CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   courant number for cfx transient run (http://www.cfd-online.com/Forums/cfx/22394-courant-number-cfx-transient-run.html)

 Mike April 4, 2006 10:04

courant number for cfx transient run

Hi Guys,

What I understood from early discussions, (I am new in this area and I am reading early discussions to understand better), Courant number is not important if I am using CFX transient run as long as I have a convergence within reasonable number of coefficient loops (like 10 coef.loop). Is that correct? I am asking this because, in my case I have courant number 999 and have convergence too. Should I worry about the courant number or not? I really need help on this and previous questions that I posted yesterday.

Thanks alot,

Mike

 Robin April 4, 2006 13:05

Re: courant number for cfx transient run

Hi Mike,

That is correct. If you are converging within 3 to 6 coefficient loops, there is no need to worry about Courant number.

Regards, Robin

 James Date April 4, 2006 13:16

Re: courant number for cfx transient run

Mike

If you are converging within a small number of coefficient loops and still have a high courant number, that suggests to me your solution is quite steady. May be restarting using the steady state solver option might give you even quicker convergence.

James

 Mike April 4, 2006 13:32

Re: courant number for cfx transient run

Robin and James, Thank you so much for your help. Actually, James, I started with steady run and after trying everything and not getting convergence to the desired values, I conclude that the flow is unsteady by its nature. That is why now I am running as transient. In my case, the Re varies alot in the domain, flow starts in a rod shape nozzle with high velocity (highest 2 m/s)(water) and it exits (expands) to a domain which the area is more than 20 times larger than tube diameter. The whole domain is order of 3 meters. Thus, even the determination of time scale is very hard. It is basically the mold flow simulation during casting steel.

I am basically using SST model, CFX with 0.01 time scale for steady run and than with this initial conditions trying to run transient with 0.05 time scale. thats why I am getting very high courant number. But I was not sure whether it is critical or not. Thank you very much again. And any additional comments, helps and ideas would be very much appreciated.

Thanks again,

Mike

 TB April 5, 2006 02:16

Re: courant number for cfx transient run

Robin,

For this comment "converging within 3 to 6 coefficient loops", what target residual are you referring to? 1e-4 MAX?

 aloha5i April 5, 2006 07:04

Re: courant number for cfx transient run

In my opinion, it depends the purpose of your simulation. If you just want to get the time averaged solution, you could ignore the importance of Courant number when you have a converged solution. HOWEVER, if you want to capture the transient of unsteady effect, you'd better have a fair small courant number.

By the way, I wasn't sure what I understand about courant number. If I am wrong, pls correct me!

 James Date April 5, 2006 10:01

Re: courant number for cfx transient run

You're spot on Aloh5i, if you want a time accurate solution you should really have a courant number of less than one, unless the problem is approaching a steady solution

 James Date April 5, 2006 10:02

Re: courant number for cfx transient run

You're spot on Aloh5i, if you want a time accurate solution you should really have a courant number of less than one, unless the problem is approaching a steady solution.

 TB April 6, 2006 23:48

Re: courant number for cfx transient run

hmmm....There're two different opinions here....

1. Courant number MUST always be less than one for time accurate solution. The question is: Is it always practical? Life is too short for me.

2. Courant number could be more than one for solver using implicit approach. The question is: Is solution always accurate?

What I will do is: Rerun a transient simulation by doubling & halving the time steps. If solution doesn't change too much, then you can reasonably assume that it's time accurate. Any one has better & quicker approach to check the time accuracy of a transient solution?

 All times are GMT -4. The time now is 13:46.