.
Nwatchok et al (2010) have developed an LBM based model to solve conservation
equations for momentum for Poisuelle Rayeigh Bernard (PRB) mixed flow. In PRB
mixed flow, there exists longitudinal pressure gradient in a horizontal rectangular channel
which is heated from below and cooled from above. PRB mixed flow simulation is an
active and vast topic of research
Myre et al (2010) have examined the performance of single component single phase
and multicomponent multiphase flows on Graphics processing unit (GPU) clusters. They
have concluded that all LBM simulations depend on effects corresponding to simulation
geometry and not on the architectural aspects of GPU .
3 Lattice Boltzmann Method
LBM is based on Boltzmann equation which describes the system by continuous
distribution function. It discretizes continuous Boltzmann equation to reduce the number
of possible particle spatial positions, microscopic momenta and time. Particle positions
are confined to the nodes of a lattice. Variations in momenta that could have been due to a
continuum of velocity directions, magnitudes and varying particle mass are reduced to
eight directions, 3 magnitudes and a single particle mass in a D2Q9 model .
D2Q9 model means 2 dimensional and having 9 velocity vectors as shown in Figure 1.
Point (0,0) represents the coordinates of particles which are at rest. Points (1,0), (0,1), (-
1,0), (0,-1) shows the position of particles which are present on coordinate axis, whereas
(1,1), (-1,1), (-1,-1), (1,-1) shows the particles present along diagonals. The fundamental
measure of length is lattice unit (lu), unit of time is time steps (ts) and of mass is mass
unit (mu). Mass of particles has uniformly taken as 1 mu throughout the domain. The
particles present on (0,0) coordinates are stationary. The velocity magnitude of 1 through
4 velocity vectors is 1 lu/ts and that of index 5 to 8 is
2
lu/ts. These velocities are
exceptionally convenient in that all of their x and y-components are either 0 or 1 .
Figure 1: D2Q9 model
The Boltzmann equation without external force term (which has been neglected
because of its lesser magnitude in comparison with intermolecular force) is given by

f
v f Q
t

+
r
(1)
Where,
f
is the distribution function,
v
r
is the velocity vector and
Q
is the collision
integral. Applying the BGK approximation to the collision integral
( )
1

eq
f
v f f f
t

+
r
(2)
Where

is known as the collision time and

eq
f is the equilibrium distribution
function. The distribution function depends on space, momenta and time. Since mass unit
has been taken as unity, so momenta is equivalent to the velocity. The velocity is
discretized to 9 directions in a D2Q9 resulting in a set of 9 distribution functions at any
point at a given time. The discretized equation is
( )
1

eq i
i i i
f
v f f f
t

+
r
(3)
Two steps involved in LBM are:
1) Streaming
2) Collision
In streaming direction specific densities (distribution functions) are moved to the
nearest neighbor lattice nodes. The streaming of distribution functions from one node to
the other is shown in Figure 2.

8
Figure 2: Streaming of particles
In collision, distribution functions collide with each other to get new distribution while
conserving momentum. Both steps can be represented by following mathematical
expression
( ) ( ) ( )
1
, , , ( , )
Streaming
Collision
eq
i i i i
f t t t f t f t f t

1 + +
]
i
x e x x x
1 4 4 4 442 4 4 4 4 43
1 4 4 4 2 4 4 4 3
(4)
Where
i
e
indicates the microscopic velocity of ith particle. Collision of the fluid
particles is considered as relaxation towards local equilibrium. This collision step is
skipped at solid nodes, because bounceback boundary condition which is applied on solid
nodes, is also a type of collision. The equilibrium distribution function
eq
i
f is defined as,
( )
( )
2
2
( )
2 4 2
. . 9 3
1 3
2 2
i eq i
i i
f w
c c c

1
+ + 1
1
]
e u e u u
x
(5)
Where the weights
i
w
are 4 9 for the rest particles ( 0 i ),1 9 for { } 1, 2, 3, 4 i
and
1 36 for { } 5, 6, 7, 8 i
. c represents the basic speed on lattice (in the simple LBM
models 1 c
1
luts

) and

stands for macroscopic density.

The macroscopic variables are calculated by using the following formulas:
Macroscopic density

8
0
i
i
f

(6)
Macroscopic velocity
u
8
0
1
i i
i
f

u e
(7)
Pressure P
P RT
(8)
Kinematic viscosity

1 1
3 2

_


,
(9)
Relaxation time

should be greater than 1/ 2 to get physically realistic results so, the

relaxation parameter which is defined as the inverse of

should be less than 2. The

values of distribution functions and equilibrium distribution functions should be positive
to meet the consistency requirements. Unlike traditional CFD techniques, pressure has
been computed by using ideal gas law rather than using Poisson equation. Moreover,
pressure does not depend explicitly on velocity.
4 Single Component Multiphase LBM
LBM ha capability to simulate single component single phase (SCSP), single
component multiphase (SCMP) and multicomponent multiphase (MCMP) flows. There
are many LBM models to simulate multiphase flow, most of them are very difficult to
apply. Here, the method proposed by Shan and Chen has been described because of its
relative ease when compared with other methods. Also, in this model, separation of two
phases takes place spontaneously whenever the magnitude of interaction strength |G|
exceeds a certain threshold value. Since, it is easy to use so it is very popular among
programmers . The other LBM models are relatively complex and difficult to apply.
In SCMP flow, an attractive force is introduced between the particles through
interaction potential and interaction strength. This force leads to a reduction in pressure in
accordance with the van der Waals Equation of state. The reduced pressure causes the
separation. On subatomic level, it can be justified as the repulsion caused by electron
cloud when they are very close to each other under the influence of external attractive
force.
In this model, equilibrium distribution function is computed as
( )
( )
2
2
( )
2 4 2
.
. 9 3
, 1 3
2 2
eq
eq eq
i
eq i
i i
f t w
c c c

1
1
+ +
1
]
e u
e u u
x
(10)
Where, the equilibrium velocity
eq
u is calculated as
10
eq

+
F
u u
(11)
The velocity
u is defined as
8
0
i i
i
f

e
u
(12)
The interparticle force Fis given by
( )
8
0
( , ) ( , ) ,
i i i
i
t G t w t t

F x x x e e
(13)
Where G is the interaction strength, being negative for interparticle attraction.
i
w
is 1/9
for { } 1, 2, 3, 4 i
, for { } 1, 2, 3, 4 i
is 1/36 for and

is the interaction potential:

( )
0
0
exp

,
(14)
0

and
0

Being the arbitrary constants. Here, the values of

0

and
0

has been
arbitrarily chosen as 4 and 200 respectively, because the model behavior with these values
has been explored thoroughly. According to equation (14) the attractive force is stronger
when density is higher, and vice versa. Hence, the liquid regions experience stronger
cohesive force than the vapor leading to the surface tension phenomenon . The variation
of

with

is shown in Figure 3.
Figure 3: Effect of density on interaction potentials
The pressure is calculated by using the following non-linear and non-ideal equation of
state:
( )
2
2
GRT
P RT + 1
]
(15)
The value of RT is fixed for both SCSP and SCMP models:
1
3
RT (16)
The second term on the right hand side of is non-ideal part that stands for attractive
force between the molecules. Since G is negative, it leads to the reduction of pressure.
When G is adequately negative, the EOS is non-monotonic that leads to the phase
separation. Using (16), the equation for pressure becomes
( )
2
6
RT
P RT + (17)
The pressure density relationship is demonstrated in Figure 4 for G = -48, -72, -92.4
(critical), -120, -144. It has been observed that when |G| exceeds a critical value of 92.4,
there corresponding to each pressure there exists two values of densities; the higher one
showing the density of liquid and lower indicating vapor density.
Figure 4: Effect of pressure reduction on density for G = -48, -72, -92.4, -120, -144
Here, only attractive forces has been considered and repulsive forces has been ignored
which become dominant when a gas is present in compressed form.
12
5 Multicomponent Multiphase LBM
In Multicomponent multiphase (MCMP) system, more than one species are combined
through interaction forces (repulsive) to form immiscible fluids. MCMP flows are of
tremendous economic and environmental importance. Their economic importance is due
to the fact that petroleum is found with water forming a MCMP system. Non aqueous
phase liquids present in the subsurface and acting as a long lived source of ground water
contamination contribute to the environmental importance of MCMP.
To simplify the system, only two components have been considered here. In LBM
algorithm, the addition of a second component is incorporated by the the addition of a
new array and a new loop. The equilibrium distribution function is calculated using the
following composite macroscopic velocityu
1
1
a a
a
a
f
u
f

e
(18)
The composite macroscopic velocity represents the flow of the bulk of fluid and differs
from the macroscopic uncoupled velocities of the individual components
u

8
0
1
a a
a
u f

e
(19)
Where,

denotes the fluid components and

= {1,2} for two component mixture.

The macroscopic density of each fluid component is given by
8
0
a
a
f

(20)
The density of both components cannot be zero at any point in the calculation domain,
otherwise physically unreakistic values of velocities will be obtained.
To obtain equilibrium velocities
eq
u , the gravity (body force) and interaction forcs
are added to u
. The repulsive interaction forc on the fluid component

is given by
( ) ( , ) ( , )
a a a
a
F x G x t w x t t

+

e e
(21)
Where, G is the interaction strength,

is the one fluid component and denotes the

other component.

and

are usually taken as the fluid densities such as

and


. The addition of this interaction forc term
F

to u
is
eq
u u F

+ (22)
eq
u F
u

(23)
The velocity increment depends on the magnitude of G and fluid densities. The
magnitude of the velocity increment in the above Equation should be kept small.
6 Boundary Conditions
The dynamics of fluid flow depends on the surrounding environment. This influence is
described via boundary conditions. So, Boundary conditions play a crucial role in
computing any meaningful results. Periodic, Bounce back and Dirichlet boundaries have
been described below.
6.1 Periodic boundary Conditions
Periodic boundary conditions are the simplest boundary conditions because system
becomes closed by the edges as if they are attached with each other. They are sufficient
where surface effects are negligible. In flow simulation in a slit, periodic boundary
condition is applied at the open end of the slit and bounce back boundary condition is
applied at the slit walls.
Periodic boundary condition is implemented in streamwise x direction by treating the
nodes on the inflow and outflow faces as the neighbors if they share common y and z
coordinates. For boundary nodes, the neighboring nodes are present on the opposite
boundary. Here are some conditionals that check if the neighboring nodes lie outside the
computational domain and then assign the appropriate node on the opposite boundary in
that case to achieve periodicity. Pseudo code to implement Periodic boundary condition in
x direction is described below (1 is the first node in both x and y direction and lx and ly
are the last nodes in x and in y direction respectively).
( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
do y=1,ly
f 1,1,y f 1,lx,y
f 5,1,y f 5,lx,y
f 8,1,y f 1,lx,y
f 3,1x,y f 3,l,y
f 6,1x,y f 6,l,y
f 3,1x,y f 7,l,y
enddo

(24)
Using this convention, the populations on all incoming links at inflow and outflow is
defined automatically by streaming process. Periodic boundary condition is imposed as
14
the natural part of streaming so that outgoing populations at one end becomes the
incoming populations at the other end.
Fully periodic boundary conditions are also applied in some cases e.g. the simulation
of multiphase flow in an infinite domain.
6.2 Bounce Back Boundary Condition
The next simplest boundary condition is the bounceback boundary condition, also
known as no-slip boundary condition. It plays a major role in making LBM popular
because it can conveniently handle the complex boundaries such as those present in
porous media. It specifies that velocity of fluid at the wall is same as that of wall.
In programming, one just has to declare a particular node as the solid object and no
special treatment is necessary. Thus it is easy to incorporate images of porous media and
immediately compute flow in them. Solids nodes are separated into two categories:
1. Boundary solids
2. Isolated solids
Boundary solids lie at the solid-fluid interface and isolated solids are those that has no
contact with fluid. By using this division it is possible to eliminate unnecessary
computations at inactive nodes; this is very important in simulating flow in fractured
media where fraction of total space occupied by open space in contact with fluids, which
needs to program is very small.
It has been assumed that solid surface is aligned with the grid. Two types of
implementations will be discussed:
1. On-grid or Full way bounceback
2. Mid grid or Half way bounceback
On-grid means that physical boundary lies exactly on grid line, whereas, in mid-grid
condition the boundary lies in between two grid lines. The first one is easy as it just
reverse all the populations sitting on the boundary node. It simply implies that for any
domain
( ) ( )
in out
f N =f N
( ) ( )
in out
f S =f S
(25)
where N and S represent north and south rows of the domain. For a D2Q9 model, it is
implemented as:
( ) ( )
( ) ( )
( ) ( )
f 6,x,y =f 2,x,y
f 7,x,y =f 3,x,y
f 8,x,y =f 4,x,y
(26)
The on-grid bounceback possess only first order accuracy because of of sided character
of streaming process at the boundary. Second order accuracy can be achieved by using the
mid grid or half way bounceback condition, which is modestly more complex than Full
way bounce back.
It is implemented as:
( ) ( )
in out
f N =f NF
( ) ( )
in out
f S =f SF
(27)
Where N and S represent a generic site on the top and bottom walls, respectively,
whereas NF,SF is the set of fluid sites connected to N,S respectively.
For a D2Q9 model it is implemented as:
( ) ( )
( ) ( )
( ) ( )
f 6,x,y = f 2,x-1,y-1
f 7,x,y = f 3,x,y-1
f 8,x,y = f 4,x+1,y-1
(28)
6.3 Dirichlet Boundary Condition
In real flow experiment, some form of regulation at inlet and outlet is needed. A
common type involves the constant pressure difference between the inlet and the outlet.
The Dirichlet boundary condition constrains the pressure/density at an inlet or outlet
boundary. To impose it, it is assumed that inflow face is oriented normal to an axis of the
lattice. It is also assumed that velocity tangent to boundary is zero and the normal to the
boundary component of velocity is evaluated.
First of all, a density
o

is specified, using it then the velocity is computed. Density

specification is equivalent to pressure specification because both are linked by equation of
state (Which for single component D2Q9 model is
P RT
, whereas RT=1/3).
In addition to the macroscopic velocity, the proper distribution function at the
boundary nodes needs to be determined. After the streaming there are three unknown
directional densities at each lattice node that point from boundary into the fluid. These
three unknowns can be solved in a way to maintain the specified pressure/density
o

at
their lattice nodes.
16
6.4 Von Neumann Boundary Condition
Von Neumann boundary condition specifies the flux at the boundary. A velocity vector
containing x and y components [ ]
0 0
= u v u
is specified from which density/pressure is
computed. Macroscopic density/pressure is only part of what needs to be computed .
After streaming there there are three unknown directional densities
4 7 8
, and f f f
at each
lattice node (pointing into the domain). These three unknowns are solved in a way to
maintain the specified velocity at lattice nodes. Then density is computed by the
summation of all distribution functions
.
7 Results And Discussions
The simplest LBM that has been implemented is for SCSP and SCMP. The
programming language used is FORTRAN and lattice model is D2Q9. TECPLOT 360
has been used as post processing tool. Here the simulation results have been presented for
poisuelle flow and multiphase flow.
7.1 Plane poisuelle Flow
The viscous flow through a channel under the action of pressure gradient is called
poisuelle flow. The flow is one dimensional incompressible and laminar. The analytical
solution gives a parabolic velocity profile with maximum velocity in the center and zero
at the walls of the channel.
In LBM, poisuelle flow has been simulated using a uniform grid, which contains thirty
lattice units (lu) in x-direction and twenty in y direction. Bounceback boundary condition
has been applied along the walls. In the flow direction Periodic boundaries has been used,
which means the fluid leaving the domain reenters in the opposite end of the channel. The
system is effectively infinite in the flow direction and there are no end effects . The time
length is 3000 time steps. The fluid is resting in the beginning and is then slowly
accelerated . The value of relaxation parameter

(the inverse of collision time) is 1.85.

The value of initial density distribution function chosen is 0.1. Figure 5 shows the
comparison of LBM and analytical results.
Parabolic velocity profile obtained show excellent agreement with analytical results of
poisuelle flow. Only slight difference has been observed at the wall due to the use of full
way bounce back boundary condition which shows first order accuracy. It can be
improved using half way bounce back boundary condition, being the second order
accurate. In the full way bounce back boundary condition the physical boundary (wall)
lies exactly on a grid line, whereas in the half way bounce back boundary condition, the
boundary lies between two grid lines.
Figure 5: Comparison of velocity profile by LBM and Analytical
To see the effect of which is a convergence parameter, velocity profiles have been
plotted for various values of , as shown in Figure 6.
18
Figure 6: Velocity profiles for different values of
It is observed that when the value of is closet to zero the magnitude of velocity obtained
is very less than actual whereas, when approaches two, the flow becomes unstable. The
value of very close to 0 shows infinite viscosity beyond the limit simulated by LBM.
The value of greater than two represents negative viscosity. Both extremes lead to
stability issues in LBM model. So, the appropriate choice of lies between one and two.
The variation of average density with time steps has been shown in Figure 7 for 4 values
of . It is observed that when is close zero, the average velocity obtained is negligible,
due to viscous effects. When is near two, the flow becomes unstable and no steady state
is achieved. When lies between one and two ( = 1.75, 1,85), a steady state is achieved
after 3000 iterations.
Fluid flowing past an object tends to drag the object in the direction of fluid flow. If
an object is moving through a stationary fluid the drag tends to slow down it and if body
is stationary in flowing fluid the drag tends to move it in the direction of flow. Boundary
layer separation occurs when fluid strikes with a blunt body such as flat plate
perpendicular to the flow.
-2.00E-02
0.00E+00
2.00E-02
4.00E-02
6.00E-02
8.00E-02
1.00E-01
1.20E-01
1.40E-01
1.60E-01
0 500 1000 1500 2000 2500 3000 3500
Avg velocity(lu/ts)
No. of iterations(ts)
w=0.5
w=1.75
w=1.85
w=1.99
Figure 7: Density variation with number of iteration for = 0.5,1.75,1.85,1.99
To observe the boundary layer separation, the obstacle has been introduced in the flow,
perpendicular to the flow at x=10 lu and the grid size has been changed to 5020 lattice to
minimize the flow disturbance as shown in Figure 8.
X (lu)
Y
(
l
u
)
10 20 30 40 50
5
10
15
20
OBST
0.95
0.9
0.85
0.8
0.75
0.7
0.65
0.6
0.55
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
Figure 8: Obstacle in flow path at x=10
The contours of velocity for Re = 50, with obstacle present at x=10, are shown in
Figure 9.
X (lu)
Y
(
l
u
)
10 20 30 40 50
5
10
15
20
VX
0.14
0.13
0.12
0.11
0.10
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
20
Figure 9: Velocity contours for Re = 50
The velocity of fluid when it strikes with obstacle becomes zero and is converted into
pressure. Hence, pressure is maximum at x = 10. Along the sides of the obstacle, since
flow area has been reduced now, to preserve continuity equation, the velocity has
increased to maximum. Just behind the obstacle, a low pressure zone is formed where
boundary layer has been separated along with formation of vortices. The kinetic energy of
the fluid has been used in vortex formation. The pressure difference between the front and
rear edge of obstacle has been illustrated in Figure 10. When the fluid strikes the obstacle,
the pressure is highest at that point, whereas just behind the obstacle a low pressure zone
is formed.
X (lu)
Y
(
l
u
)
10 20 30 40 50
5
10
15
20
PRESS
0.0343
0.0342
0.0341
0.034
0.0339
0.0338
0.0337
0.0336
0.0335
0.0334
0.0333
0.0332
0.0331
0.033
0.0329
0.0328
Figure 10: Pressure contours for Re = 50 with obstacle present at x=10
The boundary layer separation has been studied for various Reynolds numbers. The
stream traces are shown in Figure 11 for Re = 10,20,50,100,150,162. At Re = 10,
streamlines separate and recombine after the obstacle ends with no vortex formation. At
Re = 20, little vortex formation has been observed on downstream side. As Re is further
increased, vortex grows in size at the rear end. Finally, when Re has been increased from
162, instability in stream line flow has been observed. This instability is due to the
presence of large size of the obstacle.
Figure 11: Stream traces (a) Re =10, (b) Re=20, (c) Re=50, (d) Re=100, (e) Re=150 (f)
Re=162
22
a)
X (lu)
Y
(
l
u
)
10 20 30 40 50
5
10
15
20
d)
X (lu)
Y
(
l
u
)
10 20 30 40 50
5
10
15
20
b)
X (lu)
Y
(
l
u
)
10 20 30 40 50
5
10
15
20
e)
X (lu)
Y
(
l
u
)
10 20 30 40 50
5
10
15
20
c)
X (lu)
Y
(
l
u
)
10 20 30 40 50
5
10
15
20
f)
X (lu)
Y
(
l
u
)
10 20 30 40 50
5
10
15
20
7.2 SCMP Flow
Simulation of multiphase flow is an interesting and challenging task. Uniform grid has
been used having 200X200 lattice nodes. Initial density introduced is 200 lu/ts
2
plus
perturbation induced by a random number between 0 and 1. The value of interaction
strength G chosen is -120, which leads to the phase separation process. The relaxation
factor chosen is 1 (ts). Periodic boundaries have been used in all directions. The time
length is 25600 (ts). The phase separation process for various time intervals is shown in
Figure 12.
a) b)
c)
d) e) f)
g) h)
i)
The red color is indicating the denser phase (liquid) and blue phase is showing the
lighter phase (vapor).
The molecules sitting on the interface experience a net force driven by the
intermolecular distance between the two fluids , and as a result like molecules start
moving towards each other. It can be observed that the smaller droplets are relatively
more unstable than the large ones. This is due to the surface tension effects on the
interface with a larger curvature. A large pressure difference occurs in the smaller
droplets causing them to evaporate
.
Phase separation ultimately leads to a single droplet
in vapor atmosphere. Whether liquid drops or vapor bubbles are formed depends on the
24
Figure 12: Time series plot of vapor phase separation. Results for various
time steps a) = 1, b) = 100, c) = 500,d) = 1000, e) = 2000, f) = 4000, g) =
8000, h) =15000, i) 25600 ts
total mass in the domain and consequently on the initial density selected. When phase
separation occurs then the interfaces try to minimize their total area thus achieving the
shape of a circle. Fig. 13 illustrates the density profile of a droplet which is immersed in
its own vapors. At the vapor liquid interface the density changes from minimum to
maximum.
Figure 13: Density profile of a liquid droplet at time=1000 (ts)

Effect of interaction strength on phase separation process has been shown in Figure 14.
Below the critical value of |G| there is only one phase present. As the value of |G| exceeds
this threshold value, in this case, it is 92.4, two phases form having different densities.
The one having minimum density is the vapor phase and with maximum density is the
liquid phase.
Figure 14: Effect of interaction strength |G| on phase separation
Surface tension is very important for multiphase flow system. Here, it has been
calculated using Laplace law, which states that, if a liquid droplet of radius r is immersed
in its own vapors, pressure difference (
outside inside
P P P
) across the droplet is related
to the radius of the droplet by the following relationship where

denotes the surface

tension.
P
r

(29)
According to Laplace law, plot of P and 1/r should be linear and its slop represents the
surface tension. The surface tension

has been computed by measuring the radii of

droplets of various sizes, immersed in their own vapors and their pressure. The plot of P
versus 1/r is shown in Figure 15, which is linear, as stated by Laplace law. The surface
tension which is the slope of the straight line comes out to be 14.322 lu mu/ts
2
. The value
of surface tension found in literature is 14.332 lu mu/ts
2
. The error is only 0.06977%.
26
Figure 15: Pressure difference inside and outside of droplets as function of 1/r
8 CONCLUSIONS
In this report, LBM has been introduced as a competitive CFD tool to simulate the
poisuelle flow and phase transition process. For poisuelle flow, simulation has been
performed using a 3020 lattice. The velocity profiles has been studied for vaious values of

and optimum value of

has been found to be lying between 1 and two. The results of

poisuelle flow are in excellent agreement with analytical results for

=1.85.
The simulation results for SCMP have been depicted by using Shan and Chen SCMP
model, are also in good agreement with results present in literature. Phenomenon of phase
separation has been shown for various values of interaction strength which causes
reduction in pressure and hence leading to phase separation phenomenon. Density profile
in a droplet present in the vapor atmosphere has also been shown. Density shows an
abrubpt change at the interface. Surface tension computed also match quite well with the
literature values. The value of surface tension computed is 14.322 lu mu/ts
2
. This is very
close to literature value (14.332 lu mu/ts
2
). The error is only 0.06977%.
Although SCSP and SCMP flow have been simulated, LBM has capability to simulate
MCMP. The simulation of MCMP flow using LBM will be performed in the research
semester.
References
[1]. Michael C.Sukop, Daniel T.Thorne, Jr., Lattice Boltzmann Modeling An
Introduction for Geoscientists and Engineers, Springer Verlag Berlin Heidelberg,
2006.
[2]. Shan X and Chen ,Simulation of nonideal gases and liquid-gas phase transitions by
the lattice Boltzmann equation, APS, 49, 2941-2948, 1994
[3]. Buckles et al ,Toward improved prediction of reservoir flow performance, Las
Alamos science, 22, 112-121, 1994
[4]. Xiaoyi H and Luo L.S, Theory of the lattice Boltzmann method: From the
Boltzmann equation to the lattice Boltzmann equation, APS Physics, 56,6811-
6817,1997
[5]. Martys N.S and Douglas J.F, Critical properties and phase separation in lattice
Boltzmann fluid mixtures , Phys. Rev. E.63 2001
[6]. Ladd A.J.C. and Verberg R, Lattice-Boltzmann Simulations of Particle-Fluid
Suspensions, Journal of Statistical Physics, 104 2001
[7]. Baochang S, Lattice Boltzmann Simulation of Some Nonlinear Complex Equations
ICCS, 818 825 2007
[8]. Sbragaglia, et al, Generalized lattice Boltzmann method with multirange
pseudopotential, APS, 75 2007
[9]. Huang et al, Shan-and-Chen-type multiphase lattice Boltzmann study of viscous
coupling effects for two-phase flow in porous media, Int. J. Numer. Meth. Fluids,
2008
[10]. Gokaltun S and McDaniel D, Multiple-Relaxation-Time Lattice Boltzmann
Method for Multiphase Flows with High Density and Viscosity Ratios, WM
Conference, 2010
[11]. Nwatchok et al, Lattice Boltzmann simulation of the two-dimensional Poiseuille-
Rayleigh-Benard flows instability, International Journal of the Physical Sciences,
5,984-991, 2010
[12]. Myre et al, Performance analysis of single-phase, multiphase, and
multicomponent lattice-Boltzmann fluid flow simulations on GPU clusters,
Concurrency and Computation: Practice and Experience, 23, 332-350, 2011
[13]. Dieter A. Wolf Gladrow, Lattice Gas Cellular Automata and Lattice Boltzmann
Models, Lectures Notes in Mathematics, 1st Edition, Springer Verlag Berlin
Heidelberg, New York, 2000.
28
[14]. Http:Developer.nvidia.com/GPUGems2/gpugems2_chapter47.html.(retrieved on
02-05-2012).
[15]. Sauro Succi, The Lattice Boltzmann Equation for Fluid Dynamics and Beyond,
Clar-endon Press Oxford, 2001
[16]. Begum, R., Basit, M. A., Lattice Boltzmann Method and its Applications to Fluid
Flow Problems, EJSR, 22, 216-231 2008
[17]. Begum, R., Basit, M. A., Simulation of phase separation Process using Lattice
Boltzmann method, CMNSEM, 1, 2010.