next up previous
Up: Return

EGN 5456 Homework 4 Fall 1998
Numerical Solution of Heat Conduction in a Bar

1.
Fix up your program of homework 1 first so that it meets the good programming requirements.

2.
Solve the same problem as in homework 1, but this time, use the Crank-Nicholson method.

3.
Verify the unconditional stability.

4.
Verify that the numerical error is second order in both space and time.

5.
To what power of the round-off error is the numerical effort proportional?

6.
How does this compare with the explicit scheme at $\kappa\Delta
 t /(\Delta x)^2 = {1\over 2}$?

7.
How does this compare with the explicit scheme at $\kappa\Delta
 t /(\Delta x)^2 = {1\over 6}$?

8.
Convert your program to the modified Crank-Nicholson method and address all above questions again.

9.
Discuss the CFL condition for the following beloved scheme

\begin{displaymath}
{s^{n+1}_j - s^n_j\over \Delta t}
 + {u\over 2} {s^{n+1}_j -...
 ...r\Delta x}
 + {u\over 2} {s^n_{j+1} - s^n_j\over\Delta x} = 0
 \end{displaymath}

10.
Why is the CFL condition prediction of the previous question wrong?

11.
Discuss the CFL condition for the Du Fort-Frankel scheme.

12.
Is the CFL condition a condition for stability for the Du Fort Frankel scheme?



Author: Leon van Dommelen