Simple problems with ansys CFX

Luk_Fiz · October 22, 2009, 14:46

Hi all,
Since I wanted to dig a bit around basic ansys capabilities I decided to model a very simple flow through the pipe. I developed three models.
A) A pipe of 10mm diameter and length is 100x diameter so flow is fully developed. Re~50000. I was trying ke and sst models having results perfectly fitting darcy equation for pressure drop (~20800 Pa). Everyhing is ok, no problem.
B) The same pipe but Re~6000. Different meshes (from y+=0.2 up to y+=11 which is recommended for ke). Darcy equation shows around 512Pa of pressure drop but this time from CFX I got about 470Pa from ke model and various results around 530 or 540 Pa for sst. I was trying different bc's trying to establish also a fully developed flow - no change, ke underestimates.
C) Pipe shrinked to 10cm long (only 10 times a diameter). Re~6000. From darcy equation for fully developed flow it should be ten times lower than before - around 51Pa. And now suprise. Ke shows precisely 51 Pa and sst and komega is around 64Pa and even 80Pa (depending on options for intermittency).

I was really suprised by results. Can anyone tell something about how to handle it. I want to say that these solutions are perfectly converged and mesh density as also timestep undependant. How can I analyze it furtherly? This is very simple situation and in my opinion these models should shows a constant problems (going to laminar flow with low Re), and here is something strange with sst and even more strange with ke (gives wrong results for low Re long pipe and good for short one).

What turbulence model is the best in such conditions. Up to now I was thinking about sst as good for externall and internal flows in wide range of Re's and ke as good for internal ones especially for hi Re. Now I am confused.

Luk

ghorrocks · October 22, 2009, 18:26

Re=6000 is getting a little low for k-e. It does not surprise me that it starts to loose some accuracy there.

Are you sure you have a fully developed flow at the entry and exit? I suspect that is the cause of the discrepancy with the SST model. Also note you should not need a turbulence transition model as the flow should not have transition in it but be fully turbulent.

Luk_Fiz · October 23, 2009, 16:24

I was trying many options for sst and noticed of course some differences but without achieving success.
I am also interested if the ke results for short and long pipe differs (in terms of beeing in agreement with theory). In my opinion for long pipe ke should give better results, because of fully developed turbulence, however as You said Re=6000 is a bit low and it can misdirect the results. From the other side, in shorter pipe results are very good for ke - but I think it can be coincidence. I have to make a pipe shorter or longer a bit.

But what with sst? Is it normal to have not that good results with such a simple cases? I want to emphasize that I tried many options with mesh, different bc and so on.

Luk

ghorrocks · October 24, 2009, 06:52

I repeat my last comment - are you sure your simulation is fully developed flow along the entire length?

Luk_Fiz · October 24, 2009, 14:23

Yes, I am.

Luk

ghorrocks · October 25, 2009, 06:12

Are you using the turbulence transition model?

There is no inherent problem I know of in the SST model for what you are trying to do, however the accuracy of SST may be a bit reduced in the lower Re flow you are using. So I would check all the standard things - in this case that is mesh and convergence.

Luk_Fiz · October 25, 2009, 17:33

All is perfectly converged - mesh, timestep... I will try some very fine mesh, however it would be very dense, and if it changes something that it would be not usable anyway in any serious, complicated application (to many elements).

Luk

ghorrocks · October 25, 2009, 17:41

Are you using the turbulence transition model?

Luk_Fiz · October 26, 2009, 03:56

Yes, I was trying gamma-theta and others.

Luk

ghorrocks · October 26, 2009, 04:48

If you are looking at fully developed flow, why are you trying a turbulence transition model? You should turn it off and just use a normal turbulence model instead.

Luk_Fiz · October 26, 2009, 07:40

I just wan to say, that I was trying everything:
- ke, komega, sst,
- defferent meshes,
- different transition models related to ssd (and without transition model of course)
- developed and non-developed flows.

The results vary from case to case, but I was unable to obtain the proper results.

Luk

ghorrocks · October 26, 2009, 18:17

You may well then have hit a limitation of turbulence modelling in general then. It does not surprise me as all the main turbulence models are tuned for high Re flow.

The CFX documentation discusses a low Re variation of the Eddy Viscosity Transport model. Have you tried that one?

October 22, 2009, 14:46	Simple problems with ansys CFX	#1
Luk_Fiz Member Lukasz Join Date: Mar 2009 Posts: 67 Rep Power: 17	Hi all, Since I wanted to dig a bit around basic ansys capabilities I decided to model a very simple flow through the pipe. I developed three models. A) A pipe of 10mm diameter and length is 100x diameter so flow is fully developed. Re~50000. I was trying ke and sst models having results perfectly fitting darcy equation for pressure drop (~20800 Pa). Everyhing is ok, no problem. B) The same pipe but Re~6000. Different meshes (from y+=0.2 up to y+=11 which is recommended for ke). Darcy equation shows around 512Pa of pressure drop but this time from CFX I got about 470Pa from ke model and various results around 530 or 540 Pa for sst. I was trying different bc's trying to establish also a fully developed flow - no change, ke underestimates. C) Pipe shrinked to 10cm long (only 10 times a diameter). Re~6000. From darcy equation for fully developed flow it should be ten times lower than before - around 51Pa. And now suprise. Ke shows precisely 51 Pa and sst and komega is around 64Pa and even 80Pa (depending on options for intermittency). I was really suprised by results. Can anyone tell something about how to handle it. I want to say that these solutions are perfectly converged and mesh density as also timestep undependant. How can I analyze it furtherly? This is very simple situation and in my opinion these models should shows a constant problems (going to laminar flow with low Re), and here is something strange with sst and even more strange with ke (gives wrong results for low Re long pipe and good for short one). What turbulence model is the best in such conditions. Up to now I was thinking about sst as good for externall and internal flows in wide range of Re's and ke as good for internal ones especially for hi Re. Now I am confused. Luk

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
export comsol mesh into ansys or cfx	amazdeh	ANSYS Meshing & Geometry	1	June 9, 2017 10:18
Properly using symmetry with both CFX and ANSYS workbench for a FSI analysis, help!	Cirion0000	CFX	0	July 6, 2009 15:26
Water pump OpenFOAM 15 ANSYS CFX 110 comparation	waynezw0618	OpenFOAM Running, Solving & CFD	39	March 5, 2009 13:57
Ansys CFX Vs. NACA Summary Report. Mismatch Result	Santiago Orrego.	CFX	3	February 5, 2007 13:05
Temperature transferring from CFX to ANSYS?	Se-Hee	CFX	0	November 28, 2006 06:56

October 22, 2009, 18:26		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	Re=6000 is getting a little low for k-e. It does not surprise me that it starts to loose some accuracy there. Are you sure you have a fully developed flow at the entry and exit? I suspect that is the cause of the discrepancy with the SST model. Also note you should not need a turbulence transition model as the flow should not have transition in it but be fully turbulent.

October 23, 2009, 16:24		#3
Luk_Fiz Member Lukasz Join Date: Mar 2009 Posts: 67 Rep Power: 17	I was trying many options for sst and noticed of course some differences but without achieving success. I am also interested if the ke results for short and long pipe differs (in terms of beeing in agreement with theory). In my opinion for long pipe ke should give better results, because of fully developed turbulence, however as You said Re=6000 is a bit low and it can misdirect the results. From the other side, in shorter pipe results are very good for ke - but I think it can be coincidence. I have to make a pipe shorter or longer a bit. But what with sst? Is it normal to have not that good results with such a simple cases? I want to emphasize that I tried many options with mesh, different bc and so on. Luk

October 24, 2009, 06:52		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	I repeat my last comment - are you sure your simulation is fully developed flow along the entire length?

October 24, 2009, 14:23		#5
Luk_Fiz Member Lukasz Join Date: Mar 2009 Posts: 67 Rep Power: 17	Yes, I am. Luk

October 25, 2009, 06:12		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	Are you using the turbulence transition model? There is no inherent problem I know of in the SST model for what you are trying to do, however the accuracy of SST may be a bit reduced in the lower Re flow you are using. So I would check all the standard things - in this case that is mesh and convergence.

October 25, 2009, 17:33		#7
Luk_Fiz Member Lukasz Join Date: Mar 2009 Posts: 67 Rep Power: 17	All is perfectly converged - mesh, timestep... I will try some very fine mesh, however it would be very dense, and if it changes something that it would be not usable anyway in any serious, complicated application (to many elements). Luk

October 25, 2009, 17:41		#8
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	Are you using the turbulence transition model?

October 26, 2009, 03:56		#9
Luk_Fiz Member Lukasz Join Date: Mar 2009 Posts: 67 Rep Power: 17	Yes, I was trying gamma-theta and others. Luk

October 26, 2009, 04:48		#10
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	If you are looking at fully developed flow, why are you trying a turbulence transition model? You should turn it off and just use a normal turbulence model instead.

October 26, 2009, 07:40		#11
Luk_Fiz Member Lukasz Join Date: Mar 2009 Posts: 67 Rep Power: 17	I just wan to say, that I was trying everything: - ke, komega, sst, - defferent meshes, - different transition models related to ssd (and without transition model of course) - developed and non-developed flows. The results vary from case to case, but I was unable to obtain the proper results. Luk

October 26, 2009, 18:17		#12
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	You may well then have hit a limitation of turbulence modelling in general then. It does not surprise me as all the main turbulence models are tuned for high Re flow. The CFX documentation discusses a low Re variation of the Eddy Viscosity Transport model. Have you tried that one?