CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > CFX

Res. normalisation and convergence

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   September 25, 2007, 17:34
Default Res. normalisation and convergence
  #1
Felix
Guest
 
Posts: n/a
Hello,

I observed a strange convergence behavior that I can't explain, maybe someone can help with this. One specific run is not converging, RMS U-mom residuals stalling aroung 1e-4. However, in POST, the regions of residuals above 1e-5 are really small (I'd say less than 1% of the total volume ). How can the RMS residuals be of 1e-4 if almost no element is above 1e-5 ?? Could the renormalisation procedure of residuals give too much weight to some elements ?

The MAX (U-mom) residuals are around 3.3e-2 on 4 elements [2.1M total]. In their vincinity the U speed is very slow and the A.R. are large since it is close to a wall.

Thanks for your help,

Felix
  Reply With Quote

Old   September 25, 2007, 23:23
Default Re: Res. normalisation and convergence
  #2
Glenn Horrocks
Guest
 
Posts: n/a
http://www.cfd-online.com/Wiki/Ansys...gence_criteria

Glenn Horrocks
  Reply With Quote

Old   September 26, 2007, 10:24
Default Re: Res. normalisation and convergence
  #3
Felix
Guest
 
Posts: n/a
Hi Glenn,

Thanks for the link it's always good to read that for a second, third time ;-). The run I was talking about actually converged after I improved the mesh quality in the high residuals regions. Therefore my question is a more general one.

What I don't get is how can a few elements block convergence in that way ? The purpose of using a criteria on RMS residuals is to have general convergence and not to get stuck with that kind of situation where a few elements don't converge as well as the others, no ? Then I can't figure out why is the convergence stalling around 1e-4 if almost all the domain is below 1e-5.

Thanks for your input,

Felix
  Reply With Quote

Old   September 26, 2007, 10:28
Default Re: Res. normalisation and convergence
  #4
BB
Guest
 
Posts: n/a
I have the similar puzzle too.
  Reply With Quote

Old   September 26, 2007, 18:18
Default Re: Res. normalisation and convergence
  #5
Glenn Horrocks
Guest
 
Posts: n/a
Hi,

As discussed in the FAQ, a small region of high residuals may or may not be an issue for your simulation. If the region is far away from the area of interest it probably has no effect on things and is not causing significant error. Alternately, if the region is right in the middle of the key area then it could be causing problems - for instance if it occurs at a separation point (which high residuals commonly do) it could mean the separation point is not accurately placed and that can have a significant effect on results.

Glenn Horrocks
  Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On



All times are GMT -4. The time now is 20:16.