You are on page 1of 8

ME5361

NATIONAL UNIVERSITY OF SINGAPORE



FACULTY OF ENGINEERING


EXAMINATION FOR
(Semester 2 : 2011-2012)

ME5361 ADVANCED COMPUTATIONAL FLUID DYNAMICS


APRIL/MAY 2012 - Time Allowed: 2.5 Hours


___________________________________________________________________________


INSTRUCTIONS TO CANDIDATES:


1. This examination paper contains FOUR (4) questions and comprises EIGHT (8)
printed pages.


2. Answer All FOUR (4) questions.


3. All questions carry equal marks.


4. This is a CLOSED BOOK EXAMINATION.
Pg 2 ME5361

Question 1


(a) For a Steady, quasi one-dimensional, isothermal, compressible flow of an ideal gas in
a constant area duct with wall friction, as shown in the schematic below:











Starting from mass and momentum balance show that:

dp
p
pdp
A m RT
dx
D
f 2
) / (
2
4
2
+ =




Hence, or otherwise, show that:

If P decreases in the flow direction:

If P increases in the flow direction:

And:

If P decreases in the flow direction:


If P increases in the flow direction:

(8 marks)


Pg 3 ME5361

(b) Consider the Heat Diffusion equation :




Applying explicit, central difference discretization yields the following:


) u u 2 u (
x
t
u u
n
1 i
n
i
n
1 i
2
n
i
1 n
i +
+
+ + =
A
A o



Use Von-Neumann Stability analysis to establish the stability criteria (8 marks)

If we change to IMPLICIT scheme, what will be the stability criteria? (9 marks)

Pg 4 ME5361

Question 2


Considering the following 1-D Euler Equation:







Apply the Lax-Wendroff technique for the stencil as shown:

x
F
A
x
F
A
x
t
x
F
t
x t
F t
t
U
t U U
n
i
n
i
c
c
=
c
c
c
c A
+
c
c
A =
c c
c A
+
c
c
A =
+
where
) (
2 2
2 2 2
1


(a) Using a general second order backward and forward interpolation as follow to evaluate the
flux at ,
2
1
+ i and ,
2
1
i hence, derive the expression for
i
x
F
|
.
|
c
c


backward interpolation: ) (
2
1
1 2 / 1
+

+ + +
+
+ =
i i i
b
i
F F F F
forward interpolation: ) (
2
1
1 2 2 / 1

+

+

+
=
i i i
f
i
F F F F
Flux contribution:
b
i
f
i
b
i
F F F
+
+

+
+
+
+ =
2 / 1 2 / 1 2 / 1


(10 marks)


(b) Using flux splitting, derive an expression for |
.
|

\
|
c
c c
x
F
A
x



hint: ) ( ) ( [
1
1 1

+
+

+ +
+
A
= |
.
|
c
c
i i i i i i
i
F F A F F A
x x
F
A ]
(10 marks)

(c) Hence, or otherwise, derive an express for
n
i
n
i
U U
+1

(5 marks)
0 =
c
c
+
c
c
x
F
t
U
Pg 5 ME5361

Question 3


(a) Consider a two-dimensional partial differential equation (PDE) on an irregular
domain with curved boundary. When the second order finite difference scheme is
applied to solve this problem, discuss the need of coordinate transformation. This
problem can also be solved by the domain-free discretization (DFD) method without
introducing the coordinate transformation. Discuss the difference of DFD
discretization strategy and its basic solution procedure.
(5 marks)


(b) As shown in the following figure,


O A B
C
h
h
h




there are 4 mesh points, O, A, B, C, in a computational domain. The distance from
mesh points A, B and C to mesh point O is kept the same and noted as h. Suppose
that the origin of the coordinate system is placed at mesh point A. Using conventional
finite difference scheme, the first order derivative of a function u with respect to x at
point A can be approximated by the following form,

A A O O B B
A
u
a u a u a u
x
c
= + +
c
(3-1)

Determine the weighting coefficients , ,
A O B
a a a , and show the order of accuracy of
the scheme. On the other hand, the above derivative can also be approximated by the
least square-based finite difference (LSFD) scheme, which involves 4 mesh points O,
A, B, C with the following form,

A A B B C C O O
A
u
b u b u b u b u
x
c
= + + +
c
(3-2)

Determine the weighting coefficients , , ,
A B C O
b b b b . Hint: You need to apply two-
dimensional truncated Taylor series expansion (truncated to the first order
derivatives) and least square optimization.

(12 marks)
Pg 6 ME5361


(c) Suppose that the solution around point A is approximated by the following
multiquadrics (MQ),


( )
( )
2 2 2 2 2 2
1
2 2 2 2 2 2
2 3
( , ) ( )
( ) ( )
u x y c x h y c x y c
c x h y h c x y c c
= + + + + +
+ + + + +
, (3-3)


where
1 2 3
, , c c c are constants, c is the shape parameter, and the first order derivative at
point A is approximated by


A A O O C C
A
u
e u e u e u
x
c
= + +
c
(3-4)


Using the idea of local RBF-DQ method, the three weighting coefficients , ,
A O C
e e e
can be determined as

1
A
e
O
=
O
;
2
O
e
O
=
O
;
3
C
e
O
=
O
(3-5)


where O, O
1
, O
2
, O
3
are respectively the determinants of 33 dimensional matrices.
Write out explicitly the expressions of O, O
1
, O
2
and O
3
. You can use the matrix form
in your expression.
(8 marks)

Pg 7 ME5361

Question 4


(a) Flow pattern is a physical phenomenon. It can be simulated by solving Navier-Stokes
(N-S) equations. On the other hand, it can also be simulated by solving the following
lattice Boltzmann equation with BGK collision model,

( ) ( ) ( ) ( ) ( ) ( ) M i t f t f t f t f
eq
i i i t t i i
, , 1 , 0 , ,
1
, , = = + + r r r e r
t
o o (4-1)

where
i
f is the density distribution function, which depends on position r in the
physical space, the particle discrete velocity
i
e and time t;
eq
i
f is its corresponding
equilibrium state, which depends on the local macroscopic variables and u; t is
the single relaxation parameter related to the hydrodynamic viscosity,
t
o is the time
step and M is the number of discrete particle velocity. Discuss how to compute the
macroscopic variables and u from the density distribution function
i
f .
Furthermore, as the physical problem (flow pattern) is the same, discuss how to
guarantee that the solution given from (4-1) is the same as the solution of N-S
equations.
(5 marks)


(b) For the two-dimensional case, the following D2Q7 lattice velocity model is often
used.
1
2 3
4
5
6

The lattice velocity of D2Q7 model can be written as

( ) ( ) ( )
0 0
1 3 1 3 12 3 4 5 6 cos , sin , , , , ,
i
i
i i c i t t
=

=

( ( =

e
(4-2)

and its equilibrium distribution function has the following form

2 2
2 4
1
2
i i s eq i
i i
s s
e e u u c u
e u
f w
c c
o | o |
o o

| |
= + +
|
|
\ .
(4-3)

Pg 8 ME5361

where
0
1 2 1 12 12 3 4 5 6 / ; / , , , , , ,
i
w w i = = = . Start from Maxwell function

( )
( )
(


=

RT
RT t f
D eq
2
u c
exp 2 ) , c , r (
2
2 /
t , (4-4)

and explain why in the D2Q7 lattice velocity model, 12 3 4 5 6 , , , , , ,
i
w i = are taken the
same.
(5 marks)

(b) Suppose that the standard lattice Boltzmann method (LBM) with D2Q9 lattice
velocity model is applied to simulate the channel flow as shown in the following
figure, where GH is the moving boundary, and ABCDEF is the stationary wall. At
the inlet and outlet, the periodic boundary condition is imposed. On the solid wall, the
no-slip boundary condition is used. Plot the D2Q9 lattice velocity model, and show
the implementation of boundary conditions for density distribution functions at the
inlet, outlet, wall boundary BC and CD. Outline the computational sequence for the
simulation of this problem.


Outlet
Inlet Flow direction
Moving Wall
Wall
A B
C
D
E F
G H


(10 marks)


(c) The above channel flow can also be simulated by Taylor series expansion and least
square-based lattice Boltzmann method (TLLBM). TLLBM is based on multi-
dimensional Taylor series expansion. For the two-dimensional case, discuss how
many unknowns are involved in the Taylor series expansion if the designed TLLBM
has the second order of accuracy, and how many neighbouring points are often used
when D2Q9 lattice velocity model and least square optimization are applied.
Furthermore, discuss how to apply TLLBM to improve the computational efficiency
for the above channel flow.
(5 marks)




- END OF PAPER -

You might also like