CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Diffusor testcase (http://www.cfd-online.com/Forums/main/3080-diffusor-testcase.html)

Marat Hoshim January 30, 2001 04:09

Diffusor testcase
 
Hello,

I have a question concerning the testcase 8.2 of the 8th ERCOFTAC workshop in Helsinki, 1999. It's a 2D asymmetric diffusor flow.

I'm not quite sure about the inlet condition. There's said the inlet condition is a fully developed channel flow with a Reynolds-Number of 20000 based on the centreline velocity and the channel height.

Is the centreline velocity the max. velocity of the channel velocity profile ?

If my channel has a height of 1m and the fluid is air am I right if the the centreline velocity is:

v=Re*vis/rho/height=20000*1.8e-5/1.16/1=0.3134 m/s.

In the testcase there was separation but in my calcs there's attached flow. That why I'm so doubtful towards my boundary conditions.

Thanks for your help, Marat


John C. Chien January 30, 2001 04:56

Re: Diffusor testcase
 
(1). I don't know what you do with your calculation, the formulation?, the turbulence model?, etc. (2). The accurate prediction of separated diffuser flow is very difficult, because the existing two-equation k-epsilon turbulence model performs poorly in adverse-pressure gradient conditions, which exists in diffuser flows.

Marat Hoshim January 30, 2001 05:06

Re: Diffusor testcase
 
Hi John,

thanks for your advise. But what's your opinion towards my two questions ?

Regards, Marat

sylvain January 30, 2001 10:32

Re: Diffusor testcase
 
> Is the centreline velocity the max. velocity of the channel velocity profile ?

The answer is yes.

>If my channel has a height of 1m and the fluid is air am I right if the the centreline velocity is:

> v=Re*vis/rho/height=20000*1.8e-5/1.16/1=0.3134 m/s.

The answer is yes again.

> In the testcase there was separation but in my calcs there's attached flow.

As John said, this is a very difficult test case for turbulence models (so it's a good one in fact). I'm not so surprise that the numerical result doesn't show the correct behavior.

The proceedings of this workshop show how it was difficult to correctly predict the recircultaion zone.

frederic felten January 30, 2001 13:42

Re: Diffusor testcase
 
hi there,

i don't exactly know all the details of your calculation, but you should be able to find some very usefull informations in the following publication:

H.-J. Kaltenbach, M. Fatica, R. Mittal , T. S. Lund, and P. Moin. "Study of flow in a planar asymmetric diffuser using Large Eddy Simulation" J. Fluid Mech. 390, 151-185, 1999.

I hope this helps. Sincerely,

Frederic Felten CFD Lab, UT Arlington. http://utacfdb.uta.edu/

John C. Chien January 30, 2001 14:11

Re: Diffusor testcase
 
(1). I don't see anything wrong with the maximum centerline velocity, if it is fully-developend. Sometimes, people use the averaged inlet velocity. (2). My estimate of the velocity is around 1m/sec, which is close to your number.

Marat Hoshim January 31, 2001 03:18

Re: Diffusor testcase
 
John,

your estimate is around 1m/sec. But that 3 times higher than my estimate ! Where does that difference come from ?

Regards, Marat

John C. Chien January 31, 2001 05:38

Re: Diffusor testcase
 
(1). It comes from my estimate of the kinematic viscosity of the air.

Marat Hoshim January 31, 2001 12:51

Re: Diffusor testcase
 
So, what is your estimate for the kinematic viscosity of air ?

Regards, Marat

John C. Chien January 31, 2001 16:15

Re: Diffusor testcase
 
(1).In the book of Boundary Layer Theory, by Schlichting, the kinematic viscosity of air at 68degreeF, 14.7psi, is 160x10E-06 FT*FT/SEC. (2). You get 3.2 FT/SEC when U is calculated from U=Re*nu/L, with RE=20,000. So, U is around 1m/sec.

sylvain February 1, 2001 11:48

Re: Diffusor testcase
 
I'm sorry, but you get U=3.2 FT/SEC if L=1FT but if L=1m=3.2FT, it comes U=0.097FT/SEC, which leads to U=0.31m/s.

Best regards,

Sylvain

John C. Chien February 1, 2001 13:04

Eureka!
 
(1). I think, you are right. (2). That's why I said, it was an estimate. It's important to double check the result. Thank you. By now everybody knows how to compute the Reynolds number.

clifford bradford February 12, 2001 13:27

Re: Diffusor testcase
 
Marat, as John pointed out your issue is probably with your turbulence model. I guess you're using k-e (because it's so popular) unfortunately k-e is known to be poor for adverse pressure gradients. you may want to use the k-omega. See David Wilcox's "Turbulence Modelling for CFD"


All times are GMT -4. The time now is 01:20.