Scribd
Upload
191 views

Shooting Method Tutorial

This document introduces the shooting method for numerically solving boundary value problems. It uses the example of solving the boundary value problem f'' + f = 0, f(0) = 0, f(1) = 1. The shooting method reformulates the problem as an initial value problem by replacing one boundary condition with a parameter s and iteratively solves for s. It marches the solution using Euler's method and updates s using the Newton-Raphson method until the boundary condition is satisfied. The shooting method converges to s = 1.188 for this example problem.

Uploaded by

ashwin802
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
191 views

Shooting Method Tutorial

This document introduces the shooting method for numerically solving boundary value problems. It uses the example of solving the boundary value problem f'' + f = 0, f(0) = 0, f(1) = 1. The shooting method reformulates the problem as an initial value problem by replacing one boundary condition with a parameter s and iteratively solves for s. It marches the solution using Euler's method and updates s using the Newton-Raphson method until the boundary condition is satisfied. The shooting method converges to s = 1.188 for this example problem.

Uploaded by

ashwin802
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 3

Shooting Method Tutorial

Gary Parker
July 9, 2006

Let f = f(x) denote a function to be determined by the boundary value problem


f ′′ + f = 0 , f (0) = 0 , f (1) = 1 (1a,b,c)
This problem has an exact solution:
sin( x )
f (x) = (2)
sin( 1)
The shooting method can be used to find this solution numerically. Of course, a
numerical method is not necessary to solve (1). But the shooting method also
works for nonlinear boundary value problems for which there is no closed-form
solution.

Before introducing the shooting method it is useful to review the Newton-


Raphson method for solving algebraic equations iteratively. Let R(s) be some
specified function of s (e.g. R(s) = 1-e-s). We wish to solve the equation
R(s) = 0 (3)
new
Let s be a guess for the root. We can get an improved guess s by defining
∆s = snew − s (4)
and solving the linearized form based on Taylor expansion
R(s + ∆s) ≅ R(s) + R′(s)∆s = 0 (5)
to obtain the relation
R(s)
snew = s − (6)
R′(s)
Equation (6) can now be solved iteratively for s.

To implement the solution of (1a,b,c) by the shooting method we first define


g ≡ f′ (7)
so that (1a) reduces to two first-order equations,
f′ =g
(8a,b)
g′ = −f
We could solve this equation by marching from x = 0 using e.g. the Euler step
method if both boundary conditions (1b) and (1c) were specified at x = 0.
Boundary condition (1c), however, is specified at x = 1.

To overcome this problem, we replace (1c) with the condition


g(0) = s (9)
where the right value of s is the one that yields the value f = 1 at x = 1. We don’t
know what s should be yet, so we treat it as a free variable. Thus f and g are not
only functions of x, but also of s;
f = f ( x, s) , g = g( x, s) (10a,b)
The boundary value problem (1) now becomes
df
= g( x, s)
dx
dg
= −f ( x, s)
dx
f (0, s) = 0
g(0, s) = s
(11a,b,c,d)
where s is to be determined. Now (11a,b) can be solved numerically using e.g.
the Euler step method. Let
1
∆x = (12)
N
where N is an appropriately large positive integer, and let xn = (n-1)∆ x, n = 1..N
+ 1, so that x1 = 0 and xN+1 = 1. Further letting fn and gn denote the values of f
and g at xn, (11a,b) discretize to
f n +1 = f n +gn ∆x
(13a,b)
g n +1 = g n −fn ∆x
where since x1 = 0,
f1 = 0
g1 = s
(13c,d)
Equations (13a,b) are then solved by marching from n = 1 to n = N + 1.

Now the right value of s is the one that satisfies (1c), or thus
R(s) = f (1, s) −1 = 0 (14)
new
For any guess s, a better guess s is given as
R(s) f (1, s) −1
s new
=s− =s−
dR ∂f (15)
ds ∂s x =1
But in order to implement this we need to know ∂f/∂s at x = 1. So to do this we
specify the variational parameters F(x,s) and G(x,s), where
∂f
F≡
∂s
(16a,b)
∂f
G=
∂s
(Now in principle we should write ∂f/∂x in equation (11a) etc., because f is also a
function of s, but for simplicity we will retain the notation df/dx and dg/dx.) We
now take the derivatives of (11a,b,c,d) with respect to x to specify the variational
problem:
dF
= G( x, s)
dx
dG
= −F( x, s) (17a,b,c,d)
dx
F(0, s) = 0
G(0, s) =1
Now we can also solve the variational problem using the Euler step method;
Fn +1 = Fn + Gn ∆x
(18a,b)
G n +1 = G n −Fn ∆x
F1 = 0
G1 = 1
(18c,d)
Thus (14) becomes
f N +1 − 1
snew = s = s − (19)
FN +1

The problem now becomes fairly straightforward. Guess a value of s. Then


solve (13a,b) and (18a,b) out to xN = 1 to find f, g, F and G. Then get a better
guess of s from (19). Repeat until convergence is obtained in s.

Try it out. The code is easy to implement in, for example, Matlab or Visual Basic
for Applications in Excel. Since (1) has the analytical solution (2), it is easy to
verify that the correct value of s is
df d  sin( x )  1
s= = = = 1.188 (20)
dx x =0 dx  sin( 1)  sin( 1)
Try initial guesses of s that are different from 1.188 and see how many steps are
needed for convergence. The shooting method often fails if the initial guess of s
is too far from the correct value, so often one must play around to find first
guesses that lead to convergence.

You might also like