Turbulent viscosity ratio

lost.identity · May 7, 2009, 18:25

Hi,

For a flow through a pipe what are the criteria in choosing the turbulent viscosity ratio, if one were to use the Spallart-Allmares turbulent model?

Thanks.

vishyaroon · May 7, 2009, 23:38

Does SA model work well for internal flows? Just curious.

lost.identity · May 11, 2009, 06:58

Not sure mate. I'm new to CFD. So far I can't say whether SA or k-epsilon works better. The problem is with k-epsilon (for the code I'm using) I need to have be careful of the first grid node of the wall (when using wall functions), with SA, apparently, this is not an issue.

harishg · May 11, 2009, 11:57

It is also possible for you to convert your K-e model into low Re K-e model and you would be able to integrate your model through the viscous layer. Have a look at the book on Wilcox on how SA and K-e model performs for pipe flow.

lost.identity · May 11, 2009, 12:28

I think I tried that. You mean not to use wall functions with k-epsilon so that the first grid node has to be at a y+ of 100?

harishg · May 11, 2009, 12:33

There are two versions of the k-e model that you can employ. in the first version you place your first grid point in the logarithmic layer (30 < y+<100) and use the so called high Re model. In the second version, you modify the turbulent viscosity with a damping term and integrate through the viscous wall layer. In this method the first grid node has to be below y+ <5.

If you use k-omega model you can integrate through the viscous layer without damping and sometimes obtain acceptable results.

lost.identity · May 12, 2009, 06:38

cheers for that useful info.

i've decided to use SA model. I've been told this model will work whether you use wall functions or not.

btw, what's the closest you can have your first grid node from the wall? someone on this forum told me that for heat transfer problems the first grid node has to be within a y+ of much smaller than 1.

I did a grid independence study and found that when I had the first grid node at y+=1 and changed the vertical grid nodes it kept changing the solution (by solution I mean I measure the temperature, which changes slightly as I change the grid nodes). And when I had it at a very low y+ the change in solution was much small.

But I've been told by CFD code developer that it's stupid t have such a small y+.

harishg · May 12, 2009, 12:24

Using yplus less than 1 ensures grid independent solution. Look for paper by Davidson and Peng - 1995 - Intl. Jou. of Num. Methods in Fluids. They discuss the grid refinement in the context of k-omega model but the same applies to your case too.

lost.identity · May 13, 2009, 05:55

Sorry but I couldn't find that paper you mentioned. It's not even listed on Davidson's webpage. If it's possible could you let me know of the title please?

May 7, 2009, 18:25	Turbulent viscosity ratio	#1
lost.identity Senior Member Join Date: Apr 2009 Posts: 118 Rep Power: 17	Hi, For a flow through a pipe what are the criteria in choosing the turbulent viscosity ratio, if one were to use the Spallart-Allmares turbulent model? Thanks.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Turbulent viscosity ratio	Balaji	FLUENT	5	May 27, 2014 03:47
turbulent viscosity ratio	James Willie	FLUENT	1	June 21, 2005 03:35
On limiting to turbulent viscosity ratio!	varghese	FLUENT	2	November 15, 2003 08:56
Turbulent Viscosity Ratio	xiang	FLUENT	3	May 20, 2003 12:46
Problem of Turbulent Viscosity Ratio Limited	David Yang	FLUENT	3	June 3, 2002 06:13

May 7, 2009, 23:38		#2
vishyaroon Senior Member Aroon Join Date: Apr 2009 Location: Racine WI Posts: 148 Rep Power: 17	Does SA model work well for internal flows? Just curious.

May 11, 2009, 06:58		#3
lost.identity Senior Member Join Date: Apr 2009 Posts: 118 Rep Power: 17	Not sure mate. I'm new to CFD. So far I can't say whether SA or k-epsilon works better. The problem is with k-epsilon (for the code I'm using) I need to have be careful of the first grid node of the wall (when using wall functions), with SA, apparently, this is not an issue.

May 11, 2009, 11:57		#4
harishg Senior Member N/A Join Date: Mar 2009 Posts: 189 Rep Power: 17	It is also possible for you to convert your K-e model into low Re K-e model and you would be able to integrate your model through the viscous layer. Have a look at the book on Wilcox on how SA and K-e model performs for pipe flow.

May 11, 2009, 12:28		#5
lost.identity Senior Member Join Date: Apr 2009 Posts: 118 Rep Power: 17	I think I tried that. You mean not to use wall functions with k-epsilon so that the first grid node has to be at a y+ of 100?

May 11, 2009, 12:33		#6
harishg Senior Member N/A Join Date: Mar 2009 Posts: 189 Rep Power: 17	There are two versions of the k-e model that you can employ. in the first version you place your first grid point in the logarithmic layer (30 < y+<100) and use the so called high Re model. In the second version, you modify the turbulent viscosity with a damping term and integrate through the viscous wall layer. In this method the first grid node has to be below y+ <5. If you use k-omega model you can integrate through the viscous layer without damping and sometimes obtain acceptable results.

May 12, 2009, 06:38		#7
lost.identity Senior Member Join Date: Apr 2009 Posts: 118 Rep Power: 17	cheers for that useful info. i've decided to use SA model. I've been told this model will work whether you use wall functions or not. btw, what's the closest you can have your first grid node from the wall? someone on this forum told me that for heat transfer problems the first grid node has to be within a y+ of much smaller than 1. I did a grid independence study and found that when I had the first grid node at y+=1 and changed the vertical grid nodes it kept changing the solution (by solution I mean I measure the temperature, which changes slightly as I change the grid nodes). And when I had it at a very low y+ the change in solution was much small. But I've been told by CFD code developer that it's stupid t have such a small y+.

May 12, 2009, 12:24		#8
harishg Senior Member N/A Join Date: Mar 2009 Posts: 189 Rep Power: 17	Using yplus less than 1 ensures grid independent solution. Look for paper by Davidson and Peng - 1995 - Intl. Jou. of Num. Methods in Fluids. They discuss the grid refinement in the context of k-omega model but the same applies to your case too.

May 13, 2009, 05:55		#9
lost.identity Senior Member Join Date: Apr 2009 Posts: 118 Rep Power: 17	Sorry but I couldn't find that paper you mentioned. It's not even listed on Davidson's webpage. If it's possible could you let me know of the title please?