Author |
Message |
christian weber (christian_weber)
New member Username: christian_weber
Post Number: 1 Registered: 03-2006
| Posted on Wednesday, May 28, 2008 - 09:47 am: | |
Hello, I have a problem with solving a convection-diffusion problem in 3-D: div(D*grad(c))-(vx*dx(c)+vy*dy(c)+ vz*dz(c))=0 I used the Navier-Stokes-equations for modelling the velocity field. It's not possible to obtain a realistic result for the velocity field by setting the viscosity to 0.001 even if I set the density and thus the nonlinear term of the NS-equations to zero. Using a viscosity of 1000 leads to a plausible velocity field. But it seems, that this velocity field has no effect on the concentration profile. The concentration is only affected by the diffusion term. I can't see what is wrong with my script. Are there any mistakes in the equation system or have I set wrong boundary conditions? I have been trying to solve this problem since three weeks now. If anybody can help me I would greatly appreciate it. Best regards Christian PS: I use the FlexPDE Professional Version 5.0.18
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1123 Registered: 06-2003
| Posted on Wednesday, May 28, 2008 - 06:32 pm: | |
1. It looks to me like the biggest problem you have is the boundary condition at "BOTTOM" in the center pipe. You have an overall BC of value(vz)=0 on "bottom". This is over-ridden in the left and right pipe by a redefinition Natural(vz)=0. But the center pipe does not do this. So it is impossible to flow fluid through the center pipe (the C source). You can only diffuse into the disk. If you replicate the "bottom" BC over-rides used in the outer two pipes into the center one, you will be able to flow through this inlet pipe. Also, you can use a VALUE(C) boundary condition at the center pipe inlet instead of a NATURAL. (The script you sent has this BC garbled anyway). 2. You have specified constant Vz across the inlet face of the two inlet pipes, followed immediately by a Vz=0 condition on the sides. This introduces an unresolvable discontinuity in the Vz solution that will probably cause a lot of useless regridding and slow running. You should program a parabolic velocity distribution on the inlet face, so the BC matches the long-pipe distribution. We will continue to look at this problem, and may post more comments later. |
christian weber (christian_weber)
New member Username: christian_weber
Post Number: 2 Registered: 03-2006
| Posted on Thursday, May 29, 2008 - 09:49 am: | |
Thank you very much for your prompt answer. Oh, I deleted the "bottom" BC in region 3 accidentally, but this was not the only problem. I programmed a parabolic velocity profile on the inlet face and it works brilliant now. Thanks a lot for this suggestion. I have attached the corrected script. Best regards Christian
convection and diffusion - corrected.pdf test1.pde (3.7 k) |
|
|