| Author | Message | |||
     larssl Member Username: larssl Post Number: 4 Registered: 06-2010 |
Dear FlexPDE creators, I'm wondering if some numerical instabilities in my solutions are due to the limited mesh size in the student version or more likely some error in my script. The problem is of the diffusion-reaction type, and the solution is a gaussian-like peak in 1D. I've plotted the variation of the peak amplitude as a function of domain length L, attached. The oscillations increase greatly in amplitude when the domain length is increased. They can be seen to follow each other for both value and natural BC's, and are also exactly the same when varying other parameters. Are these oscillations most likely due to the limited mesh size? If necessary I'll post the program code later.
| |||
| rgnelson Moderator Username: rgnelson Post Number: 1385 Registered: 06-2003 |
It is possible that such oscillations are due to limited mesh size in the student version. A Gaussian is poorly approximated by polynomials unless the mesh is dense. If by increasing the domain length you decrease the number of cells in the Gaussian, then this could be the problem. You should post the script so that we can validate our assumptions about this. | |||
| larssl Member Username: larssl Post Number: 5 Registered: 06-2010 |
Sorry, I never got back to you on this. I ordered the 1D professional version, and that immidiately allowed me to solve the system over a much larger area without the described oscillations in the solution. In other words, the oscillations _were_ due to the limited mesh size in the student version, and this particular problem is now solved for my part. |
Plot of oscillations