CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CD-adapco (http://www.cfd-online.com/Forums/cd-adapco/)
-   -   Convergence in MARS scheme (http://www.cfd-online.com/Forums/cd-adapco/56383-convergence-mars-scheme.html)

shiv December 10, 2007 20:46

Convergence in MARS scheme
 
is it possible to get the 1e-6 order of convergence in case of compressible flow with MARS scheme for porous medium. if it is, please give me the details. if not, then what is the max convergence can achieve and how it is?

very urgent

Thanks a lot in advance

BastiL December 11, 2007 03:16

Re: Convergence in MARS scheme
 
Hi,

this is also a question of mesh quality and not only of used models.

shiv December 11, 2007 04:03

Re: Convergence in MARS scheme
 
Hex mesh with Hypermesh quality criteria is used in complete model. i checked the mesh. No problem in mesh and star-cd also accepted this. I think mesh is good. I need that is it possible to get 1e-6 order of convergence in MARS scheme. if it is, give me the suggessions.

allan December 11, 2007 06:24

Re: Convergence in MARS scheme
 
why do you need this level of convergence?

If your grid is good i.e. fine you may be picking up very small transient effects that are not possible to resolve in a steady state simulation. You may need to solve transient.

allan

ping December 11, 2007 10:40

Re: Convergence in MARS scheme
 
and the convergence data is normalized based on the first few iterations, so if the initialisation is close to the final solution convergence levels will not fall much at all. You should be monitoring other variables and plotting results to see if these change after a period of iterations.

Balduin Bankerotti December 11, 2007 18:42

Re: Convergence in MARS scheme
 
The checks only tell you that the cell shapes are valid, but nothing about the real mesh quality.


shiv December 11, 2007 19:48

Re: Convergence in MARS scheme
 
Allan, Ping, Balduin Thanks for ur response.

I cant solve this problem by transient method, b`coz number of element is so high that it may take lot of time.

I have monitored u,v,w, pressure, enthalpy at the outlet.

i know that 1e-6 is not necessary, once the solution and pressure, momentum and enthalpy equations are stable we can take the result. but still i am asking that, even after achieving stability also because of turbulence there is fluctuation in the residual and monitor values. In that case what i have to do? In MARS scheme how to get this level of convergence?

allan December 12, 2007 05:23

Re: Convergence in MARS scheme
 
You do not say whether you've achieved this level of convergence with UD. If you have and cannot achieve it with mars it's probably a grid problem.

I do not fully understand your question, but, if you're saying that your residuals are not falling and you still have a fluctuation in steady state values, then what is the value of the fluctuation 1%, 5%, 50%? If it is small, it is probably due to numerical round off error and, does it matter?

You could try solving in double precision.

allan

shiv December 12, 2007 20:03

Re: Convergence in MARS scheme
 
I have not achieved 1e-6 in UD. I monitored the values at outlet. And also fluctuation is because of recirculation near the outlet. fluctuation is also very small value and it is almost between 1% to 1.5%. all residuals are constant.

allan December 13, 2007 06:11

Re: Convergence in MARS scheme
 
how close is the recirculation to the outlet? try extending your domain so that the outlet is far field from the recirculation. A change of about 1% in your values is probably as good as you're going to get on that grid. If you want more accurracy then refine your grid, run in double precision.

allan

shiv January 31, 2008 21:27

Re: Convergence in MARS scheme
 
Hi Allan, While solving the problem using MARS scheme, how the under relaxation factor will effect on the covnergence and on the final result. Because in my problem for MARS scheme, URF for momentum is 0.2, for Pressure is 0.4, for turbulence is 0.2 Temperature is 0.5 and density os 0.1, and i ran the solution for more than 7500 iteration under this condition.

Thanks in advance


All times are GMT -4. The time now is 23:58.