Computation can't achieve convergence

dushechka · June 17, 2014, 06:01

Hi everyone!
First of all, sorry for my language mistakes. English isn't my native language.

I simulate turbulent flow in centrifugal pump channel using CFX and standard k-e model. Meshing is done using icem cfd.

the graph of RMS start to rise after 500 iterations. Is it OK? I think that I have problems with mesh..but how can I check it?

And I don't like the precession. I've tried to use different value of physical timescale, but it didn't help.

Hope you will help me. If you need some information for this I will answer the questions about my simulation.

sans · June 17, 2014, 08:51

Have you checked your plots? may be you have a separation in your domain that's causing this. Plot some variables on interest and look at their variation vs. iterations.

dushechka · June 17, 2014, 10:18

Quote:

Originally Posted by sans

Have you checked your plots? may be you have a separation in your domain that's causing this. Plot some variables on interest and look at their variation vs. iterations.

Thank you for the answer

I have no separation in domain

singer1812 · June 17, 2014, 18:07

Have your solution export the residuals. Plot the residuals in Post. Using Isovolume you can plot those that are above your criteria (whatever you want it to be).

But, this could be a transient issue. Are you sure you arent getting separation or vortex in the domain? Residuals are looking like they have a cycle to them. Might try running this as a transient case for a bit to see if it steadies out.

Are you comparing this to experiment or analytical solution? Hows the Head value you are listing compare?

ghorrocks · June 17, 2014, 20:54

This is a FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

dushechka · June 18, 2014, 05:41

Quote:

Originally Posted by ghorrocks

This is a FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

Thanr you, ghorrocks. Exactly my situation. this link is very useful.

dushechka · June 18, 2014, 06:06

Quote:

Originally Posted by singer1812

Are you sure you arent getting separation or vortex in the domain? Residuals are looking like they have a cycle to them.

Now I'm not so sure.It's really look like I get vortex or smth. Should I check it in Post. right?

Quote:

Originally Posted by singer1812

Are you comparing this to experiment or analytical solution? Hows the Head value you are listing compare?

I don't have experiment data. But I have solution after coarse mesh and Head=100,68 m. And Head=100,231 m after this simulations

ghorrocks · June 18, 2014, 07:35

The FAQ explains how to handle this.

Try a larger time step. That may get the transient behaviour under control. If not then you may be forced to consider a transient simulation - but there are lots of other things to consider..... read the FAQ

There is also an FAQ on accuracy - have a read of that while you are at it.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Convergence	Centurion2011	FLUENT	48	June 15, 2022 00:29
Effect of convergence criteria for pressure loop on computation time	mhdali101	Main CFD Forum	0	November 29, 2013 17:09
FSI/MFX: resutls after using fluid mass flux to achieve convergence	gogohusky	ANSYS	0	May 21, 2012 06:40
Force can not converge	colopolo	CFX	13	October 4, 2011 23:03
Convergence of CFX field in FSI analysis	nasdak	CFX	2	June 29, 2009 02:17

June 17, 2014, 08:51		#2
sans Member SanS Join Date: Mar 2009 Posts: 41 Rep Power: 17	Have you checked your plots? may be you have a separation in your domain that's causing this. Plot some variables on interest and look at their variation vs. iterations.

June 17, 2014, 18:07		#4
singer1812 Senior Member Edmund Singer P.E. Join Date: Aug 2010 Location: Minneapolis, MN Posts: 511 Rep Power: 21	Have your solution export the residuals. Plot the residuals in Post. Using Isovolume you can plot those that are above your criteria (whatever you want it to be). But, this could be a transient issue. Are you sure you arent getting separation or vortex in the domain? Residuals are looking like they have a cycle to them. Might try running this as a transient case for a bit to see if it steadies out. Are you comparing this to experiment or analytical solution? Hows the Head value you are listing compare?

June 17, 2014, 20:54		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	This is a FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

June 18, 2014, 07:35		#8
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	The FAQ explains how to handle this. Try a larger time step. That may get the transient behaviour under control. If not then you may be forced to consider a transient simulation - but there are lots of other things to consider..... read the FAQ There is also an FAQ on accuracy - have a read of that while you are at it.