Here are the Postscript file:homework2.ps and
the LaTeX file:homework2.tex.
Math 228C Computational Assignment 2 Spring 1999
Due: Monday, May 3, 1999
- 1.
- Solve the Neumann problem for 1D Poisson equation with GSRB (Red Black Gauss-Seidel)
|
(1) |
 |
= |
0 |
|
(2) |
 |
= |
0 |
on
= [0,1]. For N=64,128 and 256,plot
- The error
versus iteration for 1000 iterations.
- The error
versus x after 10,50,100,500 and 1000 iterations.
Experiment with different initial guesses
v0, and determine whether all initial guesses will result in a sequence of iterates that converge to a correct solution. If not, try to find a method for picking
v0 such that the iterates are certain to converge to a correct solution.
- 2.
- Use a Multigrid V-Cycle with Weighted Jacobi (
=2/3) to solve
|
(3) |
 |
= |
0 |
|
(4) |
u |
= |
 |
in 1D with N=48 and
initital guess:
|
(5) |
vj |
= |
![$\displaystyle \frac 12 [\rm {\sin}(\frac {12 j \pi}N) +\rm {\sin} (\frac {30 j \pi} N)]$](img8.gif) |
- Plot the initial guess
v0.
- Plot the error
versus x in [0,1] after 1 fine grid relaxation.
- Plot the error
versus x in [0,1] after 3 fine grid relaxation.
- Transfer the fine grid residual after 3 fine grid relaxations to the coarse grid using full weighting and relax once on the coarse grid. Plot the error
on the fine grid relaxation.
- Do the same after 3 relaxation on the coarse grid
Compare these plots to Figures (18a-e) in the Multigrid Tutorial book.
- 3.
- Plot the error
versus x in [0,1] after 1,2,5 and 10 Full Multigrid V-Cycles applied to Problem 2.
MAT 228C Webpage
|Program #1
|Program #2
|Program #3
|Program #4
Wenlong Jin
1999-05-01