Author |
Message |
Mohammad Rahmani (mrahmani)
Member Username: mrahmani
Post Number: 11 Registered: 10-2004
| Posted on Sunday, November 28, 2004 - 01:01 am: | |
Can we implement the following First order PDEs in FlexPDE: dy(U1)=f(x,y,U1,U2) dx(U2)=g(x,y,U1,U2), subject to proper BCs
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 267 Registered: 06-2003
| Posted on Monday, November 29, 2004 - 04:18 pm: | |
The finite element discretization process is based on integrals over cells. In systems where no differential operators are applied in one of the dimensions, oscillatory solutions can be accepted, as long as the integrals are correct. You can certainly pose your problem, but I would suggest adding a small diffusion term in the coordinate which has no differential terms, to stabilize the solution in that coordinate. This also provides a definition for the Natural() BC in that direction, constraining boundary oscillations. The default natural()=0 condition enforces zero normal derivative along these boundaries. You might also need to increase the upwind weight, as discussed in a previous posting. |
Mohammad Rahmani (mrahmani)
Member Username: mrahmani
Post Number: 12 Registered: 10-2004
| Posted on Tuesday, November 30, 2004 - 01:03 am: | |
Thanks for your advices. /Mohammad |
|