CFD Fundamentals Phase Test 2011 (WITH ANSWERS)

Attempt all the questions. You have 90 minutes plus 25% extra for those with dispensation.
IMPORTANT: DO NOT DETACH THE ANSWER SHEET BELOW FROM THE QUESTION SHEETS
Candidate name and programme code PRINT BELOW

...............................................................................................................................................................
Tanvir Method Example: The candidate eliminated B and E as possible right answers and allocated no
marks to them. But was then unsure if it was A, C or D. C seemed right but there was an argument for A
and D and so the 4 marks were split as shown below. C was the right answer so the candidate got 2 out
of 4 for the question. If you are sure of the right answer place 4 next to your letter choice.
question
34

A
1

C
2

D
1

mark
2

mark

question
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

TOTAL

Useful formulae
For the second order partial differential equation

A( x, y )

2
2
2

B
(
x
,
y
)

C
(
x
,
y
)
F , , x, y 0
2
2
x
xy
y
x y

B 2 4 AC

B
2A

B
2A

The Taylor expansion of a function in one variable is

1 2u
u
u x0 x, y0 u x0 , y0 x 2 x 2 ...
2 x 0
x 0
1 3u
3
4
3 x O(x )
6 x 0

Useful trig substitution

exp

1x exp 1x 2 cos x

1: Consider the partial differential equation below.

2
2
2

3
Q 2 2
6
13 5 0
2
x
xy
y
x
y
What must be the value of the missing number Q if the equation is to be parabolic?
(a)
(b)
(c)
(d)
(e)

2.25
-2.25
3
-3
1.732

Parabolic requires B 2 4 AC 0 . We have B=3 and therefore require

B 2 4 AC 9 4Q 0 , leading to Q=2.25 (a)

2: Consider the equation below.

2
2
2
5 2 2
16 2 0
x
xy
y
The two characteristics of the equation are:
(a) 5 y 8 x

y 2x

(b) 5 y 8 x

y 2x

(c) 8 y 5 x

y 2x

(d) 5 y 8 x

y 2x

(e) 8 y 5 x

y 2x

We have as the equations of the characteristics 1

B 2 4 AC 4 4.(5)(16) 324 . So 1

Integrate the two equations:

B
2A

dy 2 324

2
dx
25

dy
2 y 2x
dx

B
dy
.
, where
2A
dx

dy 2 324
8

dx
25
5

dy
8

5 y 8 x i.e. answer (d)

dx
5

3: Which one of the following equations is elliptic?

2
2
2
(a)
2
1 2 0
x 2
xy
y
(b)

2
2 2

0
x 2
xy y 2

(c)

2
2 2

0
x 2
xy y 2

(d)

2 2
2

2
0
x 2 xy
y 2

(e)

2
2
2

2
0
x 2
xy
y 2

Elliptic requires B 2 4 AC 0 . In turn we have

(a) 4 4.(1).(1) 0 parabolic
(b) 4 4.(1).(1) 8 hyperbolic
(c) 4 4.(1).(1) 0 parabolic
(d) 1 4.(1).(2) 9 hyperbolic
(e) 4 4.(1).(2) 4 elliptic

4: The X error is the error in the solution to the PDE subject to the given initial values and
boundary conditions caused by replacing the continuous problem by a discrete one and is
defined as the difference between the exact solution of the PDE (round-off free) and the exact
solution to the finite difference equations (round-off free).
What is the missing word (or words) X?
(a) significant
(b) running

(c) discretisation
(d) truncation
(e) round-off

5: A finite difference representation is said to be X if we can show that the difference between
the PDE and its finite difference representation vanishes as the mesh is refined.
What is the missing word (or words) X?
(a) stable

(b) consistent
(c) convergent
(d) over-refined
(e) ill-posed
6: Elliptic partial differential equations are X problems unlike hyperbolic and parabolic
equations.
What is the missing word (or words) X?
(a) insoluble
(b) ill-posed
(c) marching

(d) boundary value

(e) initial value

7: Consider the finite difference approximation below for the first partial derivative with
4u x0 x, y0 3u x0 , y0 u x0 2x, y0
u
O ( x 2 )
respect to x of u ( x, y ) :

x
2

x
0
u
What is the equivalent expression in terms of grid variables (i. j ) for
y 0

u 4u 3ui , j ui , j 2
(a) i , j 1
O(x 2 )
2x
y 0
u 4u 3ui , j ui , j 2
(b) i , j 1
O(x 2 )

y
2

x
0
u 4u 3ui , j ui , j 2
(c) i , j 1
O(y 2 )
2y
y 0
u 4u 3ui , j ui 2, j
(d) i 1, j
O(y 2 )
2y
y 0
3ui , j ui 2, j 2
u 4u
(e) i 1, j 1
O(y 2 )
2y
y 0
u
Note question asks for the y-derivative
y 0
4ui 1, j 3ui , j ui 2, j
4u x0 x, y0 3u x0 , y0 u x0 2x, y0
u
2

O
(

x
)

O(x 2 )

2x
2x
x 0

To find the y-derivative, swap the roles of I and j and replace x by y. So we have:
4ui , j 1 3ui , j ui , j 2
u 4u x0 , y0 y 3u x0 , y0 u x0 , y0 2y
O(y 2 )
O(y 2 )

2y
2y
y 0

i.e. answer (c).

8: Consider the function u ( x) 3exp(2 x) . What is the numerical value of the derivative du dx
when x 1.
(a)
(b)
(c)
(d)
(e)

7.389
14.778
22.167
36.931
44.334

If u ( x) 3exp(2 x) , then

du
6 exp(2 x) .
dx

When x=1 we find the value for the derivative is 44.334 so (e)
9: Now use the central finite difference approximation to the first derivative

u ui 1, j ui 1, j

.
x
2x

Evaluate the first derivative of u ( x) 3exp(2 x) when x 1 using this, with a grid spacing of

x 0.1 . The magnitude of the difference from the exact value (previous question) is
(a)
(b)
(c)
(d)
(e)

0.296
0.199
0.118
0.098
0.045

We need to evaluate the terms in the approximation

u ui 1, j ui 1, j

.
x
2x

The first term required is ui 1, j 3exp(2(1 0.1)) , giving a value 27.075.

The second term is ui 1, j 3exp(2(1 0.1)) , giving a value 18.149.
Inserting into
answer (a).

u ui 1, j ui 1, j 27.075 18.149

44.630 . The error is therefore 0.296 i.e.

x
2x
0.2

10: Which one of these schemes is not an implicit one?

(a)
(b)

u nj 1 u nj 1
2t
u

n 1
j

n 1
j

2 t

n
j 1

u nj 11 u nj 11 u nj 1

n 1
j 1

2u nj u nj 11

1
(c) uin 1 uin uin1 uin1
2
u n 1 2u n 1 u n 1
u n 2u n u n
uin 1 uin
i 1
i
i 1
i
i 1

(1 ) i 1

(d)
2
2

(e)

u nj 1 u nj
t

n 1
n
n 1
n
c u j 1 u j 1 u j 1 u j 1

0
2x
2
2

An implicit scheme must not have more than one grid point evaluated at different time-steps.
The time-steps are denoted by the superscripts and the subscripts label the grid points. As an
example the expression uin11 or uin11 in (d) for instance makes the expression implicit. The
answer is (c).
11: Consider the central difference approximation to the first derivative
u ui 1, j ui 1, j

x
2x

What is the form of the largest term that we have neglected on the right hand side of this
equation in using this approximation?

2u
x
2
x 0

(a)
(b)

1 2u
2
2 x
6 x 0
1 2u
x 2
2
3 x 0

(c)

1 3u
x 2
3
6 x 0

(d)
(e)

1 3u
2
3 x
3 x 0

We have
1 2u
1 3u
u
2
u x0 x, y0 u x0 , y0 x 2 x 3 x 3 O(x 4 )
2 x 0
6 x 0
x 0
1 2u
1 3u
u
u x0 x, y0 u x0 , y0 x 2 x 2 3 x 3 O (x 4 )
2 x 0
6 x 0
x 0

so that

1 2u
1 3u
1 2u
1 3u
u
u
u x0 , y0 x 2 x 2 3 x 3 O(x 4 ) u x0 , y0 x 2 x 2 3 x 3 O (x 4 )

2 x 0
6 x 0
2 x 0
6 x 0
ui 1, j ui 1, j
x 0
x 0

2x
2x
1 3u
u
2 x 2. 3 x3 O(x 4 )
6 x 0
x 0

2x

So the answer is
1 3u
3 x 2
6 x 0

12: Consider the scheme to solve the heat equation

u nj 1 u nj
t

u nj1 2u nj u nj1

x 2

When a von Neumann stability analysis is made of this equation using the trial solution

u nj g n e
Given that r

x 2

1 j

, the amplification factor is

(a) g 1 2r 1 cos
(b) g 1 2r 1 cos
(c) g 1 r 1 cos
(d) g 1 2r 1 cos
(e) g 1 r 1 cos
We have
u nj g n e

1 j

u nj 1 g n 1e

1 j

u nj 1 g n e

1( j 1)

u nj 1 g n e

1( j 1)

First we rewrite the equation as

u nj 1 u nj r u nj 1 2u nj u nj 1
Substituting in

g n 1e

1 j

g ne

Dividing by
g ne

1 j

gives

1 j

r g ne

1( j 1)

2 g ne

1 j

g ne

1( j 1)

g 1 r e

2 e

r (2 cos 2)

Rearranging gives answer (a)

13: Which of the following statements is false?

(a) Dispersion in the numerical solution of a PDE is generated by odd derivative terms in the
truncated PDE.
(b) Dissipation in the numerical solution of a PDE can act like viscosity to smooth the
solution
(c) Successive over-relaxation (SOR) can speed the rate of convergence of an iterative
scheme
(d) A Courant Number greater than 1 means that the wave speed exceeds the speed at
which effects propagate on the grid.

(e) The optimum relaxation factor in an SOR scheme doesnt depend on the
grid dimensions

14-17: Consider the grid below used to solve Laplaces Equation

2u 2u

0
x 2 y 2
The grid is square with grid spacing equal to 0.1 and the origin (x = 0, y = 0) corresponding to
grid coordinates (i = 1, j = 1). As an example, the grid point labelled 1 is i = 1 j = 2.
The boundary conditions are
u ( x, 0.3) 0.2 2 x u (0.2, y ) 0.1 y u ( x, 0) 0.1

u
0, y 0
x

y
y = 0.3
A
FP2

FP1

3
x = 0.2

14: From the boundary conditions deduce the value of u 3,4 at the grid point labelled A.
(a)
(b)
(c)
(d)
(e)

-0.2
-0.1
0
0.1
0.2

Applying the boundary conditions leads to the following values on the grid.

y
0.2

0.0

-0.2
A

-0.1

0.0

0.1

0.1

1
0.1

x = 0.2

15: The solution is to be found using the direct Jacobi iterative scheme
uiN, j1

1 N
ui 1, j uiN1, j uiN, j 1 uiN, j 1 .

We set initial trial values of 0.1 for the points 1 to 4 in the diagram. The new values are
evaluated in numerical order i.e. point 1 first, and then point 2 etc.
We label the values of u at these points with the shorthand
u1,2 ;u1,3 ;u2,2 ;u2,3 .
Which of these equations for the next iteration of the value at grid point 4 is correct?
(a) N 1

0.1 N

(b)

0.1 N

(c)
(d)
(e)

1
N
4
1
N 1 N
4
1
N 1 N
4
1
N 1 N
4
1
N 1 N
4

The answer is uiN, j1

i.e. answer (a).

0.1 N
0.2 N
N

1 N
1
ui 1, j uiN1, j uiN, j 1 uiN, j 1 N 1 N 0.1 0.0 N

4
4

16: We set initial trial values of 0.1 for the points 1 to 4 in the diagram. What is the value at grid
point 1 after one iteration?
(a)
(b)
(c)
(d)
(e)

0.1
0.133
0.125
0.15
0.2

The answer is uiN, j1

1 N
1
ui 1, j uiN1, j uiN, j 1 uiN, j 1 N 1 N + N N 0.1

4
4

But all the initial guesses, in particular, N and N are 0.1 so the answer is (a) since
N 1

1
1
N + N N 0.1 0.1 + 0.1 0.1 0.1 0.1.

4
4

17: The solution is now to be found using the implicit Gauss Seidel iterative scheme again
solving the grid points in the same numerical order. Which of these expressions for the next
iterated value at grid point 2 is correct?
1
0.2 N + 2 N 1

4
1
N 1 0.2 N 1 + 2 N 1
4
1
N 1 0.2 N 1 + 2 N
4
1
N 1 0.2 N 1 + N
4
1
N 1 0.2 N + N 1
4

(a) N 1
(b)
(c)
(d)
(e)

The direct scheme would be

1 N
1
1
ui 1, j uiN1, j uiN, j 1 uiN, j 1 N 1 N + N 0.2 N 2 N 0.2 N

4
4
4
In a Gauss-Seidel scheme, we can use newly evaluated iterates so when we get to , we have
already worked out an updated value for but not a new value for or . So we have
uiN, j1

N 1

1
2 N 0.2 N 1
4

18: Consider the heat equation below.

T
2T
D 2
t
x

Two of the boundary conditions are T (0, t ) 0; T ( , t ) 0.

Which of these is a possible separable solution of the equation?
(a) T ( x, t ) 3cos 4 x exp( 16 Dt )
(b) T ( x, t ) 3sin 4 x exp(4 Dt )
(c) T ( x, t ) 3cos 4 x exp(2 Dt )
(d) T ( x, t ) 3sin 4 x exp(2 Dt )
(e) T ( x, t ) 3sin 4 x exp(16 Dt )
First check the boundary conditions. Note the solutions with cos(4x) dont satisfy the boundary
conditions since cos(0) and cos(4) are equal to 1 not 0. So we have
T 3sin(4 x) exp( Dt )
T
12 D cos(4 x) exp( Dt )
x
2T
D 2 48 D sin(4 x) exp( Dt )
x
D

So we have
T
48 D sin(4 x) exp( Dt )
t
48 D
T
sin(4 x) exp( Dt ) 16
D

Or just use trial and error on (b), (d) and (e). The answer is (e).

19: Consider the simple one-dimensional wave equation

u
u
6
t
x

Which of the following is a possible solution?

(a) u ( x, t ) sin(6 x t )
(b) u( x, t ) sin( x 6t )
(c) u ( x, t ) cos( x 6t )
(d) u ( x, t ) sin( x 6t )
(e) u ( x, t ) cos(6 x t )
Again you could use trial and error on all five solutions. Otherwise consider the solution
f ( x, t ) f ( x t ) .
Then
f
f
f
f
x
t
So that
f
f

t
x
Hence the solution in this case must be of the form f ( x, t ) f ( x 6t ) i.e. (d).
Uo
y

20: Consider flow over the shallow wedge above (the slope is exaggerated for clarity). The flow
is supersonic with units chosen so that the incident speed U0 is equal to 1. The total flow is
described by the potential

x f ( x 2 y ),

where, for a wedge of slope angle 0.03 radians, we have

x 2y
f 0 for 0
f

0.03
2

for 0

The velocity is given by , which gives

(a) U 1, 0, 0 for 0

U 1 0.03, 0.03 2, 0

(b) U 1, 0, 0 for 0

0.03

U 1
, 0.03, 0 for 0
2

(c) U 1, 0, 0 for 0

0.03

U 1
, 0.03, 0 for 0
2

(d) U 1, 0, 0 for 0

0.03

U 1
, 0.03, 0 for 0
2

(e) U 1, 0, 0 for 0

U 1 0.03, 0.03, 0 for 0

for 0

For 0 , we have the potential x , so that the velocity U ,

U
,
,
(1, 0, 0)
x y z

For 0 , we have the potential x

0.03
U
,
,
, 0.03, 0)
(1
2
x y z

So the answer is (b).

0.03
x 2 y , so that the velocity U ,
2