LES of turbulent channel flows

 Register Blogs Members List Search Today's Posts Mark Forums Read

January 4, 2017, 19:38
#201
Member

Join Date: Apr 2016
Posts: 34
Rep Power: 9
Quote:
 Originally Posted by Elham I don't know if the setAverage does push the pressure difference like fvOption. Why you did not use fvOption in this case?

I have used fvOptions too but it doesn't make any difference. Nevertheless, since I am solving for incompressible flow and because of the continuity, the setAverage option must trigger the required the pressure difference.

January 4, 2017, 22:28
#202
Senior Member

Elham
Join Date: Oct 2009
Posts: 184
Rep Power: 15
Quote:
 Originally Posted by Eman. I have used fvOptions too but it doesn't make any difference. Nevertheless, since I am solving for incompressible flow and because of the continuity, the setAverage option must trigger the required the pressure difference.
How big is the Re number? Is it in the turbulent limit?

January 6, 2017, 23:23
#203
Member

Join Date: Apr 2016
Posts: 34
Rep Power: 9
Quote:
 Originally Posted by Elham How big is the Re number? Is it in the turbulent limit?

Yes. Of course. It's approximatively 10000.

January 7, 2017, 02:46
#204
Senior Member

Elham
Join Date: Oct 2009
Posts: 184
Rep Power: 15
Quote:
 Originally Posted by Eman. Yes. Of course. It's approximatively 10000.
You can try using turbulentIntensityKineticEnergyInlet BC for k and see what will happen. Please let me know if it works.

 January 9, 2017, 04:53 #205 New Member   bin xu Join Date: Jan 2016 Posts: 5 Rep Power: 9 hellow i find that the the result of channel395 in of2.4 is quite different from of3.0. I have changed the Sgs model to dylagragian model in of 2.4 ,and the result seems similar to of 3.0 . i wonder if there is something wrong in oneEqu model in of2.4 ? I think the u_tau should be equal to 0.0079 in channel395 ,but however i changed the settings, my u_tau never got right. I use both ways trying to calculate u_tau 1.u_tau=sqrt(h*dp/dx) 2.u_tau=sqrt(nu*u1/y1),u1 and y1 are data on the first point above the wall. my simulation data are like this sorry for my bad english. hope someone can help me solve this problem. thanks ! Haitham Osman CFD likes this.

January 10, 2017, 02:11
#206
Member

Join Date: Apr 2016
Posts: 34
Rep Power: 9
Quote:
 Originally Posted by Elham You can try using turbulentIntensityKineticEnergyInlet BC for k and see what will happen. Please let me know if it works.
Thanks for the suggestion. Yes, it works but the problem is that I don't know the turbulence intensity beforehand and I don't want to use synthetic b.c for the inlet. I think I better stick to a precursor simulation and forget about mapped boundary condition.

 February 10, 2017, 14:31 #207 Member   Darko Radenkovic Join Date: Oct 2015 Posts: 38 Rep Power: 9 Hello. I did LES Smagorinsky simulation with van Driest damping and I obtained good results, for turbulent flow in channel. Then I wanted to try dynamic model and I tried dynamicKEqn. IAlso I changed delta to cubeRootVol. I kept all other settings the same as for Smagorinsky, which gave good results. I had run simulation for about 90 flow-through times and after that I started averaging for about 250 flow-through times. Results are not good (velocity is in file on link below). Does anyone have an idea what could be the reason for discrepancy of this model and log law? Relevant settings are on this link https://www.dropbox.com/s/ct104fg36l...cKEqn.zip?dl=0 Any help is appreciated. Regards, Darko

February 20, 2017, 21:55
#208
New Member

bin xu
Join Date: Jan 2016
Posts: 5
Rep Power: 9
Quote:
 Originally Posted by dradenkovic Hello. I did LES Smagorinsky simulation with van Driest damping and I obtained good results, for turbulent flow in channel. Then I wanted to try dynamic model and I tried dynamicKEqn. IAlso I changed delta to cubeRootVol. I kept all other settings the same as for Smagorinsky, which gave good results. I had run simulation for about 90 flow-through times and after that I started averaging for about 250 flow-through times. Results are not good (velocity is in file on link below). Does anyone have an idea what could be the reason for discrepancy of this model and log law? Relevant settings are on this link https://www.dropbox.com/s/ct104fg36l...cKEqn.zip?dl=0 Any help is appreciated. Regards, Darko
my result with all LES delta models and LES models are similar to your second simulation result.i even tried 3 different versions of openfoam code. but my data in log law are always bigger than the DNS result.can you share your first simulation settings to help me .

Thanks!

Regards,
Xu Bin

 February 21, 2017, 03:13 #209 Member   Darko Radenkovic Join Date: Oct 2015 Posts: 38 Rep Power: 9 Xu Bin, Keep all settings the same as in my previous attached file, just change delta to vanDriest and LESModel to Smagorinsky. These settings should be the same as settings which can be found in tutorial in folder channel395 in OpenFOAM. Regards, Darko

 February 22, 2017, 03:31 #210 New Member   bin xu Join Date: Jan 2016 Posts: 5 Rep Power: 9 Darko, how do you calculate u_tau to nondimensionalize u and y to get the u+ and y+ data ? Regards Xu Bin

 February 22, 2017, 04:00 #211 Member   Darko Radenkovic Join Date: Oct 2015 Posts: 38 Rep Power: 9 Xu Bin, After using postChannel utility, I calculate friction velocity as $u_tau = \sqrt{\nu \frac{u_{fc}}{y_fc}}$. Index fc denotes first cell, and \nu is kinematic viscosity. About non-dimensional distance: https://www.cfd-online.com/Wiki/Dime...tance_(y_plus) Regards, Darko Haitham Osman CFD likes this.

March 1, 2017, 03:18
#212
New Member

bin xu
Join Date: Jan 2016
Posts: 5
Rep Power: 9
Quote:
 Originally Posted by Elham I finally could get rid of the scale difference from my results and DNS ones in wall coordinate by calculating u_tau as following: u_tau=sqrt(wallShearStressLES) and wallShearStressLES is a utility. Then I calculate y+.
Hi,Elham

do you get the correct u_tau=0.0079 for case channel 395. i find that many people get the result u_tau = 0.73. i also meat the problem.i have tried all approaches to calculated u_tau, the result is still wrong. can you help me ?

Regards

Xu Bin

March 2, 2017, 07:03
#213
New Member

Jem
Join Date: Dec 2016
Posts: 19
Rep Power: 8
Hey can you tell me how you find the first point above the wall?

Quote:
 2.u_tau=sqrt(nu*u1/y1),u1 and y1 are data on the first point above the wall.

 March 20, 2017, 22:20 #214 Senior Member   Elham Join Date: Oct 2009 Posts: 184 Rep Power: 15 Hi everyone, I have a turbulent flow in a channel. The channel is 0.3*0.05*0.075m. When I add a stationary 2mm diameter sphere in the domain the turbulent gets laminar quickly. The sphere BC is wall. I suppose the presence of sphere causes the flow to get smooth. How can I have a turbulent flow again? Thanks in advance

 July 13, 2017, 11:32 #215 New Member   Join Date: Jan 2017 Posts: 4 Rep Power: 8 Hi everyone, There is one thing I'm very curious. In the beginning of this thread, people compared their LES data obtained from Openfoam with DNS data from Moser et al. I never used openfoam but as far as I know, the accuracy of openfoam is second order in space. But Moser et al. used spectral method for spatial discretization which is more accurate than second order central scheme. If you use the same mesh resolution, you compare your LES result from openfoam with DNS data from Moser et al.,it is reasonable to have some discrepency not only because the difference between LES and DNS, but also because the numerical method in openfoam is less accurate. So why don't you compare your LES result with DNS data which used second order central difference scheme. I did a DNS simulation using an in-house code which used second order central difference scheme and compare my result with "DNS of turbulent heat transfer in channel flow with respect to Reynolds and Prandtl number effects" from Kawamura et al. In this paper they also used second order central difference scheme so my result matched very well with Kawamura et al. for u+, t+,t't',u'u',v'v',w'w',u't' v't'. So why don't you compare your result with Kawamura et al. but with Moser et al.? Best regards, Wentao Guo

August 13, 2019, 05:30
#216
New Member

Join Date: Jan 2014
Posts: 20
Rep Power: 11
Quote:
 Originally Posted by cedric_duprat Hi Philippe, Yes, you can do that, just telling that the shear stress in your periodic channel is compensated by the favorable pressure gradient. So you get u_tau to get U+ and y+. Do you plan to let your results avaiable ? I'm quite interresting in that just to compare with my results. Notice that with the value of u_tau, you can calculate a Re_tau which is better to compare with DNS results. Cedric
Dear Cedric,

Can you help me with something?

I tried to run channel395 tutorial case in OpenFoam using channelFoam solver for a larger period of time (i.e. 7000s). What I noticed is that pressure gradient value gradP wobbles over some average value, so I believe simulation can be considered statistically steady state.
However, when I tried to calculate uTau from time averaged gradP value ( uTau = sqrt(gradP) ) relative errors (compared to the DNS values) were equal to cca 20% (which is a lot!). The values reported by Eugene in his PhD were much better for similar mesh and same subgrid-scale models. Did I miss something?

P.S. I know Eugene noted in his PhD (page 163) that:
To ensure consistency, the streamwise bulk velocity, Ub, through the channel is adjusted to be equal to the average DNS value by varying an imposed streamwise constant pressure gradient, ∂P/∂x. The time averaged value of this quantity is equivalent to the mean wall shear stress and can be compared to the corresponding DNS value.
According to that, tau_w should be equal to time-averaged value of gradP and u_Tau as sqrt(gradP). Or I made some mistake?

Thank you!
Best regards!

August 13, 2019, 05:55
#217
Senior Member

Ehsan Asgari
Join Date: Apr 2010
Posts: 472
Rep Power: 17
Quote:
 Originally Posted by camel Dear Cedric, Can you help me with something? I tried to run channel395 tutorial case in OpenFoam using channelFoam solver for a larger period of time (i.e. 7000s). What I noticed is that pressure gradient value gradP wobbles over some average value, so I believe simulation can be considered statistically steady state. However, when I tried to calculate uTau from time averaged gradP value ( uTau = sqrt(gradP) ) relative errors (compared to the DNS values) were equal to cca 20% (which is a lot!). The values reported by Eugene in his PhD were much better for similar mesh and same subgrid-scale models. Did I miss something? P.S. I know Eugene noted in his PhD (page 163) that: To ensure consistency, the streamwise bulk velocity, Ub, through the channel is adjusted to be equal to the average DNS value by varying an imposed streamwise constant pressure gradient, ∂P/∂x. The time averaged value of this quantity is equivalent to the mean wall shear stress and can be compared to the corresponding DNS value. According to that, tau_w should be equal to time-averaged value of gradP and u_Tau as sqrt(gradP). Or I made some mistake? Thank you! Best regards!
Hi,

Just calculate u_tau using the mean velocity. Sure and straight.

Regards,
Syavash

August 13, 2019, 06:13
#218
New Member

Join Date: Jan 2014
Posts: 20
Rep Power: 11
Quote:
 Originally Posted by syavash Hi, Just calculate u_tau using the mean velocity. Sure and straight. Regards, Syavash
Hi Syavash,

I agree with you.

However, theory implies that same uTau value should be obtained from the pressure gradient value.

Well, I checked it and I have to say that tau_w can really be calculated from average gradP value. Moreover, I tried to calculate uTau from both, the pressure gradient value and as sqrt(nu*(uc/yc)), where uc is velocity at wall adjacent cell centre and yc is cell centre distance from the wall. The results are basically the same.

Best regards!

August 13, 2019, 11:28
#219
Senior Member

Ehsan Asgari
Join Date: Apr 2010
Posts: 472
Rep Power: 17
Quote:
 Originally Posted by camel Hi Syavash, I agree with you. However, theory implies that same uTau value should be obtained from the pressure gradient value. Well, I checked it and I have to say that tau_w can really be calculated from average gradP value. Moreover, I tried to calculate uTau from both, the pressure gradient value and as sqrt(nu*(uc/yc)), where uc is velocity at wall adjacent cell centre and yc is cell centre distance from the wall. The results are basically the same. Best regards!
Good to hear that,

Yes, there is a force balance between streamwise pressure gradient and the wall shear stress. However, I usually use the mean velocity profile to calculate the shear stress as you indicated. How is it possible to calculate the mean pressure gradient?! I guess the value written in the last time folder is the instantaneous pressure gradient, is that right?

Regards,

August 14, 2019, 09:25
#220
New Member

Join Date: Jan 2014
Posts: 20
Rep Power: 11
Quote:
 Originally Posted by syavash Good to hear that, Yes, there is a force balance between streamwise pressure gradient and the wall shear stress. However, I usually use the mean velocity profile to calculate the shear stress as you indicated. How is it possible to calculate the mean pressure gradient?! I guess the value written in the last time folder is the instantaneous pressure gradient, is that right? Regards, Syavadh
Yes. Pressure gradient value in the log file basically represents the wall shear stress. You can easily extract the pressure gradient values from the log files, perform time averaging and calculate time-averaged wall shear stress. Friction velocity value is accordingly equal to sqrt of time-averaged wall shear stress.

Best regards!