|
[Sponsors] |
August 13, 2007, 07:13 |
I use smagorinky model and cub
|
#41 |
Member
hadi tartoussi
Join Date: Mar 2009
Location: paris
Posts: 48
Rep Power: 17 |
I use smagorinky model and cubeRootVolDelta.
Can u please tell me if i am wrong: In a parallel permanent flow <u>=U(y) all the kinetic quantities depend on y only. NS: d/dy<u'v'>=d/dx<p/rho>+nu d^2/dy^2 U At the wall <u'v'>=0 integration y=0,2delta => delta d/dx<p/rho>= u_tau^2. but in my case the modeled part of <u'v'> is not 0?! Can you please tell me what relation u use between gradP and U_tau? Thanks in advance Hadi |
|
August 14, 2007, 04:45 |
Hello Eugene,
Can i get goo
|
#42 |
Member
hadi tartoussi
Join Date: Mar 2009
Location: paris
Posts: 48
Rep Power: 17 |
Hello Eugene,
Can i get good result by using cubeRoot Vol Delta? After 160 flows through time i still cant see that the mean fields tend to be symmetric Do u think that is because i didn't use vanDriest damping for the Smagorinsky model? R Field |
|
August 15, 2007, 06:42 |
Is the image above a mean fiel
|
#43 |
Senior Member
Eugene de Villiers
Join Date: Mar 2009
Posts: 725
Rep Power: 21 |
Is the image above a mean field after 160 flowthrough times? I find that highly unlikely. I suggest you restart averaging by deleting the nAvergaingSteps.raw files in the time uniform directories.
|
|
August 16, 2007, 05:14 |
hi Eugene,
This is the R fi
|
#44 |
Member
hadi tartoussi
Join Date: Mar 2009
Location: paris
Posts: 48
Rep Power: 17 |
hi Eugene,
This is the R field, which i suppose it is a running average. I know that the mean Fields are calculated when running channelOodles.(calculateAverages.H) As you said i should delete the nAveragingSteps in the saved time directory, but i don't know how to redo the averaging. I ran the simulation for 200 flows through time. Can u please help me Hadi |
|
August 16, 2007, 07:02 |
Hmmm, if its R it is the avera
|
#45 |
Senior Member
Eugene de Villiers
Join Date: Mar 2009
Posts: 725
Rep Power: 21 |
Hmmm, if its R it is the average. Unless the distribution is due to some very high initial values, this is very strange. What does Umean look like?
To restart averaging, just delete nAveragingSteps.raw and restart the calculation from your last saved time. |
|
August 16, 2007, 07:08 |
My initial velocity profile is
|
#46 |
Member
hadi tartoussi
Join Date: Mar 2009
Location: paris
Posts: 48
Rep Power: 17 |
My initial velocity profile is shown in my previous message dated Thursday, August 09, 2007
this is my Umean |
|
August 16, 2007, 07:17 |
Sorry, I don't know what to sa
|
#47 |
Senior Member
Eugene de Villiers
Join Date: Mar 2009
Posts: 725
Rep Power: 21 |
Sorry, I don't know what to say except "keep averaging". At least the latest Umean field looks more homogeneous than the previous one you posted.
I dont think your choice of wall or SGS modelling should affect the averaging times substantially. |
|
August 16, 2007, 08:28 |
Thank you Eugene,
I will ke
|
#48 |
Member
hadi tartoussi
Join Date: Mar 2009
Location: paris
Posts: 48
Rep Power: 17 |
Thank you Eugene,
I will keep averaging. I am wondering if it might be the initial conditions, maybe Ubulk is high. or i made something wrong somewhere in my simulation, although by looking to the instantaneous values of u_tau and yplus they are close to the mean value expected. I don't know if running a case in parallel can affect the results. I am using the same schemes as in channel395 tutorials. I will run channel395 to make sure that i am using channelOodles properly. Thank you for your advices, and if you have any other suggestions, i'll be glad to read it. cheers Hadi |
|
August 16, 2007, 09:27 |
Hi Eugene and Hadi,
I had r
|
#49 |
Member
diablo80@web.de
Join Date: Mar 2009
Posts: 93
Rep Power: 17 |
Hi Eugene and Hadi,
I had run some channel flow simulations using Smag model too. In my case, vanDriest damping made a lot of difference in the final Umean profile. I would say that you should use it. I got some asymmetries which took a long time to be averaged out too. Good luck, luiz |
|
August 16, 2007, 10:02 |
Thank you luiz,
Did u get s
|
#50 |
Member
hadi tartoussi
Join Date: Mar 2009
Location: paris
Posts: 48
Rep Power: 17 |
Thank you luiz,
Did u get symmetric fields with classic Smago without vanDriest damping? Regards Hadi |
|
August 16, 2007, 17:03 |
Yes, I did. After about 400 fl
|
#51 |
Member
diablo80@web.de
Join Date: Mar 2009
Posts: 93
Rep Power: 17 |
Yes, I did. After about 400 flow throw time, the symmetry errors were less than 2% (when normalized by the maximum value of the mean field in question).
I got very low "symmetry" errors for the Umean, this value (2%) was the worst case and corresponds to second order statistics, as expected. Cheers, luiz |
|
August 17, 2007, 09:56 |
Hi to all,
I just want to k
|
#52 |
Member
hadi tartoussi
Join Date: Mar 2009
Location: paris
Posts: 48
Rep Power: 17 |
Hi to all,
I just want to know after how many flow through time should i get symmetric field in channel395. Thanks in advance Hadi |
|
August 21, 2007, 10:02 |
Hi Hadi,
I dont have this inf
|
#53 |
Member
diablo80@web.de
Join Date: Mar 2009
Posts: 93
Rep Power: 17 |
Hi Hadi,
I dont have this information. I deleted my intermediate time step directories. I can only tell you that 400 flow throw time should be enough Good luck, luiz |
|
September 5, 2007, 07:10 |
Hi to all,
I have another q
|
#54 |
Member
hadi tartoussi
Join Date: Mar 2009
Location: paris
Posts: 48
Rep Power: 17 |
Hi to all,
I have another question, when i restart averaging do u think i should, set nuSgs boundary condition to zeroGradient, or i can keep the lastest values? Because i was wondering if it might affect the results. Do u think that rms values should be symmetric even if the values are not converged yet, since the fields are collapsed in streamwise and spanwise directions ? Thanks in advance Hadi |
|
September 12, 2007, 05:15 |
Hi to all,
I am sorry for b
|
#55 |
Member
hadi tartoussi
Join Date: Mar 2009
Location: paris
Posts: 48
Rep Power: 17 |
Hi to all,
I am sorry for being annoying lately, as i mentioned in my previous messages that i was not getting symmetric fields using classic smagorinsky, even after running the simulation for more than 350 flows through time, so i restarted averaging from the last saved time, and i got symmetric profile for urms, Uf and K after less than 100 flows through time, and my utau mean is only 6% off from the DNS value. I am running also smagorinsky+vanDriest damping, and even after more than 150 flows through time Uf profile is not symmetric at the walls, the only reason i can see for this unstable results in the 3 cases is the initial conditions, for my understanding i should start averaging from the time when the flow becomes established. Can anyone tell me how to know in openFoam whether the flow is established or not? Cheers Hadi |
|
September 27, 2007, 16:19 |
Is this a philosophical questi
|
#56 |
New Member
Ian Cowan
Join Date: Mar 2009
Location: London, UK
Posts: 28
Rep Power: 17 |
Is this a philosophical question, or am I reading too much into it?
As in any transient simulation, the way to tell whether the "flow is established" is to monitor the stability of statistics - monitored values at a point, over a plane, over the whole volume. Try any/all of these. If there is a stable state, then these will eventually settle down, or at least stop heading off in a particular direction. Good luck. |
|
March 13, 2008, 13:59 |
Hi
I wanna run channelOodle
|
#57 |
New Member
OpenFOAM Newbie
Join Date: Mar 2009
Posts: 20
Rep Power: 17 |
Hi
I wanna run channelOodles with several walls inside the channel. Is it possible to run this case in OpenFOAM? Which open source software would be best to mesh the case, gmsh, salome etc. |
|
March 17, 2008, 13:05 |
Hi everybody,
I have a vagu
|
#58 |
New Member
Jojo
Join Date: Mar 2009
Posts: 7
Rep Power: 17 |
Hi everybody,
I have a vague memory about a manner of imposing turbulent BC for LES using something very close to channelOodles. It was like creating a `backyard' domain behind the inlet. This backyard was semi-periodic: it ran like an infinite pipe (or any type of section geometry) flow to develop turbulence but injected this turbulent flow into the main domain. I have diged into FvPatchField where the BC are defined but without success. I remember the backyard domain is identified by OF between two plans, acting as those semi-periodic BC. Has anybody located the files, or a tutorial using them? How the targeted mean flow rate is guaranteed? Thanks a lot in advance. |
|
March 18, 2008, 03:41 |
Hi Jojo,
I think that what
|
#59 |
Senior Member
Cedric DUPRAT
Join Date: Mar 2009
Location: Nantes, France
Posts: 195
Rep Power: 17 |
Hi Jojo,
I think that what you are looking for is here : http://www.cfd-online.com/OpenFOAM_D...es/1/5685.html and for a tutorial, maybe this one should be enough : http://www.cfd-online.com/OpenFOAM_D...es/1/4198.html Enjoy, Cedric |
|
March 18, 2008, 11:52 |
Many thanks Cedric for your qu
|
#60 |
New Member
Jojo
Join Date: Mar 2009
Posts: 7
Rep Power: 17 |
Many thanks Cedric for your quick answer. It looks like what I meant and I ll dig into.
I did not search the good post to ask my question as, in my mind, doing this type of inlet bc was like sticking a channel flow to the main domain, hence I posted here. |
|
Thread Tools | Search this Thread |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Burgerbs equation non constant Boundary Conditions Initial Conditions | arkangel | OpenFOAM Running, Solving & CFD | 1 | October 2, 2008 15:48 |
Initial conditions | Shuo | Main CFD Forum | 2 | July 27, 2007 09:57 |
Initial conditions = final conditions | Chucho | CFX | 5 | December 16, 2005 18:14 |
Initial conditions | Allan | CFX | 5 | April 23, 2002 09:54 |
Initial conditions in CFX 5.5 | Astrid | CFX | 3 | December 19, 2001 00:24 |