Discussion Forum

Hi

you can go into the solver segregated nodes and group the default two (or more) groups into one, manually, by adding the parameters to the first group and then deleting the second, once all variables have been transferred. That is what COMSOL proposes by such a message, I have tried a few times, and it sometimes works

--
Good luck
Ivar

Thanks, doing that seems to have moved the calulations closer to a solution BUT

now I get: Undefined value found.
- Detail: NaN_or_Inf_found_when_solving_linear_system_using_SOR

I assume there is now an illegal opeations occuring, but how to I resolve this error??

Thanks Again..

Hi

Thats tougher, perhaps something like a 1/0 in the equations somewhere. Do not forget that often variables are fields f(x,y,z) and are mapped on the mesh elements via their locations x,y,z hence you can get easily some singularities

--
Good luck
Ivar

So I tried a new approach and attempted to recreate my model from scratch using the Conjugate Heat Transfer-Low Re K-E physics model rather than linking the separate HT and Turbulent physics.

BUT

I am not able to select any part of my geometry in the Turbulent portion of the Conjugate module in order to set my inlet, outlet, and wall conditions. I am able to select geometry to set a heat transfer boundary (set temperature), but noting for the fluids side. Is there a work around for this or some kind of unique procedure required to model in the conjugate physics???

Thank You,
Brett

Hi

that's bizarre, I can select external boundaries and apply inlet, outlets ... under the physics Laminar as well as Turbulent Flow branch (depending on turbulent switch settings) are you in the latest patch 3 ? v4.1.0.185 ? could that be it ?, in case see the main web site "Support Product updates"

--
Good luck
Ivar

Hi,

Okay I fixed the boundary selection problem (it was me, not COMSOL), but now that I have a resonable looking solution from the Conjugate HT/Low Re k-E module I keep getting this warning:

New constraint force nodes detected: these are not stored !

Are these Reynolds stresses and/or does 'new' refer to succsesive iterations of the GMRES or somethig?

I can't tell how to interept the impact of this warning on the solution generated.

Any ideas?

Thank You!

Hi

I believe that is linked to the ALE (do you use ALE), from my understanding its not the most critical warning, could imply lower precision on some of the flux calulations

--
Good luck
Ivar

thanku you Ivar.it helped me..

Hi

you can go into the solver segregated nodes and group the default two (or more) groups into one, manually, by adding the parameters to the first group and then deleting the second, once all variables have been transferred. That is what COMSOL proposes by such a message, I have tried a few times, and it sometimes works

--
Good luck
Ivar

Hi,

I am having problem with segregated solvers as well, but what you suggested, doesn't that defeat the purpose of using segregated solvers? if you put all the parameters to a single group than that would make it a (kind of) fully coupled?

Thanks

Hi

indeed its worthwile playing a little with the segregated groups, or fully coupled mode, but to sort the items, think over the logic in your moel, what is master what is slave, why ...

--
Good luck
Ivar