CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   About convergence (http://www.cfd-online.com/Forums/cfx/21300-about-convergence.html)

LQ May 30, 2005 23:29

About convergence
 
I am using CFX to solve a governning stage of a steam turbine which have three domains : the strengthening ring,stator and rotor.

I solved the flow field of the stator first,then added the rotor, finally solved all the three components as a whole stage.

During the procedure,I found that the more domains there were, the longer time will elapse for the caculation to converge.

Is there anybody else has this experience and know the reason?

LQ May 30, 2005 23:35

Re: About convergence
 
I mean "the longer time" to the iteration steps.

Glenn Horrocks May 31, 2005 18:34

Re: About convergence
 
Hi,

I assume you mean a longer total physical timestep to converge. I can see this being caused by two factors, firstly adding extra bits to the model puts longer timescale effects into the model (that is flow interaction between blade rows); and secondly the rows are probably connected with GGI interfaces and they will require a little extra time to converge.

Regards, Glenn Horrocks

Robin June 2, 2005 23:43

Re: About convergence
 
By "longer time", do you mean wall clock time? I.e. it took more time to run? If so, the answer is simple: you are solving a bigger problem.

The solver run time scales linearly with with number of equations and the number of nodes. Double the nodes, it takes twice as long. Similarly, if you add equations to solve (such as heat transfer, or an additional variable), it takes proportionally longer for each timestep. This does not necessarily mean it requires more iterations. Each run make take 100 iterations, but each iteration get's a bit longer.

Regards, Robin


All times are GMT -4. The time now is 21:21.