[PeterA_]

Hello dear forum !

I have a problem concerning the immersed solid method in CFX.

What i am trying to do:
I want to simulate the opening/closing of a valve. I build up a quasi 2-D model (2D- 5 degree wedge) and defined the fluid and immersed solid regions.

During the simulation the valve is closing and therefor the valve cone is moving 2 mm.

Everything is working fine for the first 0,8 mm. Then the results diverge and the massflow becomes high/unlimited.

When i inspect the results short before the divergence i can see strange singularities in pressure/velocity right behind the moving valve cone. (see attached image).

What i tried sofar: variation of: mesh sizes / time steps size / momentum scaling factor

I set up a very similar simulation just with another cone angle (45° instead of 38°) and it runs without any problems. After days of trial and even more errors i have no clue what i could try.

I would be very thankful for any advice what i could try now.

Many thanks
Peter

[ghorrocks]

This shows your simulation has borderline numerical stability, so you need to make it more stable. Try:
* Improve mesh quality (this is the most important one - and I can see your inflation layers are not too good)
* Double precision numerics
* Smaller time step size

[PeterA_]

Many thanks for your fast reply.

* Double precision numerics
* Smaller time step size

already tried this without success (even very very small timestep has no influence)


Improve mesh quality: What i wonder about is that the Immersed Solid moves for 0,8mm through the mesh with the "same bad quality" without problems. I have also tried a finer mesh in the region where the immersed solid moves. I will try an even finer mesh and report you if this solves the problem.

Many thanks sofar.

Peter

[PeterA_]

Hello !

I refined the mesh in the region where the valve cone is moving.

The same problem occurs - at another timestep and another position of the cone but the problem still exists.

Do you think its still a problem of mesh size/quality ? As i mentioned before a very similar modelsetup run without any problems with a much much coarser mesh.

Many thanks

Peter

[ghorrocks]

Your problem is numerical instability.

Refining a mesh makes it less stable, so problems with a finer mesh are to be expected.

The suggestions I made (mesh quality, double precision, smaller time steps, better initial condition) are the usual suspects for improving numerical stability. You can always improve mesh quality - and it will make the biggest difference if you do. In this case as this is a 2D simulation I would recommend trying hex elements as they are often more stable then tets/prisms.

There are a few settings specific to immersed solids - I am not too familiar with them so cannot make a suggestion but I would try some of them.

[Martin_Sz]

can You copy of error information which You have in solver ??

[PeterA_]

Hello !

I tried a pure hex mesh in the region, where the valve cone is moving.

Same error .....

Error message:

Parallel run: Received message from slave
-----------------------------------------
Slave partition : 6
Slave routine : ErrAction
Master location : RCVBUF,MSGTAG=1214
Message label : 001100279
Message follows below - :

+--------------------------------------------------------------------+
| ERROR #001100279 has occurred in subroutine ErrAction. |
| Message: |
| Floating point exception: Overflow |
| |
| |
| |
| |
| |
+--------------------------------------------------------------------+

Parallel run: Received message from slave
-----------------------------------------
Slave partition : 6
Slave routine : ErrAction
Master location : RCVBUF,MSGTAG=1214
Message label : 001100279
Message follows below - :

+--------------------------------------------------------------------+
| ERROR #001100279 has occurred in subroutine ErrAction. |
| Message: |
| Stopped in routine FPX: C_FPX_HANDLER


What i am confused about: If i change the cone angle from 37° to 45° the simulation run without error with nearly any meshes. This might be an annoying coincidence.

[Martin_Sz]

You can try to change turbulence model to SST if You have k-epsilon
next way - on the domain turn on mesh deformation
What do You have on outlet - opening or outlet ??
Best regards

[PeterA_]

Hey,

many thanks for your advices !!

I already use the SST-model.

The outlet is defined as outlet / no message about back flow occurs.
I will try an opening and report you if this solves my problem.

Mesh deformation: I did not tried this sofar. I have to read a bit about this topic and will also report if i could solve the problem.


Offtopic: I am still everytime amazed about the possibilities of the internet - within a second its possible to get in touch with persons - perhaps 10000 km away - and ask for an advise GREAT !!


Many thanks sofar

Stephan