CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   TIMESTEP PROBLEM IN TRANSIENT SIMULATION (http://www.cfd-online.com/Forums/cfx/19861-timestep-problem-transient-simulation.html)

George September 1, 2003 15:24

TIMESTEP PROBLEM IN TRANSIENT SIMULATION
 
Hi everybody

I have a simple doubt...

For example, if I want to do a simulation that has a total time of 60 seconds and I only wanna know the state at 30 seconds and the final state (60 sec).

Will I have different results using a time step of 5 seconds or using 30 secons time step?

Thanks

Glenn Horrocks September 1, 2003 18:15

Re: TIMESTEP PROBLEM IN TRANSIENT SIMULATION
 
Hi George,

Whether the timestep sizes you mention make a difference depends on your problem. You will have to determine that with a timestep independance test.

If you are only interested in the results after 30 and 60 seconds, set up a transient result file to only output these two timesteps.

Regards, Glenn

Forrest September 2, 2003 05:32

Re: TIMESTEP PROBLEM IN TRANSIENT SIMULATION
 
Hi, George

As Glenn mentioned, a suitable timestep selection is problem based. You need to be sure if you have used a proper timestep size to get your problem resolved accurately. If a timestep size of 30 seconds can be suitable for your case, other timestep selections smaller than 30s won't give you different results at 30s and 60s.

Regards, Forrest

Robin September 2, 2003 09:33

Re: TIMESTEP PROBLEM IN TRANSIENT SIMULATION
 
Hi George,

When running a transient simulation, a suitable timestep should be selected in order to accurately model the transient physics. This is often smaller than the periods where you desire output.

Since fluid timescales are very small, I doubt you will get meaningfull answers at a timestep as large as 30 seconds. Even 5 seconds seems to be too large.

If you look back through the posts on this forum, you should find several discussions regarding transient timestep selection. Generally, you should use a timestep which is sufficient to solve the linearized equations to a residual of ~1e-3 within 3 to 5 coefficient loops.

As for output, there is no need to write output files at every timestep. You can specify timestep intervals where output is to be written. Personally, I would write out more than just at 30 and 60 seconds for your case. Since you doing the work anyway, write out some partial results files at, say 1 second intervals, and full output files at 30 and 60 seconds.

Good luck.

Regards, Robin

BAK_FLOW September 9, 2003 10:45

Re: TIMESTEP PROBLEM IN TRANSIENT SIMULATION
 
Hi George,

along with the above advice, make sure that you use a second order accurate time stepping scheme. As with first order schemes in space (UDS), first order schemes in time often lead to unacceptable discretization errors even with small time steps. There is an excellent chapter on this in Ferziger and Peric's text: "Computational Methods for Fluid Dynamics"

Regards,

BAK_FLOW


All times are GMT -4. The time now is 16:07.