CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CFX (https://www.cfd-online.com/Forums/cfx/)
-   -   overflow problem (https://www.cfd-online.com/Forums/cfx/65455-overflow-problem.html)

 Marteusz June 15, 2009 15:16

overflow problem

Hello everyone,

At the beginning I want to mention that I have read all history about this error and the article at CFD-Wiki and all advices didn't solve that problem

Problem to solve:

http://student.agh.edu.pl/~marteusz/problemtosolve.png

simple pipe 6mm diameter, 1m long, inlet velocity 825m/s and static relative pressure 5bar

here you can see run definition settings:
http://student.agh.edu.pl/~marteusz/definition.txt

Domain: Ideal Gas
Heat Transfer: total energy
Turbulence model: K-epsilon

BC
Inlet - supersonic 825m/s and relative pressure 5bar
Outlet - supersonic

Initial values, 825m/s and temperature 20C

I had changed local timestep from 1E-3 to 1E7 to see if it help, but it doesn't.

Residuals target: 1E-05

Can anyone help me to get convergence, I will be very, very glad.

Problem can be cause because too small diameter and too big velocity.

Thanks,

Mateusz Kesek

 ghorrocks June 15, 2009 20:23

Hi,

Try using Local Timescale Factor to get the thing started. Once it has converged for a bit using that for a while go back to a physical timescale.

Also consider using the high speed numerics option. It is an expert parameter which does a second continuity loop and that occasionally helps with high speed flows.

Glenn Horrocks

 Timon June 16, 2009 03:22

Actually, the second continuity loop is activated by a separate expert parameter:

max continuity loops = 2

High speed numerics (found under compressibility control in the advanced solver control panel) does three other things. Copy-paste from the help:

"Firstly, it activates a special type of dissipation at shocks to avoid a transverse shock instability called the carbuncle effect (which may occur if the mesh is finer in the transverse direction than in the flow direction). Secondly, it activates the High Resolution Rhie Chow option to reduce pressure wiggles adjacent to shocks. Finally, for steady state flows, it modifies the default relaxation factors for the advection blend factor and gradients."

 ghorrocks June 16, 2009 08:01

Thanks for the correction Timon, it's been a while since I used that option so I forgot the details!

Glenn Horrocks

 Marteusz June 16, 2009 13:20

I still have this problem, I discovered that If I turn off the turbulance (laminar flow) or increase diameter of pipe the analysis gets convergence.

But I need to get convergence to that small tube

 Timon June 17, 2009 03:21

Have you tried to initialize your solution with lower velocities, ie. gradually increasing your boundary conditions until you reach the desired values?

 fab June 17, 2009 05:49

geometry

hi everybody,i am just a CFX-beginner,i have some questions about the Geometry,i have to design a Flowchanel!can some body help me please?
tanks lot

 ckleanth June 17, 2009 08:14

Quote:
 Originally Posted by fab (Post 219574) hi everybody,i am just a CFX-beginner,i have some questions about the Geometry,i have to design a Flowchanel!can some body help me please? tanks lot
do the tutorials before asking any questions

 LSC June 19, 2009 00:41

Hi

I do encountered convergence issue quite often (playing with the values for the past month). I realised that using Local timescale indeed helps a lot but for some of my simulations the residual graph is diving smoothly until in the 1e-4 to 1e-5 region it starts to oscillates. I tried to tune the solver fluid and mass relaxation in the expert parameters but it wont help much. Also noticed that physical time scale is much faster but oscillations are often encountered. Are there any best known methods to tackle this issue?

 ghorrocks June 19, 2009 07:02

http://www.cfd-online.com/Wiki/Ansys...gence_criteria

 LSC June 19, 2009 07:42

Hi Glenn,