Non-linear k-eps model

oj.bulmer · April 20, 2013, 09:01

One of the biggest limitations in typical 2-eqn models is that it assumes isotropic turbulence. To address this, k-eps EARSM model in CFX reportedly mimics the anisotropic physics by including non linear relationship between Reynolds stresses, mean strain rate and vorticity tensors in Boussinesq hypothesis, as against the linear relationship used in two equation models.

From the description, it sounds tempting to be used widely to capture the secondary flows, but it is rarely used. Aside from insufficient evaluation about robustness (I think), are there any caveats about the use of this model?

OJ

ghorrocks · April 21, 2013, 08:34

Give it a go in your situation. I do not have much experience with it, when I want an anisotropic turbulence model I go for either RSM or LES. So try it out and let us know how it goes.

The caveat on this model is almost certainly more difficulty in obtaining convergence. So expect to have a hard time getting it to converge.

oj.bulmer · April 21, 2013, 18:29

I certainly have tried this model in the past to assess its applicability, using data from following paper:

Younis and Chokjat 1997, "Eddy viscosity vs. 2nd. Order closures for flows in noncircular ducts," AIChE Journal 43, pp. 893-901.

Interestingly, the k-eps EARSM was able to capture the secondary flows in a non-circular duct, at par with RSM (though over-predicted). Whilst, k-eps RNG failed.

If I remember correctly, the model did give me a hard time while converging. In fact, I had to coarsen (!) the mesh a little to have some reasonable convergence. But then, I thought that it was the flow, and not the model, that was to blame. Somewhere I suddenly read about this model and all those memories poured in...

In fairness, if the non-linear definition of eddy-viscosity here is matured and the model made more robust, I think this model provides a good trade-off between a regular 2-eqn model and RSM.

OJ

ghorrocks · April 21, 2013, 19:49

Quote:

In fairness, if the non-linear definition of eddy-viscosity here is matured and the model made more robust, I think this model provides a good trade-off between a regular 2-eqn model and RSM.

Absolutely! And if you can manage it you will win a Nobel prize

. Do not underestimate the difficulty of it.

oj.bulmer · April 24, 2013, 04:50

Indeed, it is tricky but then, given how CFD has evolved over the years, there is always a room for wishful thinking

OJ

April 20, 2013, 09:01	Non-linear k-eps model	#1
oj.bulmer Senior Member OJ Join Date: Apr 2012 Location: United Kindom Posts: 473 Rep Power: 20	One of the biggest limitations in typical 2-eqn models is that it assumes isotropic turbulence. To address this, k-eps EARSM model in CFX reportedly mimics the anisotropic physics by including non linear relationship between Reynolds stresses, mean strain rate and vorticity tensors in Boussinesq hypothesis, as against the linear relationship used in two equation models. From the description, it sounds tempting to be used widely to capture the secondary flows, but it is rarely used. Aside from insufficient evaluation about robustness (I think), are there any caveats about the use of this model? OJ

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
about Subgrid-scale model	impecca	OpenFOAM Running, Solving & CFD	4	December 20, 2013 11:36
Low pressure drop with the same model	zhoubinwx	OpenFOAM Running, Solving & CFD	5	June 16, 2009 06:10
Water pump OpenFOAM 15 ANSYS CFX 110 comparation	waynezw0618	OpenFOAM Running, Solving & CFD	39	March 5, 2009 13:57
LES and combustion model	Margherita Cadorin	CFX	0	October 29, 2008 06:24
linear eddy model	wind_flowcfd	Main CFD Forum	0	November 4, 2005 14:51

April 21, 2013, 08:34		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,852 Rep Power: 144	Give it a go in your situation. I do not have much experience with it, when I want an anisotropic turbulence model I go for either RSM or LES. So try it out and let us know how it goes. The caveat on this model is almost certainly more difficulty in obtaining convergence. So expect to have a hard time getting it to converge.

April 21, 2013, 18:29		#3
oj.bulmer Senior Member OJ Join Date: Apr 2012 Location: United Kindom Posts: 473 Rep Power: 20	I certainly have tried this model in the past to assess its applicability, using data from following paper: Younis and Chokjat 1997, "Eddy viscosity vs. 2nd. Order closures for flows in noncircular ducts," AIChE Journal 43, pp. 893-901. Interestingly, the k-eps EARSM was able to capture the secondary flows in a non-circular duct, at par with RSM (though over-predicted). Whilst, k-eps RNG failed. If I remember correctly, the model did give me a hard time while converging. In fact, I had to coarsen (!) the mesh a little to have some reasonable convergence. But then, I thought that it was the flow, and not the model, that was to blame. Somewhere I suddenly read about this model and all those memories poured in... In fairness, if the non-linear definition of eddy-viscosity here is matured and the model made more robust, I think this model provides a good trade-off between a regular 2-eqn model and RSM. OJ

April 24, 2013, 04:50		#5
oj.bulmer Senior Member OJ Join Date: Apr 2012 Location: United Kindom Posts: 473 Rep Power: 20	Indeed, it is tricky but then, given how CFD has evolved over the years, there is always a room for wishful thinking OJ