[liu]

Hi all,

My LES of turbulent channel flow Retau=395 finally goes to turbulent (or at least it looks like it). What I am trying to do is to reproduce Eugene's case in his thesis.

First, I compared the velocity profile with DNS data. Again, it looks OK. The only thing I am not so happy is channelFoam never really get Retau=395. It stays at about 384. Maybe we should not impose bulk velocity, but the pressure gradient.

Second, I tried to compare the spectrum. What I don't understand is how Fig. 5.7 in Eugene's thesis was produced. To be specific, I have the following doubts:

1. What is the meaning of k? It says "total resolved turbulent energy"? Is it the k field in the output? Or just k=0.5*(Ux*Ux+Uy*Uy+Uz*Uz), where Ux, Uy, and Uz are velocity components?

2. Is this a spatial spectrum or temporal spectrum? I suppose this is the spatial spectrum. What he did is to average the U over z direction in a xz plane (y+=8) parallel to the wall. Then U_avg(x_i) is only a function of x_i, where x_i is the center of the cells (Say we have 45 cells in x direction, then x_i = 1 to 45). Then we do spectrum on this U_avg(x_i). Am I right?
Anyway, I didn't get the same plots as in Fig 5.7. I know something is wrong along the way.

3. Also the wave number on x-axis (Fig. 5.7), how it f and U defined?

Thanks for you help!

[eugene]

Hi,

1. It is neither k = 0.5 (magSqr(U - Umean))
2. It is a temporal spectrum.
3. f is frequency [Hz], U is the phase velocity = Umean at the measurement location

Hope this helps,

Eugene

[liu]

Thanks.

I will try it.

How long did you take the sample for spectrum? Did you normalize E(k) with something? The order of magnitude for k is much smaller than 1 when calculated.

[eugene]

I really cannot recall how long the sampling ran for. It would have to be long enough to capture the lowest frequencies, so somewhere of the order of 1 second simulated time.

What order of magnitude do you get for E? Unfortunately, I cant help you with the definition, but it is unlikely that it is anything but turbulent energy since I would have provided some kind of equation if that were the case.(Note that the "k" in "E(k)" refers to the wavenumber, not the turbulent energy.)

[lakeat]

Hi Liu,

Here are some of my ideas,

1. I was recommended some years ago to look deep into Pope's "Turbulence Flows" to get a basic idea what turbulence spectrum is.
2. Taylor frozen turbulence field hypothesis is used, since the spatial resolution is far from enough in common channel flow simulation.

Bye,

[cheng1988sjtu]

Hi Liu,

How long (simulate flow time) have you run the case to get a turbulence?

Did you change the code? or the initial condition? k for example?

Right now, I'm basically doing the same thing, i.e. reproduce the log law and turbulence intensities. However, I just can't get the turbulence.

Maybe you can see the link to my thread:
http://www.cfd-online.com/Forums/ope...tml#post293152

Basically, the pressure gradient is dropping, though much slower when run it to 100 seconds. so I just wonder how you set up the case for it. If it's no bother for you, can you show me your case? via email or posting a reply as you like.

Thank you!

[aka]

Originally Posted by liu

Hi all,

My LES of turbulent channel flow Retau=395 finally goes to turbulent (or at least it looks like it). What I am trying to do is to reproduce Eugene's case in his thesis.

First, I compared the velocity profile with DNS data. Again, it looks OK. The only thing I am not so happy is channelFoam never really get Retau=395. It stays at about 384. Maybe we should not impose bulk velocity, but the pressure gradient.

Second, I tried to compare the spectrum. What I don't understand is how Fig. 5.7 in Eugene's thesis was produced. To be specific, I have the following doubts:

1. What is the meaning of k? It says "total resolved turbulent energy"? Is it the k field in the output? Or just k=0.5*(Ux*Ux+Uy*Uy+Uz*Uz), where Ux, Uy, and Uz are velocity components?

2. Is this a spatial spectrum or temporal spectrum? I suppose this is the spatial spectrum. What he did is to average the U over z direction in a xz plane (y+=8) parallel to the wall. Then U_avg(x_i) is only a function of x_i, where x_i is the center of the cells (Say we have 45 cells in x direction, then x_i = 1 to 45). Then we do spectrum on this U_avg(x_i). Am I right?
Anyway, I didn't get the same plots as in Fig 5.7. I know something is wrong along the way.

3. Also the wave number on x-axis (Fig. 5.7), how it f and U defined?

Thanks for you help!

Hi Liu,
What modification did u do to get turbulent flow in your channel? I am also getting the same problem in my channel simulations. My nuSgs is close to nu (molecular viscosity) in most of the flow domain. However, I run the same case using k-omega model and nut values are reasonable and are about 1000 times larger than nuSgs in some spatial locations ( mainly close to the mid-section of the channel).

I also appreciate if anyone has compared nut and nuSgs for the same channel flow cases.

Any advice please?

Thanks,