Author |
Message |
M. Cube (m_cube)
Member Username: m_cube
Post Number: 7 Registered: 09-2007
| Posted on Wednesday, November 21, 2007 - 06:12 am: | |
Hello Mr Nelson, First, thank you for your constant support! I appreciate it. I have the following issue. I want to simulate a laminar fluid flow (Re<1) in two connected channels. I used the simplified Navier-Stokes model from Backstrom "Fluid Dynamics" in 3D mode. The medium shall be flowing in y direction, whereas the boundary conditions for fluid flow shall be valid for the channels only. I tried to configure it with NATURAL and VALUE settings, but the medium still keeps "migrating" in the outer region. Can you help isolating it in the channels only? The other issue concerns calculation time. I tried to run it over night, where at some point my RAM was exhausted. Is even the simplyfied version a heavy duty calculation task? Thanks, M. Kubon
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1002 Registered: 06-2003
| Posted on Wednesday, November 21, 2007 - 03:49 pm: | |
1) you have an inconsistent setup: Vy is specified as nonzero at the y-end faces, but zero on the top and bottom z-faces. This creates a serious shear, but you have only one mesh cell in the channel to resolve the velocity shape. If you want to model this drag on top and sides of the channel, you will need enough mesh cells in the channel to resolve the shape of the velocity field. 2) You have a system in which the solution for Vx and Vz is exactly zero. You have defined no THRESHOLD value for either of these, which means the error tolerance will always be 0.001 of whatever nonzero values are computed. In other words, FlexPDE will try to resolve Vx=0 and Vz=0 to machine precision, an impossible feat. Put in a THRESHOLD defining what precision you want to allow in defining zero (1e-6, for example). 3) Your vector(vy) plot is not plotting a vector. Vy is a scalar, which will be interpreted as the x-component of a vector with zero y-component. Plot Vector(vx,vy). 4) You can decouple solutions at interfaces using the CONTACT boundary condition (q.v.). |
Marek Nelson (mgnelson)
Moderator Username: mgnelson
Post Number: 5 Registered: 07-2007
| Posted on Thursday, November 22, 2007 - 12:38 am: | |
Here is an example that keeps all the flow within the channels and runs about 90 minutes on my machine. I have increased the grid density around the channels in x and y dimensions using MESH_SPACING and in the z dimension by adding a few extra layers. The extra layers force extra grid density in the z dimension without increasing the density too much in the x and y (since your z dimension is much smaller than x and y). I have also changed the boundary conditions to VALUE(VY)=0 everywhere except on the ends of the channels. Although with these BCs, the two side regions could be deleted. I also combined the two channel regions and the center region into one. |
M. Cube (m_cube)
Member Username: m_cube
Post Number: 8 Registered: 09-2007
| Posted on Tuesday, November 27, 2007 - 12:20 pm: | |
Hi Marek and Robert Nelson, Your help was perfect as usual! Everything works fine now. I realized now how to set the boundary conditions correctly. Thanks as well for always corresponding to the German descriptions! M. Kubon |
|