CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Effect of iterations on result (http://www.cfd-online.com/Forums/main/91289-effect-iterations-result.html)

sheth August 5, 2011 15:53

Effect of iterations on result
 
Hi folks I was wondering what is the effect of iterations on the result.

Does increasing iteration increase accuracy of result?

cfdnewbie August 8, 2011 19:59

generally:no! why should it?
exception: implicit code, where your matrix inversion has not converged yet....

what kind of iteration (physical / computational) did you mean?

husker August 9, 2011 08:25

Hi,

Could you let me ask what the physical iteration is?

seyedashraf August 9, 2011 14:56

if you know how to make the whole system and the results to converge
yes more iterations could bring better results

but you may also consider the time/cost of the accuracy

cfdnewbie August 9, 2011 17:20

Quote:

Originally Posted by husker (Post 319466)
Hi,

Could you let me ask what the physical iteration is?

I was sloppy with my language there. What I meant was a physical time step in your iteration, i.e. a delta t.

Ford Prefect August 10, 2011 16:07

Do you run a transient or steady state solver?

If the convergence is well behaved then more iterations will generally lead to lower residuals. This is a good thing (to a point) in a steady state solver, but it might not be necessary within a transient solver. I would suggest that you put an iteration number so that you will reach convergence for your variables within most of the time-steps.

sheth August 15, 2011 12:49

Thanks for answers everyone.

ertan August 20, 2011 18:40

Quote:

Originally Posted by Ford Prefect (Post 319648)
Do you run a transient or steady state solver?

If the convergence is well behaved then more iterations will generally lead to lower residuals. This is a good thing (to a point) in a steady state solver, but it might not be necessary within a transient solver. I would suggest that you put an iteration number so that you will reach convergence for your variables within most of the time-steps.


If I am wrong please correct me.
If you set an "iteration number" (I guess, you meant maximum iteration) and the solution does not converge to a desired value within a time step, then iteration error will start accumulating. Unless those errors are convected out with the flow, and the solution starts converging to the desired value, your simulation will have high iteration error. For an accurate simulation, generally the minimum number of (non-linear) iterations is set.


All times are GMT -4. The time now is 22:49.