CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Please help me out: Linear solver overlow (http://www.cfd-online.com/Forums/cfx/26546-please-help-me-out-linear-solver-overlow.html)

Hao October 20, 2008 01:24

Please help me out: Linear solver overlow
 
I am running the simulations of a turbulent jet flow. The simulition is successfully finished when i use 1e-3 sec as the time step. However, when i change to use 1e-4 sec (with the same setup of the others), there is always a overflow problem. I thought it was due to the mesh quality. So i checked it in Ansys ICEM. It said the quality was good. Does anyone confront this kind of problem? please give me some clue. Cheers!

John S. October 20, 2008 11:19

Re: Please help me out: Linear solver overlow
 
This can sometimes happen if you have too fine a mesh as well. Check your domain advection time. A typical time step is usually 1 advection time whereas 1/3 of the advection time would be considered small but can aid convergence.

Glenn Horrocks October 20, 2008 21:29

Re: Please help me out: Linear solver overlow
 
Hi,

Using either a smaller timestep or finer mesh often causes the simulation to loose stability. There are a number of reasons but the most common one is that the mesh is now starting to resolve turbulent eddies. This leads to a minimum size allowable for timesteps. That is why it is recommended to increase the timestep agressively in steady state simulations after the initial transient is under control.

Glenn Horrocks

Hao October 21, 2008 00:18

Re: Please help me out: Linear solver overlow
 
Thanks, Glenn. When you said 'increase the timestep agressively in steady state simulations after the initial transient is under control', do you mean that i should do a steady-state simulation first to get some results to set as the initial conditions for the later transient simulation?

Hao October 21, 2008 00:44

Re: Please help me out: Linear solver overlow
 
Thanks for the response, John. I have a rookie question: how to check the advection time of my domain? cheers.

Glenn Horrocks October 21, 2008 18:53

Re: Please help me out: Linear solver overlow
 
Hi,

You should model the physics as they are in the real world. If it starts from a steady state condition then yes, start with a steady state simulation. If it starts from a known condition (ie from rest, or some other known condition) then start straight into the transient simulation.

Glenn Horrocks

pmp77 November 12, 2008 06:08

Re: Please help me out: Linear solver overlow
 
Hao,

I'm new here but I have found this:

http://www.cfd-online.com/Forum/cfx_...cgi?read=28402


All times are GMT -4. The time now is 10:12.