Oundary Alue Roblems: Dr. Johnson

Review
Boundary Value Problems

Newtons Shooting Method
Summary
B OUNDARY VALUE P ROBLEMS
Dr. Johnson
School of Mathematics
Semester 1 2008
Dr. Johnson MATH65241

Review
Summary
O UTLINE
1 R EVIEW
2 B OUNDARY VALUE P ROBLEMS
Condition at Both Ends
Taking a guess
3 N EWTONS S HOOTING M ETHOD

Correcting the guess
The Secant Method
Newtons Method
4 S UMMARY

Review
Summary
O UTLINE
1 R EVIEW
Taking a guess

The Secant Method
Newtons Method
4 S UMMARY

Review
Summary
O UTLINE
1 R EVIEW
Taking a guess

The Secant Method
Newtons Method
4 S UMMARY

Review
Summary
O UTLINE
1 R EVIEW
Taking a guess

The Secant Method
Newtons Method
4 S UMMARY

Review
Summary
I NTIAL VALUE PROBLEMS FOR ODE S
So far we have have encountered initial value problems

We can start at one end and march to the other
We have seen different methods of differing order
We will select a method based upon:
Accuracy – the order of errors in the method
Expense – how time consuming a method is
Stability – yet to talk about this!!

Review
Summary
I NTIAL VALUE PROBLEMS FOR ODE S
So far we have have encountered initial value problems

We can start at one end and march to the other
We have seen different methods of differing order
We will select a method based upon:
Accuracy – the order of errors in the method
Expense – how time consuming a method is
Stability – yet to talk about this!!

Review
Boundary Value Problems Condition at Both Ends
Newtons Shooting Method Taking a guess
Summary
O UTLINE
1 R EVIEW
Taking a guess

The Secant Method
Newtons Method
4 S UMMARY

Review
Summary
B OUNDARY CONTITIONS
All ODEs and PDEs require boundary conditions in order
that a solution may exist
In initial value problems, the boundary conditions are all on
one side
This is not always the case, for instance take the following
problem:
d 2y dy
2
+κ + xy = 0
dx dx
with the boundary conditions
y (0) = 0 and y (1) = 1.
then we have conditions at both ends!

Review
Summary
one side
problem:
d 2y dy
2
+κ + xy = 0
dx dx
y (0) = 0 and y (1) = 1.

Review
Summary
one side
problem:
d 2y dy
2
+κ + xy = 0
dx dx
y (0) = 0 and y (1) = 1.

Review
Summary
R EDUCE THE PROBLEM TO AN ODE

We can rewrite the problem above as a system of first
order ODEs:
Y1 = y (x);
dy
Y2 = ;
dx
dY1
= Y2 ;
dx
dY2
= −κY2 − xY1 .
dx
and the boundary conditions are:
Y1 (x = 0) = 0 and Y1 (x = 1) = 0.
Review
Summary
W HAT NEXT ?
In order to solve the problem by marching through x we
need to assign a value to Y2 (x = 0)
How to choose a value of Y2 – our choice must satisfy the
boundary condition
2
1.8
1.6
1.4
1.2
y1(x)
1
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
x
Review
Summary
W HAT NEXT ?
In order to solve the problem by marching through x we
need to assign a value to Y2 (x = 0)
How to choose a value of Y2 – our choice must satisfy the
boundary condition
2
1.8
1.6
1.4
1.2
y1(x)
1
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
x
Review
Summary
O UTLINE
1 R EVIEW
Taking a guess

The Secant Method
Newtons Method
4 S UMMARY

Review
Summary
TAKING A GUESS
Start by making a guess, g, to Y2 (0) so that

µ ¶ µ ¶
Y1 (0) 0.
Y (0) = =
Y2 (0) g
Next, solve using your favourite method, to get

µ ¶ µ ¶
Y1 (1) β1
Y (1) = =
Y2 (1) β2
then by comparing the value β1 to our boundary condition,

we can see how good the guess was

Review
Summary
U P A BIT... DOWN A BIT...
In order to make our guess better, we want to know

whether we have shot above or below
Let us look at the value
φ(g) = Y1 (1; g) − 1,
then the guess is right when φ = 0.

This now looks like a root finding problem
which leads us to Newtons Shooting Method

Review
Summary
U P A BIT... DOWN A BIT...
In order to make our guess better, we want to know

whether we have shot above or below
Let us look at the value
φ(g) = Y1 (1; g) − 1,
then the guess is right when φ = 0.

This now looks like a root finding problem
which leads us to Newtons Shooting Method

Review
The Secant Method
Newtons Method
Summary
O UTLINE
1 R EVIEW
Taking a guess

The Secant Method
Newtons Method
4 S UMMARY

Review
The Secant Method
Newtons Method
Summary
C ORRECTING THE GUESS

Suppose we have taken a guess ĝ, then seek a correction
dg such that
φ(ĝ + dg) = 0
Then take a Taylor expansion around φ(ĝ) to obtain
dφ
φ(ĝ + dg) = φ(ĝ) + dg (ĝ) + O(dg 2 )
dg
So equating this to zero and rearranging we find our

correction is
φ(ĝ)
dg = −
φ′ (ĝ)

Review
The Secant Method
Newtons Method
Summary
R EMINDER ...
Newton’s root finding algorithm has quadratic convergence

The formula is:
xn+1 = xn − f (x)/f ′ (x)
0.8
Start with a guess x0 = 0.8
0.6
0.4 In order to generate the
0.2 green line, we need the
f(x)
0 gradient of f at x0 = 0.8
-0.2
The new guess is
-0.4
x1 = 0.55625
-0.6
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x

Review
The Secant Method
Newtons Method
Summary
R EMINDER ...
Newton’s root finding algorithm has quadratic convergence

The formula is:
xn+1 = xn − f (x)/f ′ (x)
0.8
Start with a guess x0 = 0.8
0.6
0.4 In order to generate the
0.2 green line, we need the
f(x)
0 gradient of f at x0 = 0.8
-0.2
The new guess is
-0.4
x1 = 0.55625
-0.6
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x

Review
The Secant Method
Newtons Method
Summary
O UTLINE
1 R EVIEW
Taking a guess

The Secant Method
Newtons Method
4 S UMMARY

Review
The Secant Method
Newtons Method
Summary
S ECANT M ETHOD
Our formula to generate new guesses is:
φ(gn )
gn+1 = gn −
φ′ (gn )
How do we find φ′ (gn )?

We could estimate it by
φ(gn ) − φ(gn−1 )
φ′ (gn ) =
gn − gn−1
to give the Secant Method
φ(gn )(gn − gn−1 )

gn+1 = gn −
φ(gn ) − φ(gn−1 )
Review
The Secant Method
Newtons Method
Summary
S ECANT M ETHOD
Our formula to generate new guesses is:
φ(gn )
gn+1 = gn −
φ′ (gn )
How do we find φ′ (gn )?

We could estimate it by
φ(gn ) − φ(gn−1 )
φ′ (gn ) =
gn − gn−1
to give the Secant Method
φ(gn )(gn − gn−1 )

gn+1 = gn −
φ(gn ) − φ(gn−1 )
Review
The Secant Method
Newtons Method
Summary
O UTLINE
1 R EVIEW
Taking a guess

The Secant Method
Newtons Method
4 S UMMARY

Review
The Secant Method
Newtons Method
Summary
I NITIAL VALUE P ROBLEM FOR g

Consider again
dY1
=Y2
dx
dY2
= − κY2 − xY1
dx
with Y (0) = (0, g)T

Then we can write
∂Y1
φ′ (gn ) = (1; gn )
∂g
This suggests we should differentiate the original

equations with respect to g
Review
The Secant Method
Newtons Method
Summary

Consider again
dY1
=Y2
dx
dY2
= − κY2 − xY1
dx
with Y (0) = (0, g)T

Then we can write
∂Y1
φ′ (gn ) = (1; gn )
∂g
This suggests we should differentiate the original

equations with respect to g
Review
The Secant Method
Newtons Method
Summary

Then the initial value problem for φ′ is given by
½ ¾ Ã ∂Y2
!
d ∂Y ∂g
= ,
dx ∂g −κ ∂Y ∂Y1
∂g − x ∂g
2
¯ µ ¶
∂Y ¯¯ 0
= .
∂g x=0
¯ 1
Then we may recover φ′ since
∂Y1
φ′ (gn ) = (1; gn )
∂g
and use newton’s formula to generate new guesses

Review
The Secant Method
Newtons Method
Summary

Then the initial value problem for φ′ is given by
½ ¾ Ã ∂Y2
!
d ∂Y ∂g
= ,
dx ∂g −κ ∂Y ∂Y1
∂g − x ∂g
2
¯ µ ¶
∂Y ¯¯ 0
= .
∂g x=0
¯ 1
Then we may recover φ′ since
∂Y1
φ′ (gn ) = (1; gn )
∂g
and use newton’s formula to generate new guesses

Review
The Secant Method
Newtons Method
Summary
T HE N EWTON -AUGMENTED S YSTEM
Simultaneously solve for Y and φ′ , setting
Y1 = y , Y2 = y ′ , Y3 = ∂y /∂g and Y4 = ∂y ′ /∂g
Resulting system looks something like
   
Y2 0
dY  −κY2 − xY1   g 
= , Y (x = 0) =  0 .
 
dx  Y4 
−κY4 − xY3 1

Review
Summary
S UMMARY
Boundary value problems are commonplace in CFD and

mathematics in general
Newton’s Shooting method is one of the most common
methods for solving these types of problem
Newton’s method is quadratic, so should converge in 5-15
iterations
Start worrying if it doesn’t!

Review
Summary
S UMMARY
Boundary value problems are commonplace in CFD and

mathematics in general
Newton’s Shooting method is one of the most common
methods for solving these types of problem
Newton’s method is quadratic, so should converge in 5-15
iterations
Start worrying if it doesn’t!

Oundary Alue Roblems: Dr. Johnson

Uploaded by

Copyright:

Available Formats

Oundary Alue Roblems: Dr. Johnson

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Oundary Alue Roblems: Dr. Johnson

Uploaded by

Copyright:

Available Formats

Review

Boundary Value Problems

B OUNDARY VALUE P ROBLEMS

Dr. Johnson MATH65241

3 N EWTONS S HOOTING M ETHOD

Dr. Johnson MATH65241

3 N EWTONS S HOOTING M ETHOD

Dr. Johnson MATH65241

3 N EWTONS S HOOTING M ETHOD

Dr. Johnson MATH65241

3 N EWTONS S HOOTING M ETHOD

Dr. Johnson MATH65241

I NTIAL VALUE PROBLEMS FOR ODE S

So far we have have encountered initial value problems

Dr. Johnson MATH65241

I NTIAL VALUE PROBLEMS FOR ODE S

So far we have have encountered initial value problems

Dr. Johnson MATH65241

3 N EWTONS S HOOTING M ETHOD

Dr. Johnson MATH65241

y (0) = 0 and y (1) = 1.

then we have conditions at both ends!

y (0) = 0 and y (1) = 1.

then we have conditions at both ends!

y (0) = 0 and y (1) = 1.

then we have conditions at both ends!

R EDUCE THE PROBLEM TO AN ODE

3 N EWTONS S HOOTING M ETHOD

Dr. Johnson MATH65241

Start by making a guess, g, to Y2 (0) so that

Next, solve using your favourite method, to get

then by comparing the value β1 to our boundary condition,

Dr. Johnson MATH65241

U P A BIT... DOWN A BIT...

In order to make our guess better, we want to know

then the guess is right when φ = 0.

Dr. Johnson MATH65241

U P A BIT... DOWN A BIT...

In order to make our guess better, we want to know

then the guess is right when φ = 0.

Dr. Johnson MATH65241

3 N EWTONS S HOOTING M ETHOD

Dr. Johnson MATH65241

C ORRECTING THE GUESS

Then take a Taylor expansion around φ(ĝ) to obtain

So equating this to zero and rearranging we find our

Dr. Johnson MATH65241

Newton’s root finding algorithm has quadratic convergence

xn+1 = xn − f (x)/f ′ (x)

Dr. Johnson MATH65241

Newton’s root finding algorithm has quadratic convergence

xn+1 = xn − f (x)/f ′ (x)

Dr. Johnson MATH65241

3 N EWTONS S HOOTING M ETHOD

Dr. Johnson MATH65241

How do we find φ′ (gn )?

to give the Secant Method

φ(gn )(gn − gn−1 )

How do we find φ′ (gn )?