CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Large Eddy Simulation inlet (http://www.cfd-online.com/Forums/openfoam-solving/67821-large-eddy-simulation-inlet.html)

prapanj August 27, 2009 06:10

Large Eddy Simulation inlet
 
Hi,

I am trying to simulate a wind tunnel case of flow over cylinder at a range of high Reynolds numbers. I was trying to simulate a certain length of the cylinder and cap the domain by symmetry Plane boundary conditions. I have values for inlet turbulent intensity, length scale etc. But I am not able to simulate a proper inlet as I don't have any idea how I can achieve a turbulent inlet. Could some one please suggest me some options ?

Thank You
Prapanj.

lakeat August 27, 2009 06:40

Is turbulent inlet crucial for your case? :D
What's your your les model?
What is the Re number?
Have you read the usersguide and noticed a turbulentInlet entry, will it meet your need?
Would you mind if the inlet is mapped from a channel flow case?

prapanj August 27, 2009 07:22

Hi lakeat,

In fact I was just reading another forum entry from you :)

I am not aware of a turbulent inlet entry. Is it just specification of the turbulent kinetic energy ? Because I think I will need upstream turbulence for this case.

How do I generate upstream turbulence using channel. I am aware of mapping fields but not aware how channel flow could get me some turbulence.

The reynolds number I am working is of the order of 10^6.

Do you think You have a suggestion?

Thanks,
Prapanj.

prapanj August 27, 2009 07:23

Answering the left out questions,

I am using dynOneEqnEddy model.

Yes upstream turbulence is crucial for my case as I am dealing with coefficient of drag values. And I see these values are affected by my upstream turbulence.

Thanks

prapanj August 27, 2009 07:35

Daniel,

I just checked the turbulentInlet and it looks like a good starting point. Do you think I could achieve better solutions with channel mapping?

What do they mean by a reference field and fluctuation scale? In other words how do we relate these to turbulence Intensity and average velocity? I hope I should get what I need by giving this BC to my velocity field right?

Thank You

prapanj.

sandy August 27, 2009 09:19

Quote:

Originally Posted by lakeat (Post 227650)
Is turbulent inlet crucial for your case? :D
What's your your les model?
What is the Re number?
Have you read the usersguide and noticed a turbulentInlet entry, will it meet your need?
Would you mind if the inlet is mapped from a channel flow case?

Hi lakeat, if my case is an external flow but an internal flow, how can I get inlet k in the oneEqnEddy model?

bjj August 29, 2009 08:46

I have had some good results with the directMapping BC for generating the turbulent inlet for LES. This will also generate k on the inlet. Have a look in the pitzDailyDirectMapped tutorial.

You might also need to trigger the turbulence in the initial condition for example by using foamCalc randomize.

lakeat August 29, 2009 10:31

Quote:

Originally Posted by prapanj (Post 227656)
Answering the left out questions,

I am using dynOneEqnEddy model.

Yes upstream turbulence is crucial for my case as I am dealing with coefficient of drag values. And I see these values are affected by my upstream turbulence.

Thanks

Since you see they are affected by upstream turbulence, then I guess it would be a problem of turbulence transition, and then, will you get good prediction without set your 1st grid point around y+=1. And if so, have you considered whether you would afford for such a large computation? what is your grid numbers?

prapanj August 30, 2009 12:01

Hi Bjj,

I see using direct Mapped boundary should be an option. But my concern is, what primer domain do I use to get the turbulence? I hear a lot about the channel flow that generate turbulence. But how does this primer region generate turbulence?

I am using OF 1.5.x. I don't see a tutorial that has pitzDailyDirectMapped. Or my chief concern is, where do I map my inlet from?

As for foamCalc randomize, don't I need some sort of regression to generate turbulence that has some time dependant properties? Will generating random numbers for U inlet help?

Thanks
Prapanj.

prapanj August 30, 2009 12:03

Hi Daniel,

I hope I can go upto 2 million cells with my current computational ability. Still, I know my deltaT could be as low as 0.0001 for this kind of a problem for the Co to be within bounds. I am still ready to try that.
My chief concern is inlet turbulence though.

Thanks,
Prapanj.

bjj August 30, 2009 13:52

Dear Prapanj,

With the directMapped BC you are recycling the turbulent quantatives from a mapping plane inside the domain back to the inlet. The mapping plane is defined with an offset distance from the inlet patch. I remember there are some different threads in the forum discussing this.

The pitzDailyDirectMapped tutorial is from OF 1.6 so you can try to have a look in that to see how it is set up. You can also find some description of this BC and other methods for generating turbulent inlet BC in Eugene De Villiers PhD. In general this is a good reference for les in OF.

With foamCalc randomize you can generate a pertubation in the U field as the initial condition simply to trigger the generation of turbulence. I think Eugene covers this as well in his PhD.

Regards,
Bjarne

roth August 31, 2009 08:45

Quote:

Originally Posted by prapanj (Post 227939)

I am using OF 1.5.x. I don't see a tutorial that has pitzDailyDirectMapped. Or my chief concern is, where do I map my inlet from?

OpenFOAM-1.5.x/tutorials/oodles/pitzDailyDirectMapped

prapanj August 31, 2009 09:06

Thanks Mike :)

Btw, I was trying to use turbulentInlet BC. I assume the fluctuation scale is the turbulent intensity in each direction. The simulation seems fine, except that I can't make any sense out of pressure field. For a initial pressure field of 1e5, inlet zeroGradient and outlet fixedValue 1e5, I was getting a pressure field ranging from 9.6e4 to 1e5. And because of this, my Cd values kept oscillating from iteration to iteration. Should I be using a constant total pressure at inlet for a turbulentInlet in velocity ? (it should be working fine with zerogradient at inlet right? )

Thanks,
Prapanj.


All times are GMT -4. The time now is 15:24.