| Author | Message | ||||
     Chris Walton (ccwalton) Member Username: ccwalton Post Number: 11 Registered: 01-2008 |
This question is about how overwriting part of one region with another region (in a heat conduction problem) affects boundary conditions. I believe it's supposed to work this way: if I first write Region 1 with some thermal BCs, enclosing the whole domain, then overwrite part of it with Region 2 and give no BCs for Region 2, then FlexPDE will do two things: a) for boundaries of region 2 that are inside Region 1, T will be made continuous across the boundary, modeling perfect thermal contact and b) for boundaries of Region 2 that lie along a boundary of Region 1, the previously-declared BC from Region 1 will be used. If on the shared boundary a new BC is declared for Region 2, it replaces the previous BC from Region 1, and also remains in effect for the rest of the definition of Region 2 until changed. I ran the attached simple case to test this: a cylinder with constant T0bot on the bottom, convective loss on the curved surface, and constant T0top on the top, then the bottom half is overwritten with a new region with different thermal properties. That works fine (VersionA attachment). But if as an experiment I write natural(temp) = 0 at the start of Region 2 (VersionB attachment), it would seem it would make the bottom, side and top of Region 2 all adiabatic and Region 2 should stay isothermal. Instead the natural BC seems only to change the convective loss out the side - the fixed T0bot at the bottom remains in effect. Is there a rule for some BCs getting overwritten by later regions and others not? It seems this is more complex than the documentation indicates. -Chris Walton
| ||||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 1300 Registered: 06-2003 |
We reccommend that users make region 1 the outer system boundary and place all BC's there specifically to avoid such uncertainties. Natural()=0 is the default condition on all boundaries, so on boundaries where there is a Value() condition and a Natural()=0 condition, the Natural is ignored. If you declare a second Value() condition on region 2, then you should get an error diagnostic "conflicting BC". Admittedly, this seems to be inconsistent with the Region property of overlaying and hiding, but that's the way it's been for many years, so we're stuck with it. We'll consider changing it in version 7. | ||||
| Chris Walton (ccwalton) Member Username: ccwalton Post Number: 12 Registered: 01-2008 |
Thanks - that clarifies it. Is the following a full and accurate summary? It's best to put in the final BCs of all exterior surfaces from the start, but when overwriting a BC: 1) A natural() will override a previous natural() 2) Value() BCs are dominant over natural() BCs: a value() will override a previous natural(), and a natural() trying to override a value() is ignored 3) A value() trying to override a previous value() will give an error. May I suggest writing that in the manual somewhere? I guess one can find it all now with enough reading, but it seems fundamental enough to be worth stating clearly in one place. Unless I missed it. Thanks, Chris Walton |
2-layer thermal slab Version A