Urans solution does not "converge"

luca1989 · January 8, 2015, 04:31

Hi everyone. I'm trying to simulate a 3D sphere in the unsteady case. The steady case converged fine, the field and the coefficients seems to be close enought to the real. Using the same boundary and initial conditions and changing only the fvchemes, fvsolution and controldict (witch i took and manipulate from the motorbike pisofoam tutorial) the solution seem to converge at the beginning but the pressure is oscillating strongly and after 50 iterations the velocities start to diverge and then start oscillating. I tried both starting from initial field of p=0 and U=(1 0 0) and from the resolved stationary case.

Any ideas? Thank you

RodriguezFatz · January 8, 2015, 05:48

1) What means "the velocities start to diverge and then start oscillating"?
2) Can you post some log output of the "diverging" time steps?

luca1989 · January 8, 2015, 06:47

Hi thank you for your quick reply. I should have specified that the residuals start to diverge. I will link my gnuplot output so you can see what I mean.

I also tried to use the setup from this page
http://www.cfd-online.com/Forums/ope...omega-sst.html
as his case is quite similar to mine (obviuolsy I changed the empty boundary with mine) the pressure acts in the same way velocities does not diverge but the dt decreases fastly to very small values (1e-6)

RodriguezFatz · January 8, 2015, 07:11

Can you post some log output of the "diverging" time steps?

luca1989 · January 8, 2015, 07:19

The file is too large for this site so I give you the dropbox link

https://www.dropbox.com/s/q77r6gne8j2lqxt/log.txt?dl=0

RodriguezFatz · January 8, 2015, 07:45

1) I would use a fixed time step of 1e-4 with maybe 4 pressure corrections in piso.
2) Did you set laplacian to "Gauss linear corrected"? It looks like that... Try "Gauss linear uncorrected".
3) I would start with "backward" time differencing.

RodriguezFatz · January 8, 2015, 08:01

If you try it... post the log output after some iterations...

luca1989 · January 8, 2015, 08:02

I tried many different solution that I took from differnt tutorial. I used also backward scheme for time but nothing changed. I also tried the set up from thies page as the case is similar to mine (exept from it is 2D ad I changed the empty BC)

http://www.cfd-online.com/Forums/ope...omega-sst.html

I will post the fvchemes and fvsolution.

May it be the mesh? I've done it with ansa and it has some non orthogonal faces but running the checkmesh is OK

RodriguezFatz · January 8, 2015, 08:09

Ok, additionally you should change the divScheme to "default Gauss upwind phi;" and delete all of them except the "nuEff*..."

RodriguezFatz · January 8, 2015, 09:41

Now, what do you get with these modifications?

January 8, 2015, 04:31	Urans solution does not "converge"	#1
luca1989 New Member luca Join Date: Dec 2014 Posts: 8 Rep Power: 11	Hi everyone. I'm trying to simulate a 3D sphere in the unsteady case. The steady case converged fine, the field and the coefficients seems to be close enought to the real. Using the same boundary and initial conditions and changing only the fvchemes, fvsolution and controldict (witch i took and manipulate from the motorbike pisofoam tutorial) the solution seem to converge at the beginning but the pressure is oscillating strongly and after 50 iterations the velocities start to diverge and then start oscillating. I tried both starting from initial field of p=0 and U=(1 0 0) and from the resolved stationary case. Any ideas? Thank you

January 8, 2015, 05:48		#2
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 27	1) What means "the velocities start to diverge and then start oscillating"? 2) Can you post some log output of the "diverging" time steps? __________________ The skeleton ran out of shampoo in the shower.

January 8, 2015, 07:11		#4
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 27	Can you post some log output of the "diverging" time steps? __________________ The skeleton ran out of shampoo in the shower.

January 8, 2015, 07:45		#6
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 27	1) I would use a fixed time step of 1e-4 with maybe 4 pressure corrections in piso. 2) Did you set laplacian to "Gauss linear corrected"? It looks like that... Try "Gauss linear uncorrected". 3) I would start with "backward" time differencing. __________________ The skeleton ran out of shampoo in the shower.

January 8, 2015, 08:01		#7
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 27	If you try it... post the log output after some iterations... __________________ The skeleton ran out of shampoo in the shower.

January 8, 2015, 07:19		#5
luca1989 New Member luca Join Date: Dec 2014 Posts: 8 Rep Power: 11	The file is too large for this site so I give you the dropbox link https://www.dropbox.com/s/q77r6gne8j2lqxt/log.txt?dl=0

January 8, 2015, 08:09		#9
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 27	Ok, additionally you should change the divScheme to "default Gauss upwind phi;" and delete all of them except the "nuEff..." __________________ The skeleton ran out of shampoo in the shower.*

January 8, 2015, 09:41		#10
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 27	Now, what do you get with these modifications? __________________ The skeleton ran out of shampoo in the shower.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Asymmetrical solution using RANS	Marli	FLUENT	2	January 9, 2015 10:26
grid dependancy	gueynard a.	Main CFD Forum	19	June 27, 2014 22:22
Analytic solution for 2D steady Euler equations	jojo81	Main CFD Forum	0	October 15, 2012 13:05
IcoFoam parallel woes	msrinath80	OpenFOAM Running, Solving & CFD	9	July 22, 2007 03:58
Wall functions	Abhijit Tilak	Main CFD Forum	6	February 5, 1999 02:16