Convergence problems turbo machinary

baddy · October 17, 2012, 03:26

hi,

i have a modell devided by 4 (90°) my MAX residuals are never under e-02.
coupled rotating domain with stationary domain. air as ideal gas / total energy.
physical time step 0.00001.
interface between domains frozen rotor.

can some one help me to get an convergence?

thanx

mvoss · October 17, 2012, 05:08

export the residuals (CFX-Pre) within the normal result files->solve again starting from your current solution and run just 1 Iter.->visualize the "highest" residuals in CFX-Post as a Isovolume --> i bet you´ll find some mesh issue OR some trans. effect

Is there a part. reason for chossing such a small physical TS?

neewbie

baddy · October 17, 2012, 05:16

Yes if not such a ts, error comes 'overflow'

mvoss · October 17, 2012, 05:27

tried 'local' timestepping?

baddy · October 17, 2012, 05:29

thanx

for quick answers

no physical ts

energy382 · October 17, 2012, 07:46

what kind of turbomachinery? turbine?

you'vve to provide more information to achieve serious advice.

as it is a stationary simulation, it's not a "real" timestep, but rather a pseudo.

if your mesh has good quality, stat/rot interface is well placed and there were no setup failure (e.g. neglect wall as counter-rotating etc.), try to use a 1st order scheme (also turbulence) or try to use a coarse mesh to achieve convergence.

is there also a diffuser anywhere?!

baddy · October 17, 2012, 07:59

yes it is a stationary simulation. no turbine. it is a asynchrone machine. i am solving the inner convetive cooling... air produces also heat. my mesh has a good quality but could be better. the geo is very difficult. the inter face i ok ( frozen rotor) no countering wall exist.

eng_norouzi · October 17, 2012, 11:33

can you send an image of your residuals?
are they like a smooth line or not?
are they changing?
and also you should check your results if they are logical residual is not a big deal

baddy · October 18, 2012, 06:16

Which wall function should be taken? CFX makes it automatically??
what about les ?

energy382 · October 18, 2012, 06:54

Quote:

Originally Posted by baddy

Which wall function should be taken? CFX makes it automatically??
what about les ?

as I mentioned before, it's hard to give you serious advice without any pics.

One for sure: you need a very good mesh for this kind of simulation. y+ should be <1!
that means, that you can't use a wall function. use sst as turbulence model and ensure, that your near-wall boundary layer is resolved accurate. regarding heat transfer issues, you have to be meticulous about that!

y+ values currently?

please provide some pics

ghorrocks · October 18, 2012, 18:18

Sorry to jump in late but the original question here is an FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

baddy · October 19, 2012, 02:55

ghorrocks i read this time ago. i did not ask before reding this. so i am refinig the mesh.

energy382 · October 22, 2012, 06:40

As I mentioned before (after looking at your residuals), you have problems with energy equation. It's highly proable that you didn't resolve boundary layer accurate and/or use heat model, that doesn't apply to real physics.

But it's difficult to give serious advice without further details.

What kind of mesh do you use? what y+ values do you have? which setup do you use?

Quote:

Originally Posted by baddy

ghorrocks i read this time ago. i did not ask before reding this. so i am refinig the mesh.

baddy · October 23, 2012, 06:03

my mesh is patially stuctured, partially unstructured. but some geom is nor possible to mesh it with icem (to much time). i do not unterstand the setup? what do you mean with this?

my physical ts is 1*e-6 now is this ok or to small?

October 17, 2012, 03:26	Convergence problems turbo machinary	#1
baddy New Member baddy Join Date: Mar 2012 Posts: 8 Rep Power: 14	hi, i have a modell devided by 4 (90°) my MAX residuals are never under e-02. coupled rotating domain with stationary domain. air as ideal gas / total energy. physical time step 0.00001. interface between domains frozen rotor. can some one help me to get an convergence? thanx

October 17, 2012, 07:59		#7
baddy New Member baddy Join Date: Mar 2012 Posts: 8 Rep Power: 14	yes it is a stationary simulation. no turbine. it is a asynchrone machine. i am solving the inner convetive cooling... air produces also heat. my mesh has a good quality but could be better. the geo is very difficult. the inter face i ok ( frozen rotor) no countering wall exist. Last edited by baddy; October 17, 2012 at 10:13.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Fluent Convergence Problems	Vidya	FLUENT	24	January 1, 2016 14:01
Convergence problems!!!	king lui	FLUENT	11	June 15, 2011 11:30
NACA0012 Convergence Problems	StudentAndrew	CFX	6	November 21, 2005 07:49
convergence problems	jeremy	FLUENT	7	May 30, 2002 07:41
convergence problems	david aquilina	Siemens	4	October 27, 2000 10:18

October 17, 2012, 05:08		#2
mvoss Senior Member Matthias Voß Join Date: Mar 2009 Location: Berlin, Germany Posts: 449 Rep Power: 20	export the residuals (CFX-Pre) within the normal result files->solve again starting from your current solution and run just 1 Iter.->visualize the "highest" residuals in CFX-Post as a Isovolume --> i bet you´ll find some mesh issue OR some trans. effect Is there a part. reason for chossing such a small physical TS? neewbie

October 17, 2012, 05:16		#3
baddy New Member baddy Join Date: Mar 2012 Posts: 8 Rep Power: 14	Yes if not such a ts, error comes 'overflow'

October 17, 2012, 05:27		#4
mvoss Senior Member Matthias Voß Join Date: Mar 2009 Location: Berlin, Germany Posts: 449 Rep Power: 20	tried 'local' timestepping?

October 17, 2012, 05:29		#5
baddy New Member baddy Join Date: Mar 2012 Posts: 8 Rep Power: 14	thanx for quick answers no physical ts

October 17, 2012, 07:46		#6
energy382 Senior Member Christoph Join Date: May 2011 Location: Germany Posts: 182 Rep Power: 18	what kind of turbomachinery? turbine? you'vve to provide more information to achieve serious advice. as it is a stationary simulation, it's not a "real" timestep, but rather a pseudo. if your mesh has good quality, stat/rot interface is well placed and there were no setup failure (e.g. neglect wall as counter-rotating etc.), try to use a 1st order scheme (also turbulence) or try to use a coarse mesh to achieve convergence. is there also a diffuser anywhere?!

October 17, 2012, 11:33		#8
eng_norouzi Member ali Join Date: Feb 2011 Posts: 37 Rep Power: 15	can you send an image of your residuals? are they like a smooth line or not? are they changing? and also you should check your results if they are logical residual is not a big deal

October 18, 2012, 06:16		#9
baddy New Member baddy Join Date: Mar 2012 Posts: 8 Rep Power: 14	Which wall function should be taken? CFX makes it automatically?? what about les ?

October 18, 2012, 18:18		#11
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,872 Rep Power: 144	Sorry to jump in late but the original question here is an FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

October 19, 2012, 02:55		#12
baddy New Member baddy Join Date: Mar 2012 Posts: 8 Rep Power: 14	ghorrocks i read this time ago. i did not ask before reding this. so i am refinig the mesh.

October 23, 2012, 06:03		#14
baddy New Member baddy Join Date: Mar 2012 Posts: 8 Rep Power: 14	my mesh is patially stuctured, partially unstructured. but some geom is nor possible to mesh it with icem (to much time). i do not unterstand the setup? what do you mean with this? my physical ts is 1*e-6 now is this ok or to small?