CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Question about convergence (http://www.cfd-online.com/Forums/cfx/115654-question-about-convergence.html)

Clay April 4, 2013 03:45

Question about convergence
 
I am new to CFX and I am simulating a plate heat exchanger. Which parameter/variable should I be looking at for the convergence? By default, the plot of momentum and mass of two fluids and heat transfer are generated. The plot of momentum and mass shows sinusoidal pattern. While heat transfer cannot achieve convergence. Is it necessary for these three parameters to reach convergence? I am confused.

ghorrocks April 4, 2013 04:40

For conjugate heat transfer simulations convergence by both residuals and imbalances is a good idea. Often the residuals are tightly converged but the overall imbalances are miles off. This is very common in CHT simulation due to the different time scales in the fluid and solid domains.

Clay April 4, 2013 04:59

Quote:

Originally Posted by ghorrocks (Post 418265)
For conjugate heat transfer simulations convergence by both residuals and imbalances is a good idea. Often the residuals are tightly converged but the overall imbalances are miles off. This is very common in CHT simulation due to the different time scales in the fluid and solid domains.

I'm sorry but I still do not understand. I would like to show the plots but the solver is still running. It has been 27 hours for 190 iterations. The momentum and mass shows sinusoidal pattern and I don't think they will converge even when the calculation is complete.

ghorrocks April 4, 2013 06:16

Oh, in that case your question is a FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

Clay April 4, 2013 11:39

I had read that a few times. But I'm trying to ask if I'm looking at the correct parameter to converge? And is it necessarily for momentum and mass plot to converge?

oj.bulmer April 4, 2013 15:50

Isn't it obvious? Yes they do need to converge, but it depends on definition of "convergence".

1) Typically, your residuals for continuity, momentum, k, epsilon, energy etc should be reduced to O(1e-4). Smaller if you want better results.
2) Your monitors should be flat within the tolerance you are happy with and imbalances should be less than 1%

#1 can be difficult sometimes and you may see oscillating pattern and stay between 1e-3 and 1e-4. As long as the monitors are sufficiently flat and imbalances small, you should be alright.

OJ

zegtuhetmaar April 5, 2013 05:32

Try to "play" during the simulation with the time scale (Even steady calculations have time steps). Try making it larger. In case the convergence lines goes up and down every time step, you have too large timescale.


All times are GMT -4. The time now is 19:03.