CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Convergence tips (http://www.cfd-online.com/Forums/cfx/21553-convergence-tips.html)

 John August 25, 2005 12:02

Convergence tips

Hello,all

In CFX10 help file, I found exactly the same convergence tips for multiphase modelling(bubble column) as in CFX5:

(ANSYS CFX 10 Best Practices Guide for Multiphase, page 132): "Sometimes, physical instabilities (such as recirculation zones) can result in slow or stalled convergence. In these cases, you can still obtain an indicator of convergence for a global quantity by creating a monitor point at some point in the domain. As a result, you can determine whether values for selected global quantities (such as gas hold-up) are meaningful."

Does this mean that the convergence critira(eg. RMS) is not so important(or critical) for multiphase problems?

I asked the same question in a previous post, but I still fell confused about these words.

The answer to this question is important to my current work. I would appreciate your inputs!

Regards!

John

 John August 26, 2005 09:49

Re: Convergence tips

It may be a naive question but I hope someone can explain it. In literatures I did see some people use liquid height level(eg. for sieve trays) or holdup as indicator parameters to terminate the run.

Regards!

John

 Jeff August 28, 2005 00:41

Re: Convergence tips

Multiphase simulations are, by nature, tansient. However, the transients can be very localized. Usually this is due to small flow oscillations or oscillating bubble streams. When this happens, the RMS residuals will drop to a certain level and then show a repeating oscillating pattern. The residuals can't be reduced any further because there simply isn't a non-oscillating steady-state solution. You have two recourses:

1) after the residuals have "bottomed out", plot velocities at various locations, and total gas volume (hold up). When these values become steady, or reach a repeatable oscillating "pseudo-steady-state" pattern, you're probably converged.

2) Switch the simulation to transient, and plot the same monitor points as for the steady-state run until they reach a pseudo-steady-state behavior. I prefer this method, since I can show that my transient residuals are 10^-5 or lower, and that a stable solution has been reached through the monitor point values.

Hope this helps.

Jeff

 John August 28, 2005 01:15

Re: Convergence tips