| Author | Message | |||
     Alan J. Horston (alan) New member Username: alan Post Number: 1 Registered: 07-2004 |
I try to solve my pde task with FlexPDE 4.0.7a. This task demands to import the extrusion function from the table data format. And sometimes I am getting a little problem -- when I set the boundary conditions on region 2, the mesh generator fail. But without this conditions, all work fine. In attachment there are 2 scripts: "seError.pde" with boundary conditions and "seOK.pde" with commented boundary conditions. Can be for this task more smooth table function is required?
| |||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 182 Registered: 06-2003 |
1. I'm not sure why you chose to use tables to describe your surfaces, they appear to be analytic spheres (see attached figures). Furthermore, FlexPDE tables are rectangular with linear interpolation. This will generate many ragged edges where your curved shapes intersect the table coordinates. You would be better off to describe the surface shapes analytically. 2. FlexPDE maps a triangular mesh onto the declared boundaries of the surfaces. It does not attempt to search the surface definition for breaks in slope. In your case you seem to have a sphere imbedded in a truncated sphere, but you have given FlexPDE no clue that there are sharp slope breaks at the periphery of the inner sphere and the truncation plane. You should declare a new region or a feature to delineate the slope breaks of surfaces. For instance, declare three regions: the outer boundary, the truncation edge, and the inner sphere; define the surfaces with different analytic descriptions in each of these regions. FlexPDE will then create a mesh that exactly tracks the breaks in the surfaces. 3. Is the outer boundary of your figure really supposed to be a ten-sided polygon, or are you merely approximating a circle? You can specify a circle directly, using the ARC boundary. There are meshing difficulties in the points of the polygon where the spherical surface becomes vertical or vanishes. Be sure your boundary does not extend beyond the sphere. 4. Your problem appears to be very similar to our 3d_Shell.pde example. You might use that problem as a guide for setting up this one.     | |||
| Alan J. Horston (alan) New member Username: alan Post Number: 2 Registered: 07-2004 |
Thank you for the detailed answer. We redesign our task according to you recommends – we introduce the analytic functions. But the some problems still exist. Our task (in attach) has some troubles in version 4 of FlexPDE. But in version 3 this task successfully solved. What we have to do for solving this task in version 4?
| |||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 186 Registered: 06-2003 |
Well, you got point #1, but not point #2. And it doesn't look to me like a correction of the original design, but a totally different beast. But whatever. My previous point #2 was merely this: If you have sharp breaks in your surface slope, you must give the mesh generator a trace of these breaks (by putting boundaries there), so that the mesh can be made to conform to the shape. In this new problem, you have a sphere imbedded in the figure, but no clue to the mesh generator where the break between the sphere and the flat dividing plane occurs. Your square is nowhere near it. Version 3 runs the problem (sort of), but if you use the domain review button, you will see that the surface of the sphere is really crudely modeled. Your output plots show the same message: the sphere is really badly delineated. The solution is merely to put a bounding curve at the diameter of the sphere. See the attached file.
| |||
| Alan J. Horston (alan) Junior Member Username: alan Post Number: 3 Registered: 07-2004 |
Thank you very much! It works. |
se pde problem
ellipse.pde