Author |
Message |
Mehdi Naderi (mnader4)
Member Username: mnader4
Post Number: 61 Registered: 12-2006
| Posted on Monday, April 20, 2009 - 12:36 pm: | |
Dear All, I am wondering if somebody helps me. the attached is my 3D code. I want to have fine mesh in th regions 1, 2 and 3. The rest of them, mesh should be very coarse (but not violate the solution procedure and without error). I tried mesh density and mesh spacing, but I would like to see more coarse mesh . Also, how can I increase the mesh in the thickness of regions 1, 2 and 3? I apprecaite your helps. Regards, Mehdi |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1247 Registered: 06-2003
| Posted on Tuesday, April 21, 2009 - 06:04 pm: | |
1. The effort in FlexPDE has been directed to providing a reliable solution, not to minimizing the mesh size. Automatic mesh refinement in areas of sharp curvature allows a balance between these two goals, but its effect is to increase mesh density in areas of need, not to eliminate initially created cells in other areas. The NGRID selector allows you to guide the mesh generator in creating a crude or dense initial mesh, but the NGRID request is always over-ridden by other mesh density requirements, like the size of domain features. Furthermore, there are continuity requirements in mesh cell sizes that do not allow cell sizes to grow too abruptly between coarse and fine regions. This means that there is an effective minimum density controlled by the sizes of features in your domain. You have disabled the automatic mesh refinement in your script, possibly because it led to massive mesh density in some locations. Such dense mesh refinement is indicative of improper initial conditions, and should not be ignored or over-ridden. You should instead cure the cause of the difficulty in your problem statement. 2. FlexPDE tries to create cells as nearly equilateral as possible in the given domain. A thin layer will therefore be one cell layer thick unless the overall cell size is less than the layer thickness. There are two ways to defeat this policy: a) manually introduce additional layers within the layer in question. b) scale the layer thickness to make FlexPDE believe it is large. (See Tech Note: Coordinate Scaling). |
|