[biro]

Hello,

I have problems with my simulation in CFX.
I start a simulation with a rough mesh. The simulation starts, but there alternating variables (See the attachment).
With the results from my first simulation I think, that I have to refine my mesh.

After refining, the simulation stops after second iteration (" c_fpx_handler: Floating point exception: Overflow").
In my opinion the mesh has sufficient quality. But I'm a beginner with CFX and ICEM, so I can be wrong.
Can anybody help me ? Are there some more details required ?
Should I upload some mesh data ?

Thanks for your help

Greets biro

[Martin_Sz]

Can You write about BC's which You use in simulation ?

[biro]

I use the k-epsilon turbulence-model.
On the inlet I pretend a mass flow.
The outlet is executed as opening with pressure and temperature specification.

Are there more information you need ?

[Far]

What is the problem with mesh? did you check quality metrics?

[biro]

I don't know exactly the error.
With the rough mesh my simulation will work, but with the refine mesh the simulation doesn't work.
Based on the error message in CFX and the fact, that the simulation with the rough mesh is running, I thought its a problem with the mesh.

There are no bad angles and the quality is better 0.2.
The 3x3x3 determinant is better then 0.6 and when I checked the mesh in ICEM, there will be no errors.
I'm new with ICEM and maybe there are more facts, that could be responsible for the problem, but I don't know them.

[Far]

Might be problem is transient. In bad mesh (coarse mesh) flow unsteady is not captured and hence no problem in convergence. But with fine mesh unsteadiness is creating problem for steady solver .


Past some pictures of mesh...

[biro]

There are some pictures

mesh1 - mesh complete
mesh2_inlet - part of the inlet
mesh3 - intersection to box
mesh4 - intersection to pipe
mesh4 - intersection to pipe (zoom)

[biro]

Originally Posted by Far

Might be problem is transient. In bad mesh (coarse mesh) flow unsteady is not captured and hence no problem in convergence. But with fine mesh unsteadiness is creating problem for steady solver .

How can I solve the problem ?
Are there any issues I have to follow by refining the mesh ?

[Far]

what you are trying to model at the outlet with cylinder?

[biro]

It's the geometry of the test-track.
At the end there is that cylinder

[Martin_Sz]

On the first step I will check this simulation for SST turbulence model.
In that kond of model a convergence criteria is more elastic.
When that change dont give sucess results - next case I will change outlet BCs for outlet with static pressure.
Best regards

[biro]

I found a solution.
I have made a variation of element size. So I discovered, that the simulation works only in one case.
In this case the mesh is rough and at the wall its a bit more defined.

Then i can start the refined mesh by choosing the rough solution as start solution.

[gerardosrez]

In cfx generally the alternating residuals during iteration process is due to a large timestep (false timestep if steady state simulation). You can try to reduce it one order and check how your iteration process go....

Best regards...

[biro]

Thank you for you answer.

But I don't understand steady state and timestep? What does "timestep" means in this context ? And what are the criteria to choose it ?

Still any other question:
Is it better to use a previous simulation as initial values at all time, even if the simulation will run without initial values ?
Or should I start such simulation without initial values ?

Greets