Y+ value for SST turbulence model

beyonder1 · September 17, 2015, 15:49

Hi,
I have cascade of airfoils with inlet speed=20m/s and chord length=15.2. The airfoils are oscillating few degrees and hence there is mesh deformation. The reynolds no. is around 0.2 million. I am running the simulation on cfx. I did some research, read wiki pages and concluded that SST model could be best choice. But i have some doubt regarding choosing Y+ value as SST asks for low value. I dont know if my mesh is fine enough.
I also know about mesh independence study as a way of finding out but i would like some expert suggestion about choosing turbulence model and Y+ importance for my case(explained in first line).

ghorrocks · September 17, 2015, 18:39

Generate a mesh which you think is close and run it. Then generate the mesh again with all mesh length settings halved. This should generate a mesh with approximately 8 times as many nodes. Run it and compare it to the first one. If the two results are similar enough that you are happy with it then you can use the coarser mesh. If not then you need to halve it again until you do get an accuracy you are happy with.

You can also do this for the boundary layer mesh only to work out what y+ you need.

There are also more sophisticated ways of doing this using Richardson extrapolation.

beyonder1 · September 18, 2015, 03:55

I understand what you are saying and i am currently doing it. I want to know if there is some restriction on Y+ value(less or greater than some value) for SST to work efficiently. Also does SST an appropriate choice or do i have to vary the turbulence model too?

ghorrocks · September 18, 2015, 05:46

Yes, SST is an appropriate choice.

SST is theoretically valid for all y+ so do not restrict yourself to a y+ value. But numerical reality means that if y+ is too big you will not have adequate resolution of the boundary layer and if y+ is too small you will not converge due to numerical round off errors.

urosgrivc · January 6, 2016, 07:28

As i understand, SST switches to DNS near wals if y+ is lower than ~11. if it is from 0 to 300 it should be sufficient, but with y+ values above 100 I dont suspect geting any relevant results from the simulation, I usualy keep it mostly under 5 (train brake rotor cooling). And if you are simulating aerofoils where there are high gradients of pressure and velocity, good near wall resolving is wery important, I stay belov 1 with y+ when near wall results are important.
As you probably need forces for cd as a result, the smaller the better but under a surtan walue results wont change anymore and that is enough. Problem is that with such small elements node count gets quite big and than computational time begins to be a problem.

please correct me if I'm wrong

ghorrocks · January 7, 2016, 18:30

SST does not switch to DNS. It is a RANS turbulence model. It does however switch to integration to the wall.

Your other comments are generally correct, but I think you have missed the point a little bit. If wall functions adequately model the flow then there is no need to use y+<11 and integrate to the wall. So it is whether wall functions are adequate or not which determine whether you should integrate to the wall or use wall functions.

September 17, 2015, 15:49	Y+ value for SST turbulence model	#1
beyonder1 Member beyonder Join Date: May 2015 Posts: 67 Rep Power: 10	Hi, I have cascade of airfoils with inlet speed=20m/s and chord length=15.2. The airfoils are oscillating few degrees and hence there is mesh deformation. The reynolds no. is around 0.2 million. I am running the simulation on cfx. I did some research, read wiki pages and concluded that SST model could be best choice. But i have some doubt regarding choosing Y+ value as SST asks for low value. I dont know if my mesh is fine enough. I also know about mesh independence study as a way of finding out but i would like some expert suggestion about choosing turbulence model and Y+ importance for my case(explained in first line).

January 6, 2016, 07:28		#5
urosgrivc Senior Member urosgrivc Join Date: Dec 2015 Location: Slovenija Posts: 365 Rep Power: 11	As i understand, SST switches to DNS near wals if y+ is lower than ~11. if it is from 0 to 300 it should be sufficient, but with y+ values above 100 I dont suspect geting any relevant results from the simulation, I usualy keep it mostly under 5 (train brake rotor cooling). And if you are simulating aerofoils where there are high gradients of pressure and velocity, good near wall resolving is wery important, I stay belov 1 with y+ when near wall results are important. As you probably need forces for cd as a result, the smaller the better but under a surtan walue results wont change anymore and that is enough. Problem is that with such small elements node count gets quite big and than computational time begins to be a problem. please correct me if I'm wrong Last edited by urosgrivc; January 7, 2016 at 08:58.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Error in Two phase (condensation) modeling	adilsyyed	CFX	15	June 24, 2015 19:42
Overflow Error in Multiphase Modelling with Two Continuous Fluids	ashtonJ	CFX	6	August 11, 2014 14:32
An error has occurred in cfx5solve:	volo87	CFX	5	June 14, 2013 17:44
Wrong calculation of nut in the kOmegaSST turbulence model	FelixL	OpenFOAM Bugs	27	March 27, 2012 09:02
Low Reynolds k-epsilon model	YJZ	ANSYS	1	August 20, 2010 13:57

September 17, 2015, 18:39		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	Generate a mesh which you think is close and run it. Then generate the mesh again with all mesh length settings halved. This should generate a mesh with approximately 8 times as many nodes. Run it and compare it to the first one. If the two results are similar enough that you are happy with it then you can use the coarser mesh. If not then you need to halve it again until you do get an accuracy you are happy with. You can also do this for the boundary layer mesh only to work out what y+ you need. There are also more sophisticated ways of doing this using Richardson extrapolation.

September 18, 2015, 03:55		#3
beyonder1 Member beyonder Join Date: May 2015 Posts: 67 Rep Power: 10	I understand what you are saying and i am currently doing it. I want to know if there is some restriction on Y+ value(less or greater than some value) for SST to work efficiently. Also does SST an appropriate choice or do i have to vary the turbulence model too?

September 18, 2015, 05:46		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	Yes, SST is an appropriate choice. SST is theoretically valid for all y+ so do not restrict yourself to a y+ value. But numerical reality means that if y+ is too big you will not have adequate resolution of the boundary layer and if y+ is too small you will not converge due to numerical round off errors.

January 7, 2016, 18:30		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	SST does not switch to DNS. It is a RANS turbulence model. It does however switch to integration to the wall. Your other comments are generally correct, but I think you have missed the point a little bit. If wall functions adequately model the flow then there is no need to use y+<11 and integrate to the wall. So it is whether wall functions are adequate or not which determine whether you should integrate to the wall or use wall functions.