Heat conduction, discontinuous boundary

Numerical methods and mathematical models of Elmer
c_ross
Posts: 12
Joined: 24 Feb 2021, 19:32
Antispam: Yes

Re: Heat conduction, discontinuous boundary

Post by c_ross »

Progress report, comments, questions:

I have made progress introducing a thermal discontinuity. My problem has nine bodies but there is only one face of one body where I am interested in a discontinuity. I define 8 bodies as "master" and one as "slave".

I believe I have to define the whole boundary of the single body as discontinuous? I do so by defining two boundaries: one for the face I want to study and one for all the rest.

I have noted that HeatSolveVec is almost 10 times faster than HeatSolve for my problem with no discontinuities. Is this expected?

The heat transfer coefficient for the interface is specified using the key word "Heat Gap Coefficient". Earlier examples seemed to use "Heat Transfer Coefficient"? In any case, there seems to be no documentation on either "Heat Gap" or "Heat Gap Coefficient".

Any feedback appreciated.

Thanks, Carl
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Heat conduction, discontinuous boundary

Post by raback »

Hi

It is faster because this code includes many improved techniques that we have developed over the years. Still, ten seems bit extreme but I would be happy with that. The solver has not been documented since it is mainly just modernized version of HeatSolver. The ability to deal with this discontinuity stuff is probably the only added feature. There are many missing features compared to HeatSolver which will recieve a Fatal if used.

The "Heat Gap Coefficient" is basically the same as "Heat Transfer Coefficient" except it is between two boundaries whereas latter is between some external temperature. Basically these could be different even though they probably are not. One thing is that heat gap is only possible when solved as a DG system and hence for error checking it is wise to separate them. Otherwise the code does not know whether the user needs discontinuities.

-Peter
c_ross
Posts: 12
Joined: 24 Feb 2021, 19:32
Antispam: Yes

Re: Heat conduction, discontinuous boundary

Post by c_ross »

Thanks, Peter.

"Heat Transfer Coefficient" and "Heat Gap Coefficient" give very different results. I had started with "Heat Transfer Coefficient" because I had seen it used in older examples. However, the result made no physical sense.

"Heat Gap Coefficient" gives results that are consistent with expectations (and it is used in recent examples).

Carl
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Heat conduction, discontinuous boundary

Post by raback »

Well, what I meant is that they are physically the same parameters but have different reference temperature: the other side vs. given external temperature. That is why they should also give different values. -Peter
c_ross
Posts: 12
Joined: 24 Feb 2021, 19:32
Antispam: Yes

Re: Heat conduction, discontinuous boundary

Post by c_ross »

I have made some progress with discontinuous interfaces.

With a number of master bodies (8) and 1 slave body, things seem to work as expected.

However, I have two bodies with a common discontinuous interface. If I define them as two slave bodies then the common interface is not considered to be discontinuous.

I thought perhaps I could use "DG Reduced Basis Mapping" to avoid defining Master and Slave bodies. However, I do not understand the description associated with this keyword and ElmerSolver objects, no matter what values I use.

Can I avoid these keywords and define EVERY interface as discontinuous? For my problem, I do not have a ridiculously large number of interfaces.

(Maybe I have a mental block when it comes to the documentation in ElmerSolver Section 14.5)

Thanks, Carl
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Heat conduction, discontinuous boundary

Post by raback »

Hi

Maybe you could draw a picture...

The most discontinuous field is of course the initial Discontinuous Galerkin field. This reduced basis was developed as a compromise between fully discontinuous and fully continuous. Maybe you would be best served with the fully discontinuous flavour. The solver should work with "Discontinuous Galerkin = True" with the other funny stuff removed.

-Peter
Post Reply