Advanced Mathematics For Engineering and Science

In Example 7.9 4, we have seen that due to round-off errors, the shooting method does not work when the P clet number is in the range of 1 and 500. This is an example of instability. The round-off errors (or truncation errors) amplify as the integration proceeds and the magnitude of the errors exceeds the solution. We illustrate this situation further by considering the following example.
Write the finite difference equation for the system
| (7.0 1a) | |
| (7.0 1b) | |
Examine the stability of the system.
The exact solution is
| (7.0 2) | |
We note that y is a decreasing function of x.
Using Equation (7.9 22b), the finite difference equation is
| (7.0 3) | |
Instead of solving Equation (7.10 3) numerically, we seek an analytic solution. The solution is of the form
| (7.0 4) | |
Substituting Equation (7.10 4) into Equation (7.10 3), we obtain
| (7.0 5) | |
From Equation (7.10 5), we deduce that ? 1,2 are the roots of
| (7.0 6) | |
The values of ? 1, 2 are
| (7.0 7) | |
Since h is assumed to be small, we expand
in powers of h.
The roots are then approximately given by
| (7.0 8a) | |
| (7.0 8b) | |
The solution is given approximately by
| (7.0 9) | |
Replacing j by x j (x j=jh), and noting that
| (7.0 10a) | |
| (7.0 10b) | |
The solution y j can be written as
| (7.0 11) | |
Applying boundary condition (7.10 1b), we find C 2 to be zero and y j is a decreasing function of x j.
In an actual computation, there is round-off error and the value of C