CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > FLUENT

different results between serial solver and parallel solver

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

Like Tree1Likes
  • 1 Post By wlt_1985

Reply
 
LinkBack Thread Tools Display Modes
Old   April 30, 2011, 05:09
Default different results between serial solver and parallel solver
  #1
Senior Member
 
raymond
Join Date: Nov 2009
Posts: 149
Rep Power: 6
wlt_1985 is on a distinguished road
I run my model with serial solver and parallel solver respectively. I found that the results especially the velocity does have large difference in power of magnitude. For example, serial solver 2m/s but parallel solver 20k m/s. What is wrong with the parallel solver? In addition, turbulent viscosity limited to viscosity ration does appear in parallel solver but none in serial solver. Pls help.

Thanks a lot
stuart23 likes this.
wlt_1985 is offline   Reply With Quote

Old   May 1, 2011, 17:14
Default
  #2
New Member
 
Join Date: Jan 2011
Posts: 25
Rep Power: 5
zhenglun.wei is on a distinguished road
I also found some difference between the serial and parallel version, but not that severe. Hope you'll find the solution asap.
zhenglun.wei is offline   Reply With Quote

Old   May 2, 2011, 16:46
Default
  #3
Senior Member
 
raymond
Join Date: Nov 2009
Posts: 149
Rep Power: 6
wlt_1985 is on a distinguished road
Quote:
Originally Posted by zhenglun.wei View Post
I also found some difference between the serial and parallel version, but not that severe. Hope you'll find the solution asap.
Thansk for care.
wlt_1985 is offline   Reply With Quote

Old   May 2, 2011, 21:18
Default
  #4
Senior Member
 
Raashid Baig
Join Date: Mar 2010
Location: Bangalore, India
Posts: 136
Rep Power: 6
cfd_newbie is on a distinguished road
Quote:
Originally Posted by wlt_1985 View Post
I run my model with serial solver and parallel solver respectively. I found that the results especially the velocity does have large difference in power of magnitude. For example, serial solver 2m/s but parallel solver 20k m/s. What is wrong with the parallel solver? In addition, turbulent viscosity limited to viscosity ration does appear in parallel solver but none in serial solver. Pls help.

Thanks a lot
Hi Raymond,
Are you sure you have inputed the exact same initial and boundary conditions in both the cases ? If the results are still different then you should report the problem to Fluent. They have a dedicated set of people who look into this issue and they might give you a patch to fix this problem.
Raashid
cfd_newbie is offline   Reply With Quote

Old   May 4, 2011, 08:50
Default
  #5
Senior Member
 
raymond
Join Date: Nov 2009
Posts: 149
Rep Power: 6
wlt_1985 is on a distinguished road
Quote:
Originally Posted by cfd_newbie View Post
Hi Raymond,
Are you sure you have inputed the exact same initial and boundary conditions in both the cases ? If the results are still different then you should report the problem to Fluent. They have a dedicated set of people who look into this issue and they might give you a patch to fix this problem.
Raashid

Dear Raashid,

I have found the causes of the problem. There were no much differences in FINAL RESULTS for both cases. Since the calculation for parallel solver was much faster than the serial solver, hence the WRONG RESULTS due to the model setting have been shown earlier than the serial soler. These errors were due to the TURBULENT VISCOSITY RATIO LIMITATION during iteration. Initially, the limit of turbulent viscosity ratio is set to 1e+05 and there was no such warning of limitation after the initiation has been activated. However, after hundred iterations, this warning of limitation has been shown and the velocity has suddenly been increased after that particualr moment. I have gone through few posts in this forum. These errors were due to poor mesh and they could be solved by having grid adaption. I tried to have grid adaption. However, the residuals of continuity was keeping increasing and going to diverge. I have no idea on solving my problem. Can please help me for my problem? Thanks for help.

Regards
Raymond
wlt_1985 is offline   Reply With Quote

Old   May 4, 2011, 21:01
Default
  #6
Senior Member
 
Raashid Baig
Join Date: Mar 2010
Location: Bangalore, India
Posts: 136
Rep Power: 6
cfd_newbie is on a distinguished road
Quote:
Originally Posted by wlt_1985 View Post
Dear Raashid,

I have found the causes of the problem. There were no much differences in FINAL RESULTS for both cases. Since the calculation for parallel solver was much faster than the serial solver, hence the WRONG RESULTS due to the model setting have been shown earlier than the serial soler. These errors were due to the TURBULENT VISCOSITY RATIO LIMITATION during iteration. Initially, the limit of turbulent viscosity ratio is set to 1e+05 and there was no such warning of limitation after the initiation has been activated. However, after hundred iterations, this warning of limitation has been shown and the velocity has suddenly been increased after that particualr moment. I have gone through few posts in this forum. These errors were due to poor mesh and they could be solved by having grid adaption. I tried to have grid adaption. However, the residuals of continuity was keeping increasing and going to diverge. I have no idea on solving my problem. Can please help me for my problem? Thanks for help.

Regards
Raymond
Hi Raymond,
The gird adaptation technique may or may not work depending on problem to problem. According to my experience (more experienced Fluent users can correct me), mesh adaptation can work well for heat transfer kind of problems but for external aerodynamics problems you should have a good grid to start your problem from the beginning.
Raashid
cfd_newbie is offline   Reply With Quote

Old   May 9, 2011, 08:06
Default
  #7
New Member
 
litter
Join Date: May 2011
Posts: 17
Rep Power: 5
litterfree is on a distinguished road
the unit for the two cases are same?
__________________
never ever give up
litterfree is offline   Reply With Quote

Old   May 9, 2011, 11:21
Default
  #8
Senior Member
 
raymond
Join Date: Nov 2009
Posts: 149
Rep Power: 6
wlt_1985 is on a distinguished road
Quote:
Originally Posted by cfd_newbie View Post
Hi Raymond,
The gird adaptation technique may or may not work depending on problem to problem. According to my experience (more experienced Fluent users can correct me), mesh adaptation can work well for heat transfer kind of problems but for external aerodynamics problems you should have a good grid to start your problem from the beginning.
Raashid
Dear Raashid,

Thanks a lot. I am trying to do it with better grid from the beginning.

Best regards
wlt_1985 is offline   Reply With Quote

Old   May 9, 2011, 11:24
Default
  #9
Senior Member
 
raymond
Join Date: Nov 2009
Posts: 149
Rep Power: 6
wlt_1985 is on a distinguished road
Quote:
Originally Posted by litterfree View Post
the unit for the two cases are same?
Dear Litterfree,

Yes. Both cases are same units.

Best regards
wlt_1985 is offline   Reply With Quote

Old   May 12, 2011, 16:36
Default
  #10
Member
 
Casey
Join Date: Jun 2009
Posts: 92
Rep Power: 7
meangreen is on a distinguished road
Make sure your choice of partitions is not affecting your solutions.
meangreen is offline   Reply With Quote

Old   April 11, 2012, 15:25
Default
  #11
Member
 
nazareno mancinelli
Join Date: Mar 2009
Location: argentina
Posts: 35
Rep Power: 7
yochule is on a distinguished road
dear raymond, i hope you are still in the forum.
i'm havin some problems with local parallel and dynamic gradient adaption. my problem it's that when i set up the gradient adaption scheme, the parallel computing stop working and the next iterations became in serial... so the solutions take so much time. can you tell me if you did found the same issue?
i'm runing in gnu/linux ubuntu amd64, whith the command: ./fluent 2ddp -t4 -pnet
you can check my post here: local parallel - dynamic gradient adaption

hope you can help me.
thanks a lot (and excuse me to get into your post)
yochule is offline   Reply With Quote

Reply

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
"Define_Profile" UDF for parallel solver Antoine Fluent UDF and Scheme Programming 7 February 26, 2013 04:22
Working directory via command line Luiz CFX 4 March 6, 2011 20:02
Solver crashes (PVM parallel) Marco Müller CFX 2 February 4, 2009 02:51
Differences between serial and parallel runs carsten OpenFOAM Bugs 11 September 12, 2008 11:16
Help: Serial code to parallel but even slower Zonexo Main CFD Forum 4 May 14, 2008 10:26


All times are GMT -4. The time now is 13:37.