Author |
Message |
Alan J. Horston (alan)
Member Username: alan
Post Number: 4 Registered: 07-2004
| Posted on Wednesday, August 04, 2004 - 10:38 am: | |
Hello! First of all, thank you very much for you help. The mesh is built successful. But I have the question about calculation process. In attachment there is a testing sample for this. In region 2 of boundaries section I set the natural boundary condition on surfaces 2 and 3. In plots section I display the normal(grad(u)) value on these 2 and 3 surfaces. To my mind, these values on surfaces 2 and 3 in different sections (boundaries section and plots section) must be equal. But in plots section the normal(grad(u)) value on surface 2 is wrong (on surface 3 normal(grad(u)) is near of definition). How can we overcome the difficulties? |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 201 Registered: 06-2003
| Posted on Wednesday, August 04, 2004 - 04:30 pm: | |
The derivative of the solution is double-valued at an interface. Absent an explicit command, FlexPDE evaluates plot quantities on the upper layer (if there are two). For your surface 3 plot, the upper layer is outside the small sphere. For your surface 2 plot, the upper layer is inside the sphere. Therefore, the displayed values are different; one represents the value outside, the other the value inside the small sphere. (Note that surface 3 does not exist on the cutplane, because it has been merged with surface 2). A NATURAL boundary condition specification at an interior boundary does not demand that the flux on both sides of the interface match the given value. What it prescribes is the difference between the flux values on the two sides of the interface. In an electrostatic problem, this means a surface charge (See an electrostatic text: D2.n-D1.n=sigma). In principle, you can cure this display problem by specifying the layer for evaluation of the plot function, by adding an ON LAYER 1 qualifier to the plot statement. In fact, this exposes a bug in the plot distiller related to limited regions, and version 4.0.7a crashes. I will fix this problem in the next version update. |
Alan J. Horston (alan)
Member Username: alan
Post Number: 5 Registered: 07-2004
| Posted on Wednesday, August 11, 2004 - 10:49 am: | |
Thank you. When we shell hope to see the next release? |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 204 Registered: 06-2003
| Posted on Wednesday, August 11, 2004 - 05:17 pm: | |
Version 4.1.0 is being tested for release this week or next. |
Alan J. Horston (alan)
Member Username: alan
Post Number: 6 Registered: 07-2004
| Posted on Sunday, August 22, 2004 - 05:42 am: | |
Thank you for version 4.1.0, the "on layer" qualifier works well. But our problem unfortunately still exists. The "normal(grad(u))" on surface "02" strongly differs from the definition values on surface "02" in "boundaries" section (attached file). We want to fix this mistake. Is it possible?
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 215 Registered: 06-2003
| Posted on Monday, August 23, 2004 - 02:06 pm: | |
If you read my posting of Aug 4 and implement the suggestions made there, you will see that the normal of grad(u) is consistently 5 in the outer material (see attached). The value inside the inner sphere is necessarily zero, because it is inside a charged sphere. Your recently posted script did not include the "layer 1" or "layer 3" controls necessary to require evaluation in the outer material, so the upper surface was again evaluated inside the inner sphere. You have defined "K" but it is unused. If you want to consider material property differences, you must use Maxwell's equation div(D)=rho.
|
Alan J. Horston (alan)
Member Username: alan
Post Number: 7 Registered: 07-2004
| Posted on Saturday, September 04, 2004 - 07:18 am: | |
Thank you very mach! The programm workes well! |