[fredrikh]

Hi

I have implemented the sliding mesh technique into coodles, and it seems to work fine. I will use the sliding mesh technique to simulate a radial turbine, and it works when I run the code in seriel mode, but I have problem when I trying to run it in parallel. I have tried different strategies when I decompose the case with different results:

-The moving part and all cells connected to the sliding interfaces on one processor, the solver starts, but it stops after it had read thermodynamics package (before the first time step )

-I have decomposed the geometry in the axial direction; the code runs in parallel but the slidingInterfaces seems to treated as a wall with no slip

-I have decomposed the geometry in sectors; the code runs but the slidingInterfaces seems to treated as a wall with no slip, and

- I have tried the metis function, and the code doesn't run at all.

- I have also tried different way of decomosing the mixer2D case without any succes

Can somebody give me some advice how to decompose the geometry when I use the sliding mesh technique?

Thanks in advanced

Fredrik

[hjasak]

Sliding interface in parallel, where the processor boundary cuts the sliding interface is currently not supported. I have been looking for someone to sponsor this work but no success so far.

As you may imagine, there's quite a lot to do in order thandle this without limitations on the decomposition - if you'd like to get involved, please let me know.

Hrv

[fredrikh]

Hi Hrv

I have to talked to my professor before I promise anything....

I understand that it can be some problem when the processory boundary cuts the sliding interface, but why doesn't it work when I have all sliding interfaces, adjecent cells and the whole rotating part on one processor?

/Fredrik

[hjasak]

Well, put it this way: I wrote all the code but I don't run topo changes in parallel just yet. It is almost certainly a synchronisation issue - one CPU decides to do a topo change but others don't know about it - meaning it should be pretty trivial to fix if you have the patience to find out exacly what happens.

In any case, some digging through the source may be required and I'm not sure if you fancy doing it.

Hrv

[fredrikh]

Hi Prof.Jasak

I have started to look on the sliding mesh technique and how to get it running in parallel.

If I have understand it right, one problem is that the number of faces is changed (due to sliding mesh algorithm), which causes problem with the processor boundary patches. Right or wrong?

Do you have any suggestions how to solve this problem (if this is the problem).


Best regards

Fredrik

[sradl]

Dear Frederik,

I'd like to also use an (incompressible) LES solver with sliding mesh (stirred tank simulation) with use of parallel computers.

Have you made any progress with the parallel computation issue?

Can you give me as a departue point your implementation of the sliding mesh technique for oodles?

Cheers
Stefan

mfg
Stefan Radl