CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Use of Timestep in obtaining solution. (http://www.cfd-online.com/Forums/cfx/22223-use-timestep-obtaining-solution.html)

 hagupta February 27, 2006 07:53

Use of Timestep in obtaining solution.

Hello all, I have the following queries regarding a STEADY STATE simulation in CFX.

1) Is "timestep x no of iterations" an estimate of the time required for the flow to evolve into a steady state flow?

2) Even if the RMS residuals have reached beyond the required level(say 10^-5) and the global imbalances below 1%, can the results(of atleast certain parameters) still change significantly if more iterations are given?

3) Does it make a difference if I give a higher value of timestep as long as the results converge? Is it ok to start with the auto timestep option and keep on increasing the timestep value regulary as the results start converging?

 Robin February 27, 2006 13:53

Re: Use of Timestep in obtaining solution.

1. Not necessarily, because a steady state is not an accurate transient.

2. Yes. You may have locally high residuals (check the MAX residual) indicating local changes and these could evolve into changes in the entire system. That said, if the global energy and momentum imbalances are very low, further changes are unlikely. I have seen cases where everything looks good, but a boundary layer is slowly growing and suddenly separates, causing a major change in the solution.

3. At convergence, the transient terms cancel out (left hand side equals the right hand side), so the timestep has no influence on the final solution. Timestep will effect the rate of convergence, however.

Since some features may take some time to develop, I usually recommed running with as big a timestep as you can get away with. This will help accelerate the slower transport processes, such as heat condution and viscous momentum transport. If the global balances are low and if you are still having trouble with rapidly changing residuals, you might consider dropping the timestep. A good way to estimate what timestep to use is to calculate streamlines from your inlets in Post and look at the time on the streamline. I'll often do this on a backup file and set the timestep to be equal to either the max or mean time on my streamlines, but you probably shouldn't go much lower than 1/10th to 1/50th of this characteristic time.

Regards, Robin

 Gab February 27, 2006 14:37

Re: Use of Timestep in obtaining solution.

Hi, Robin

How to calculate the time on the streamline in cfx-post?

Thanks!

Best regards!

Gab

 Gab February 27, 2006 14:46

Re: Use of Timestep in obtaining solution.

Another question to Robin:

Sometimes I need to set many monitoring points(and each point may have several monitoring parameters) in cfx-pre. However in cfx-solver I do not want to show all the user monitoring results. My question is, is there a function key to disable all the user monitoring terms and then allow me to just select one from the list?

Best regards!

Gab

 Glenn Horrocks February 27, 2006 17:26

Re: Use of Timestep in obtaining solution.

Hi Gab,

If you don't want to show all the monitor points then delete the default plot monitor showing all the point results and create a new one just showing what you want.

Regards, Glenn

 Gab February 27, 2006 18:15

Re: Use of Timestep in obtaining solution.

Thanks a lot, Glenn. It works.

Best regards!

Gab

 hagupta February 28, 2006 12:07

Re: Use of Timestep in obtaining solution.

Thanks Robin for clarifying my doubts. Very well explained!!

 Robin February 28, 2006 14:14

Re: Use of Timestep in obtaining solution.

Hi Gab,

When you compute a streamline, say "Streamline 1", Post adds a variable named "Time on Streamline 1". The streamline can be colored by this variable or you can select it in the function calculator when you have the streamline selected. Note that this variable is only avialable on the streamline it was created from.

Regards, Robin

 All times are GMT -4. The time now is 07:10.