CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > OpenFOAM Running, Solving & CFD

rhoCentralFoam for channel flow

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

Reply
 
LinkBack Thread Tools Display Modes
Old   June 6, 2014, 20:25
Default
  #21
Member
 
Fluid Dynamics
Join Date: Mar 2013
Posts: 41
Rep Power: 5
cfd.with.openfoam is on a distinguished road
Hi Felipe,

Hope you are doing well.

Sorry for such a long delay but too many things are happening -> other than work .

Some questions -
  • How much computational power (both for rhoCentral and rhoPimple) did it take for you to get through your finest case i.e 256*256*256? e.g 5000 hours. I saw that you only used 48 decompositions for your case so I don't think it would have taken more than 10000 hours? I think pimple would have taken around 10 times less than central?
  • What was the delta x+, y+ and z+ for your finest case. Since you said rhoCentralFoam is too dissipative I am guessing you were referring to these values. Wall y+ as we know should be <1 but I am very curious about your x+ and z+ values. I will try to estimate them using a RANS calc. as well but your answer will help.
  • Did you try your cases where you got compression waves with the Gamma interpolation scheme? I am aware that you got rid of your issues by playing with the mesh and not by changing the numerical schemes.
  • Did you try your cases with the Euler time scheme. I have heard that LES should not be done with Euler ever and I am curious (if we can get away with it or not) because rhoCentral behaves very strangely with backward e.g the solutions are not smooth and sometimes there is a network of waves like you saw.
  • Favre or Reynolds avgd. --- I am not sure if rhoCentral writes a favre avgd. field in the time directories (e.g U) or it writes reynolds avgd. fields. Any idea there?
Thank you very much. I know I am asking too much here but hopefully you won't mind.

Awaiting your response.
cfd.with.openfoam is offline   Reply With Quote

Old   June 8, 2014, 15:51
Default
  #22
Member
 
Felipe Portela
Join Date: Dec 2012
Location: London
Posts: 64
Rep Power: 5
fportela is on a distinguished road
Hi there,

I haven't looked at this in a long time so I\ll have to try to answer your questions from (not so reliable) memory...
  • I only ran the 256^3 case once or twice (I was more interesting in the coarser cases). If I recall correctly, for the finest grid I actually used 96 processors for about a week, so something in the order 11k processing hours. I never really tried this case on rhoPimpleFoam so I can't comment on how it performs...
  • I just did a quick check (I hope I didn't get my maths wrong :P ) and for the 128^3 case I would expect the y+ of the first cell to be smaller than 1 (with a uniform grid it would be about 4, but since I applied stretching the number should go down). From this you can just check the domain dimensions from the Coleman paper. I'm quite confident the issue here is not resolution as I got near-DNS results with the other solvers I used (even in coarser grids).
  • I do not recall if I change the interpolation scheme, sorry...
  • I'm not sure if using a first order scheme would be interesting here as we are interested in good resolution (in time and space). Maybe the first order scheme would kill the strange waves, but it could also kill some (physical) features of the flow.
  • I'm pretty sure that the means are Reynolds' averaged. I think the routine that calculates statistics is solver-independent so it should simply take the running average of the fields.

Sorry I can't be more detailed in my answers, but I haven't really thought about this problem in almost a year and I'm not even using OpenFOAM atm x)

Cheers,
Felipe

Quote:
Originally Posted by cfd.with.openfoam View Post
Hi Felipe,

Hope you are doing well.

Sorry for such a long delay but too many things are happening -> other than work .

Some questions -
  • How much computational power (both for rhoCentral and rhoPimple) did it take for you to get through your finest case i.e 256*256*256? e.g 5000 hours. I saw that you only used 48 decompositions for your case so I don't think it would have taken more than 10000 hours? I think pimple would have taken around 10 times less than central?
  • What was the delta x+, y+ and z+ for your finest case. Since you said rhoCentralFoam is too dissipative I am guessing you were referring to these values. Wall y+ as we know should be <1 but I am very curious about your x+ and z+ values. I will try to estimate them using a RANS calc. as well but your answer will help.
  • Did you try your cases where you got compression waves with the Gamma interpolation scheme? I am aware that you got rid of your issues by playing with the mesh and not by changing the numerical schemes.
  • Did you try your cases with the Euler time scheme. I have heard that LES should not be done with Euler ever and I am curious (if we can get away with it or not) because rhoCentral behaves very strangely with backward e.g the solutions are not smooth and sometimes there is a network of waves like you saw.
  • Favre or Reynolds avgd. --- I am not sure if rhoCentral writes a favre avgd. field in the time directories (e.g U) or it writes reynolds avgd. fields. Any idea there?
Thank you very much. I know I am asking too much here but hopefully you won't mind.

Awaiting your response.
fportela is offline   Reply With Quote

Old   June 10, 2014, 20:14
Default
  #23
Member
 
Fluid Dynamics
Join Date: Mar 2013
Posts: 41
Rep Power: 5
cfd.with.openfoam is on a distinguished road
Hi Felipe,

Thanks for your response.

Actually I decided to simulate an even more basic case first i.e the incompressible turbulent channel with Retau=395 - which has been extensively discussed here.

However I have a couple of questions and I think I might get an answer more quickly here from you. Once things are clear I can post everything on the link above so that everybody can benefit.

The questions are -
  • The DNS uRMS profiles plotted by everybody in this post differ from mine. Eugene's thesis also had different profiles for DNS uRMS, please see Fig. 5.4 on page 168. I accessed the DNS data from this link. I am not sure what's wrong - it is at the same reTau i.e 395.
  • I have attached my LES results and DNS profiles. The problem is that I am missing the subgrid components of shear stresses from my 1st, 2nd and 3rd subplots and I don't know how to get them. Basically I used the One Equation TKE LES model. So, I know that all the SGS stuff is connected to k but I don't know how. Can you guide me here? This question has been asked earlier e.g here, as well but there was no answer.
Awaiting your response.

Thanks
Attached Files
File Type: pdf LESvsDNS_Channel.pdf (36.9 KB, 40 views)
cfd.with.openfoam is offline   Reply With Quote

Reply

Tags
channel flow, channelflow les, compressible flow, compressible solver, rhocentralfoam

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
transient, impregnating flow problem fgommer FLUENT 0 February 29, 2012 17:10
Flow meter Design CD adapco Group Marketing CD-adapco 3 June 21, 2011 08:33
mass flow in is not equal to mass flow out saii CFX 2 September 18, 2009 08:07
potential flow vs. Euler flow curious ... Main CFD Forum 23 July 21, 2006 07:40
Plug Flow Franck Main CFD Forum 3 September 4, 2003 05:57


All times are GMT -4. The time now is 17:52.