[mbranag]

Hi all,

I am running a thin film hydrodynamic bearing model and am having some issues with the turbulence model. When you include a turbulence model under laminar conditions, should the results be equal to that of the case with no turbulence model.

Thanks

[ghorrocks]

It depends on the turbulence model. k-omega based turbulence models degenerate to zero turbulent viscosity when k=0, so they will be pretty close. k-epsilon based models cannot handle k=0 so they do not degenerate to laminar flow well. There are modifications to the k-e model to help this, but then things start getting even more complex.

In short, if your flow is near transition then a k-omega based model handles it much better. If you use the SST model you also have the benefit of a turbulence transition model. But note the turb transition model has not been tuned for thin film bearings so its predictions could be considerably wrong.

[mbranag]

I am seeing a drop of 1/4-1/3 drop in peak pressure using k-omega model. The Reynolds number of the flow as reported by a 3rd party software is 250 so the flow should be laminar. Does this sound strange?

Thanks

[BlnPhoenix]

Try modelling it in laminar. Also as Ghorrocks mentioned, k-w SST may not be tuned to your specific task (thin film)

[icornejo]

Basically all eddy viscosity models will artificially induce turbulence where there is not (in laminar flows). It is just how they work. One of the basic assumptions of k-w, k-e, etc models is that the flow is fully turbulent. Even if you set the zone as a laminar zone, you will just kill the turbulence production. Eddy viscosity models does not predict transition right, you shouldnt trust in any result where the turbulence is low. If it is a thin domain, with low Re, why dont you just use laminar flow? Your Re number appears to be way below the critical one for a thin layer.

Sent from my SM-T810 using CFD Online Forum mobile app

[ghorrocks]

If you have a laminar flow and you run it with a laminar flow model and a k-omega turbulence model the results will be just about the same for most cases. This is because k-omega models degenerate to laminar flow models when k=0. They usually do add a bit of dissipation so the effective viscosity is slightly higher than a laminar model, but for many cases this is small with the result that a laminar model and k-omega based model will give essentially the same result.

So if you know the flow is laminar then it is best to use a laminar model, but if you are not sure then use k-omega (or one of its derivatives).

[mbranag]

In the problem that I am trying to investigate, I am trying add features to the the stationary surface which may induce turbulence. I posted it in another thread but I am seeing a pressure drop occur that should be a pressure rise. This pressure drop occurs well into the laminar regime. I have tried the zero order model, the k-omega model, and the BSL Reynolds stress model and am seeing the same behavior from all three of them.

[farzadmech]

Hello there and kOmegaSST give 30% higher values for drag coefficient(Cd). please see the figure and read my post;
laminar Flow over a sphere(laminar vs KOmegaSST simulation)




Thanks,
Farzad

Originally Posted by ghorrocks

If you have a laminar flow and you run it with a laminar flow model and a k-omega turbulence model the results will be just about the same for most cases. This is because k-omega models degenerate to laminar flow models when k=0. They usually do add a bit of dissipation so the effective viscosity is slightly higher than a laminar model, but for many cases this is small with the result that a laminar model and k-omega based model will give essentially the same result.

So if you know the flow is laminar then it is best to use a laminar model, but if you are not sure then use k-omega (or one of its derivatives).

[Opaque]

Activate beta features, see Edit/Options.

Open your domain, and take a look at your turbulence model selection. You may see an additional option named Blended Near Wall Treatment, select it and compare your results one more time.

[ghorrocks]

Also, have you done a sensitivity analysis on your mesh? If not then your results are inaccurate and comparing against published results is meaningless. See https://www.cfd-online.com/Wiki/Ansy..._inaccurate.3F

[farzadmech]

Dear Opaque
I am using Openfoam and I think you are talking about fluent? am I right?

In Sphere simulation Cd is almost 30% higher, but flow seems reasonable when I compare laminar and turbulent flow(see sphere.jpg). but for airfoil like geometry at angle of attach 30 degree at Re<500 even flow visualization seems to show wrong answers. Do you know why?


Thank,
Farzad

Originally Posted by Opaque

Activate beta features, see Edit/Options.

Open your domain, and take a look at your turbulence model selection. You may see an additional option named Blended Near Wall Treatment, select it and compare your results one more time.

[farzadmech]

Dear ghorrocks
Laminar result for sphere is very good, but I want to use turbulent model. I fact I am going to change the shape to airfoil like geometry and I do not know if my flow is laminar or turbulent there, so I want a model to simulate it whether it is laminar or turbulent. By the way, I am using openFoam and I have used a very fine mesh as you see in the attach figure. I just want to know that kOmegaSST is capable of handling laminar flow?

Thanks,
Farzad

Originally Posted by ghorrocks

Also, have you done a sensitivity analysis on your mesh? If not then your results are inaccurate and comparing against published results is meaningless. See https://www.cfd-online.com/Wiki/Ansy..._inaccurate.3F

[ghorrocks]

My post #2 answers your question on whether SST can model laminar flow.

We cannot help you with the details of OpenFOAM - try the OF forum for that. This is the CFX forum.

I have no idea why your flat plate results are weird. It will be something in your setup, and as it is OpenFOAM we cannot help you there.
If you want accurate drag numbers in the regime where the sphere has a Von Karman vortex street you will need to resolve the vortices. So you will need a transient model and time average the drag. You should be able to model this using SST reasonably well, at least until effects like laminar to turbulent transition become important.