CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   residuals (http://www.cfd-online.com/Forums/cfx/19716-residuals.html)

Karl June 23, 2003 14:32

residuals
 
For time dependant simulations in CFX, the residuals are just plottted with their value at the end of each time step. I would like to see how the residuals are reducing with each inner iteration within the time step. Is there any way to make them plot like this?

Also, CFX says that it does not recommend using more than 5 inner iterations per time step, and that you will probably not get increased accuracy with more. This seems like a rather low value to me. How can you tell if you are truly converged after each time step? From my experience with segregated solvers, you should continue inner iterations untill residuals stop decreasing -- but cfx does not plot residuals at inner iterations, and the form of the *.out file does not make it easy or fast to plot the inner iteration residuals.

Robin June 23, 2003 15:10

Re: residuals
 
Hi Karl,

The CFX-5 solver is not a segregated solver. It uses a fully implicit coupled multigrid method, which is much more effective. This also allows you to use a larger timestep than a segregated solver will allow.

I would advise you to start with a very small timestep. If you require fewer than 3 coefficient loops, you can increase the timestep. If you require more than 5, decrease it.

Regards, Robin


Neale June 25, 2003 00:13

Re: residuals
 
The rule of thumb for transient calculations is to use a smaller timestep before you increase the number of coefficient loops. You will always be better off this way.

For now, the only way to "see" inner loop convergence is to look at the output file during the run.

If you are not reaching your residual target or conservation targets within the coefficient loop then you need to decrease your timestep.

Neale


All times are GMT -4. The time now is 16:13.