| Author | Message | |||
     Armin Wiegner (praktaw) New member Username: praktaw Post Number: 1 Registered: 09-2007 |
Hello Mr. Nelson, I try to simulate some Electrical Fields. I wanted to validate my differential equation by simulating a problem with an very easy geometry. It works, but only in some frequency domains. In other frequency-domains the simulation time increases hundreds of times, I get chaotic contours and to high values. In the example I attached, the critical domains are located between 3 and 105Hz. It is the value 'f' under the heading "values". I use Flex PDE 4.2.15 Professional version and I don't know what the trouble is.
| |||
| Armin Wiegner (praktaw) New member Username: praktaw Post Number: 2 Registered: 09-2007 |
I collected some data, by trying to understand the problem. This is an Excel-sheet with my formulas to analyze the Problem mathematically and a graph which shows the results as well of the FlexPDE-simluation as of my analyses. I have another very similar Simulation and a corresponding Exelsheet, which shows similar but not the same behaviour. Thank you very much!
| |||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 965 Registered: 06-2003 |
I suspect that you are suffering from round-off troubles. Your equation is effectively Div(K*grad(Vr)-eps*omega*grad(Vi))=0. In some of your materials, (eps*omega)/K is like 1e-9, so the dependency on Vi is lost in the numerical noise. You might be able to extract a solution by forcing a Newton's method solution. Use SELECT NONLINEAR. This causes frequent re-testing of a trial solution by insertion in the PDE, and limitation of each change to the step giving the best improvement of the solution. It will not help the running time issue, but it might force a real answer. Because of the tiny nature of the cross-terms, you might want to look for some kind of perturbation treatement, rather than a direct solution. |
PDE-File Validation
Excel-Sheet Validation