Multiphase flow through a nozzle: turbulence model

leonardoasfora · May 30, 2016, 14:02

Hi,

Im modeling a homogeneous two phase flow through a convergent nozzle (liquid water as continuous phase an air as discrete phase). This study is part of my master degree. I just recently started using CFX (two months now) and i didn't have time to learn all its peculiarities and necessary knowledge (to be honest, i just have three more months).

My question is about which turbulence model should i use in order to model this kind of flow. I read that it's only necessary to use a low reynold number models (like SST) when there's, for instance, adverse pressure gradient or flow separation. Is this case i would need inflation layers with y+~1. In a convergent nozzle that's not the case, and i was wondering if k-e model with scalable wall function would be fine, as long as i guarantee 30 < y+ < 200 (is that right)?

Sorry if that's a stupid question, but as I said, i didn't have much time to learn everything. Sorry for my bad english

Thanks in advance.

ghorrocks · May 30, 2016, 22:21

My advice is always "Use SST unless you have a very good reason not to." You have not stated a clear reason why not to use it - so use SST.

Some other comments:

Quote:

I read that it's only necessary to use a low reynold number models (like SST) when there's, for instance, adverse pressure gradient or flow separation.

Low Re models do not necessarily handle adverse pressure gradients or separations better.

Quote:

Is this case i would need inflation layers with y+~1. In a convergent nozzle that's not the case, and i was wondering if k-e model with scalable wall function would be fine, as long as i guarantee 30 < y+ < 200 (is that right)?

This is a very common misconception with SST - you do not need y+=1 for SST. With automatic wall functions it works fine using wall functions and can handle y+>30 and maybe up to 200 (or further).

So my recommendation is:
* Use SST
* Use wall functions, so y+ in the range of 30-200
* Do a mesh sensitivity study to see if that is OK. If it is then you are fine. Only go to y+=1 if you fail to get mesh insensitivity.

leonardoasfora · May 31, 2016, 10:42

Thank you very much

I didn't want to use SST because it took more time to converge, but it isn't a big deal, so i'll use it.

Opaque · May 31, 2016, 11:43

I am a bit surprised by your observations.

It is known that k-omega based models converge better than k-epsilon based models. The stiffness of the epsilon equation, and its behavior near the wall makes it nearly impossible to solve with extremely fine meshes. You can try that by solving the k-epsilon for a fully developed velocity profile in a pipe. It is a 1-D equation, and the solution is hard to converge as the Reynolds number is increased.

As Glenn suggested, stick to SST as long as it works for your case.

Just my 2 cents..

May 30, 2016, 14:02	Multiphase flow through a nozzle: turbulence model	#1
leonardoasfora New Member Leonardo Asfora de Oliveira Join Date: May 2016 Posts: 9 Rep Power: 10	Hi, Im modeling a homogeneous two phase flow through a convergent nozzle (liquid water as continuous phase an air as discrete phase). This study is part of my master degree. I just recently started using CFX (two months now) and i didn't have time to learn all its peculiarities and necessary knowledge (to be honest, i just have three more months). My question is about which turbulence model should i use in order to model this kind of flow. I read that it's only necessary to use a low reynold number models (like SST) when there's, for instance, adverse pressure gradient or flow separation. Is this case i would need inflation layers with y+~1. In a convergent nozzle that's not the case, and i was wondering if k-e model with scalable wall function would be fine, as long as i guarantee 30 < y+ < 200 (is that right)? Sorry if that's a stupid question, but as I said, i didn't have much time to learn everything. Sorry for my bad english Thanks in advance.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
mass flow in is not equal to mass flow out	saii	CFX	12	March 19, 2018 06:21
Multiphase Flow with NIST Real Gas Model	Annka	FLUENT	7	May 23, 2017 08:58
Flow over a blunt body - SST turbulence model	bharath	CFX	5	January 29, 2016 11:16
compressible flow in a counterflow nozzle	d.vamsidhar	FLUENT	0	November 24, 2005 02:45

May 31, 2016, 10:42		#3
leonardoasfora New Member Leonardo Asfora de Oliveira Join Date: May 2016 Posts: 9 Rep Power: 10	Thank you very much I didn't want to use SST because it took more time to converge, but it isn't a big deal, so i'll use it.

May 31, 2016, 11:43		#4
Opaque Senior Member Join Date: Jun 2009 Posts: 1,880 Rep Power: 33	I am a bit surprised by your observations. It is known that k-omega based models converge better than k-epsilon based models. The stiffness of the epsilon equation, and its behavior near the wall makes it nearly impossible to solve with extremely fine meshes. You can try that by solving the k-epsilon for a fully developed velocity profile in a pipe. It is a 1-D equation, and the solution is hard to converge as the Reynolds number is increased. As Glenn suggested, stick to SST as long as it works for your case. Just my 2 cents..