Basic Notes Project 2

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

Numerical PDEs: General notes for Project 2

Consider a function:
f(x, y)
which represents the falling of a liquid through the air under the inuence of
gravity. For the purposes of this paper, let it follow the PDE:
d
dx
(W(x) 1)
df
dx
)
d
2
f
dt
2
= 0
Specications for Variables:
1) Gravity works only in the x direction.
2) W(x) is the Weber number:
W(x) =
(x)V (x)

Where:
is the liquid density.
For this paper, let = 1
(x) is the thickness of the curtain of falling liquid.
For this paper, let
0
= 1
V (x) is the speed at which the liquid is falling.
is the surface tension of the liquid.
The Weber number has the following qualities:
1. It is dimensionless.
2. It is the ratio of the ux per unit width to the surface tension force per
unit width.
We know our curtain liquid has the following qualities:
1. When it rst comes out of the hopper, it is going to be the shape of the slot.
The shape of the liquid as it leaves is f(0, y) and for many applications
this is just set to 0.
2. The curtain is xed at the sides. Therefore: f(x, 0) = f(x, 100) = 0.
3. There is no side to side movement of our liquid as it goes down. Therefore:
df
dy
(0, y) = 0
Note that this problem is based in space, not time. We are dealing with a
curtain where x represents the vertical axis of the curtain, and y represents the
horizontal dimenion of the curtain.
1
However, x can be thought of as a time-like component so we can treat this
like a hyperbolic system.
y =
2
3

2
0
V
2
0
g

W(x) 1(W(x) + 2) + k
Project Goals
Show that the PDE is hyperbolic if W(x) 1 and that it is elliptic if
W(x) < 1.
Graph characteristic grids on 0 < x < 300, 0 < y < 100 for:V
0
=
8.7, 20, 200
Write a computer problem using two dierent FDMs where at least one is
explicit.
Determine the CFL conditions for the schemes and describe the accuracy.
Find the orders of accuracy for the methods by comparing them over
several dierent grids to the exact solution.
Considerations for Methods:
Our text lists several dierent methods. I need to pick an explicit one. One
possibility would be Leapfrog:
U
n+1
j
= U
n1
+ 25AU
n
An implicit method could be the upwind method as shown in (10.25) on page
211:
U
j
n+1
= U
n
j

ak
2h
(U
n
j+1
U
n
j1
) +
ak
2h
(U
n
j+1
2U
n
j
+ U
n
j1
)
2
Implementing Leap Frog:
(1) In the text, leapfrog is u
t
+au
x
= 0. For this problem we have a = (1W(x)).
(2) Recall one of the basic problems we had from Project 1: That we couldnt
just used the center space method at the boundary. In this case I need to come
up with a similar method to use at the boundary which is second order in both
space and time. In this case we would use the coecients:
j j+1 j+2
-3/2 2 -1/2
The result of this would be replacing the terms as follows
u
x
=
3U
N
j
+4U
N
j+1
U
N
j+2
2
This produces a forward leapfrog method:
U
N
= U
N1

W(x)k
h
(
3U
N
j
+ 4U
N
j+1
U
N
j+2
2
)
Qualities of the Leapfrog Method:
From page 220 in the book:
3 level explicit method
Second order is space and time.
Stable provided
ak
h
< 1.
For our purposes this is
W(x)k
h
< 1 is a reasonable statement. We can just
pick the k and h step sizes so that it is within the stability range for and
value of W(x).
The value of k
p
will always be on the boundary of the stability region
for the method. There is no growth or decay of the eigenmode.
What is an eigenmode?
Graphically this means that the wave phenomenon that is modelled by
this FDM will not decay as time goes on. This has the advantage that we
make the same assumption for the system we are modelling.
One the other hand, this may also lead to high level oscillations that
actually make the FDM a poor stand-in for the system.
What to watch out for:
(1) Variable Coecients
(2) Nonlinearities in the system.
3

You might also like