| Author | Message | |||
     yucheng zhang (stanley_z) Junior Member Username: stanley_z Post Number: 3 Registered: 05-2006 |
Hi, I tried to run a simple 1-D diffusion problem. The script is attached. The plot of the result I have got becomes rough at the distance ranging from 50-80. The RMS Err below is shown as 1.7e+9. I am wondering is there any way I can reduce the RMS err and get a more smooth curve? many thanks
| |||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 596 Registered: 06-2003 |
The "roughness" you are seeing on you log plot is noise in the seventh decimal place relative to the sizes of C in your problem. FlexPDE computes a numerical approximation to the solution, not an analytic solution of infinite precision. When the solution falls from some C0 to a value of zero, there will be a small amount of ringing around zero. In your case, this is in the sixth decimal place. You can decrease the size of the ringing by decreasing ERRLIM. This will make the problem run longer and require a denser mesh. The plot will look very much the same, but the absolute size of the "roughness" will be smaller. Or you can plot on linear scale, where the "roughness" will not appear. Incidentally, when I ran this on version 5.0.10 it reported an RMS error of 1e-6, not 1e9. What version are you using? |
script for 1-D time-dependent diffusion problem