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

Monitoring residuals , is it safe ?

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

Like Tree13Likes

Reply
 
LinkBack Thread Tools Display Modes
Old   July 13, 2017, 10:26
Default Monitoring residuals , is it safe ?
  #1
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
Sponsored Links
Hi, i read that monitoring only the residuals can be treaky, and we should monitoring also values on interest, in my case velocity and temperature. By monitoring those values, how can we say that the solution is converged ? When the values dont change any more or when it becomes close to experimental results ? Or what ? Also, what are the acceptable values of residuals that can be published for continuity, velocity and energy
medaouarwalid is offline   Reply With Quote
Sponsored Links

Old   July 13, 2017, 10:37
Default
  #2
Senior Member
 
Join Date: Aug 2014
Location: N/A
Posts: 156
Rep Power: 5
BlnPhoenix is on a distinguished road
Transient or steady state simulation?
medaouarwalid likes this.
BlnPhoenix is offline   Reply With Quote

Old   July 13, 2017, 10:46
Default
  #3
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
Transient LES

Sent from my F1f using CFD Online Forum mobile app
medaouarwalid is offline   Reply With Quote

Old   July 13, 2017, 11:01
Default
  #4
Senior Member
 
Join Date: Aug 2014
Location: N/A
Posts: 156
Rep Power: 5
BlnPhoenix is on a distinguished road
You need to define what convergence means to you.

e.g. if you monitor temperature you can define convergence as: the calculated solution for temperature in each time step changes only in the second decimal place behind the comma. You need to get a feeling how exact your solutions per time step need to be. If you study convergence levels you will see that the change in temperature in each time step correlates to the level the residuals reach. So you can set automatic convergence detection if you know let's say for example that 1e-4 in residuals for energy translates to a max change in temperature in the xyz place behind the comma. You have to find that out for yourself.
medaouarwalid likes this.
BlnPhoenix is offline   Reply With Quote

Old   July 13, 2017, 11:07
Default
  #5
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
But when i took the mesures of temperature in experiance, i didnt took it with a fix time step so i have no idea how much it must be changing each timstep, althought i have an idea how much it changes every z/d , because i took temperature mesurements with velocity, each time i mesured velocity i mesured also temperature. I hope you understood me

Sent from my F1f using CFD Online Forum mobile app
medaouarwalid is offline   Reply With Quote

Old   July 13, 2017, 11:08
Default
  #6
Senior Member
 
Join Date: Aug 2014
Location: N/A
Posts: 156
Rep Power: 5
BlnPhoenix is on a distinguished road
Also for continuity. I usually accept a mass imbalance of < 0.1% as acceptable. Again this translates into a value for continuity residuals. Maybe something like 1e-4 or less. You need to check this.
BlnPhoenix is offline   Reply With Quote

Old   July 13, 2017, 11:11
Default
  #7
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
Quote:
Originally Posted by BlnPhoenix View Post
Also for continuity. I usually accept a mass imbalance of < 0.1% as acceptable. Again this translates into a value for continuity residuals. Maybe something like 1e-4 or less. You need to check this.
I am doing an example wright now, all residuals are less than 10^-5 exept for continuity it dident reach 10^-4 i dont know why

Sent from my F1f using CFD Online Forum mobile app
medaouarwalid is offline   Reply With Quote

Old   July 13, 2017, 11:15
Default
  #8
Senior Member
 
Join Date: Aug 2014
Location: N/A
Posts: 156
Rep Power: 5
BlnPhoenix is on a distinguished road
Quote:
Originally Posted by medaouarwalid View Post
But when i took the mesures of temperature in experiance, i didnt took it with a fix time step so i have no idea how much it must be changing each timstep, althought i have an idea how much it changes every z/d , because i took temperature mesurements with velocity, each time i mesured velocity i mesured also temperature. I hope you understood me

Sent from my F1f using CFD Online Forum mobile app
How long is your experimental time? When did you measure and how often?
You can simulate until maybe you first measured your temperature in experiment. If you measured first after 5 mins. simulate 5 mins. Also how exact is you measurement device? If you now the uncertainty in your measurements, you do need to be more exact in simulation than your measurement device.
BlnPhoenix is offline   Reply With Quote

Old   July 13, 2017, 11:17
Default
  #9
Senior Member
 
Join Date: Aug 2014
Location: N/A
Posts: 156
Rep Power: 5
BlnPhoenix is on a distinguished road
Quote:
Originally Posted by medaouarwalid View Post
I am doing an example wright now, all residuals are less than 10^-5 exept for continuity it dident reach 10^-4 i dont know why

Sent from my F1f using CFD Online Forum mobile app
Check mass imbalance in your system, if your satisified. It sounds OK to me.
medaouarwalid likes this.
BlnPhoenix is offline   Reply With Quote

Old   July 13, 2017, 11:20
Default
  #10
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
Time of experimental was about an hour Because i took the velocities in 20 axial stations and for each axial station a go for 10 stations radially . I dont think that i can do a 1 hour of simulation because i am setting the time step = 10^-5 . If i can reac 10 seconds of simulation i am lucky 😥
The device is velocicalc
[IMG]lghttp.17652.nexcesscdn.net/808B15/magento/media/catalog/product/cache/8/image/500x500/ae8e27adf667b20d5f3db4b7e3a38cc2/9/5/9555_app.jpg[/IMG]
Sent from my F1f using CFD Online Forum mobile app
medaouarwalid is offline   Reply With Quote

Old   July 13, 2017, 11:20
Default
  #11
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
Quote:
Originally Posted by BlnPhoenix View Post
Check mass imbalance in your system, if your satisified. It sounds OK to me.
What mass imbalance means ?

Sent from my F1f using CFD Online Forum mobile app
medaouarwalid is offline   Reply With Quote

Old   July 13, 2017, 11:22
Default
  #12
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 3,415
Rep Power: 39
FMDenaro will become famous soon enoughFMDenaro will become famous soon enough
Since you are doing an LES, what you call residuals are only the terms you have in each time step in the formulation for solving the system of the equations. Now that depends on the choice of the formulation, are you using a full compressible flow solver? In such a case what about the Mach number? Are the residuals normalized?
medaouarwalid likes this.
FMDenaro is offline   Reply With Quote

Old   July 13, 2017, 11:29
Default
  #13
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
Since you are doing an LES, what you call residuals are only the terms you have in each time step in the formulation for solving the system of the equations. Now that depends on the choice of the formulation, are you using a full compressible flow solver? In such a case what about the Mach number? Are the residuals normalized?
I am suppozing that the flow is incompressible because the velocity inlet is lot less than speed of sound. (6.66 m/s) so much number is 0.02
Residuals are not normalized

Sent from my F1f using CFD Online Forum mobile app
medaouarwalid is offline   Reply With Quote

Old   July 13, 2017, 11:47
Default
  #14
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 3,415
Rep Power: 39
FMDenaro will become famous soon enoughFMDenaro will become famous soon enough
Your case will be stiff using a compressible flow model. Are you preconditioning the equations? Could you switch to a pure incompressible solver? Try to evaluate the normalized residuals
FMDenaro is offline   Reply With Quote

Old   July 13, 2017, 11:53
Default
  #15
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
Your case will be stiff using a compressible flow model. Are you preconditioning the equations? Could you switch to a pure incompressible solver? Try to evaluate the normalized residuals
Excuse me but I didn't understand .

Sent from my F1f using CFD Online Forum mobile app
medaouarwalid is offline   Reply With Quote

Old   July 13, 2017, 12:21
Default
  #16
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 3,415
Rep Power: 39
FMDenaro will become famous soon enoughFMDenaro will become famous soon enough
Quote:
Originally Posted by medaouarwalid View Post
Excuse me but I didn't understand .

Sent from my F1f using CFD Online Forum mobile app

I think I cannot help you if you are not aware about these issues? What about your background in CFD?
FMDenaro is offline   Reply With Quote

Old   July 13, 2017, 12:35
Default
  #17
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
I think I cannot help you if you are not aware about these issues? What about your background in CFD?
Not very good, everdy day i learn something new, what should i learn now?

Sent from my F1f using CFD Online Forum mobile app
medaouarwalid is offline   Reply With Quote

Old   July 13, 2017, 12:46
Default
  #18
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 3,415
Rep Power: 39
FMDenaro will become famous soon enoughFMDenaro will become famous soon enough
Quote:
Originally Posted by medaouarwalid View Post
Not very good, everdy day i learn something new, what should i learn now?

Sent from my F1f using CFD Online Forum mobile app

You should attend a couple of course in fluid mechanics and CFD ... there are a lot of CFD textbooks, each one has some peculiarity and some issues more detalined than others. However, it requires some effort and some years to be good in CFD....
FMDenaro is offline   Reply With Quote

Old   July 13, 2017, 13:19
Default
  #19
Senior Member
 
dilaw meda
Join Date: Jun 2017
Location: algeria
Posts: 128
Rep Power: 2
medaouarwalid is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
You should attend a couple of course in fluid mechanics and CFD ... there are a lot of CFD textbooks, each one has some peculiarity and some issues more detalined than others. However, it requires some effort and some years to be good in CFD....
I stied and i am studying but of corse i can not know every thing. thats why i ask lot of​ questions, like this i am progressing every day. I have a simulation to do , i can not wait until i read every thing, i only have to understand things that have relation with my case, and with time i will learn more and more about other things

Sent from my F1f using CFD Online Forum mobile app
medaouarwalid is offline   Reply With Quote

Old   July 13, 2017, 13:59
Default
  #20
Senior Member
 
Join Date: Aug 2014
Location: N/A
Posts: 156
Rep Power: 5
BlnPhoenix is on a distinguished road
Quote:
Originally Posted by medaouarwalid View Post
Time of experimental was about an hour Because i took the velocities in 20 axial stations and for each axial station a go for 10 stations radially . I dont think that i can do a 1 hour of simulation because i am setting the time step = 10^-5 . If i can reac 10 seconds of simulation i am lucky 😥
The device is velocicalc
[IMG]lghttp.17652.nexcesscdn.net/808B15/magento/media/catalog/product/cache/8/image/500x500/ae8e27adf667b20d5f3db4b7e3a38cc2/9/5/9555_app.jpg[/IMG]
Sent from my F1f using CFD Online Forum mobile app

So did you find some average temperature in your experiment? Have you measured exactly after one hour? Try to describe your experimental findings. Is the temperature constant over time, if yes what do you expect if you would measure after five minutes? Same temp as after one hour?

If you cannot afford the simulation to run to one hour, i would concider redoing your measurements and run it till you have experimental data to compare.

Mass imbalance means that the mass flow in the systems equals the mass flow out of the system. It's very important to ensure mass balance in every time step. So please check for that and as mentioned set your target residual values accordingly. Also for temperature and velocities as i described above. Your residuals sound good but try to look for the actual change in values to see what they actually mean.
BlnPhoenix 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
ANSYS Workbench on "Uncertified" Linux Distros hsr CFX 264 April 27, 2017 11:46
Monitoring Residuals - Initial or Final? BigPapi34 OpenFOAM Running, Solving & CFD 1 November 3, 2014 06:31
Monitoring Residuals - Initial or Final? BigPapi34 Main CFD Forum 0 November 3, 2014 06:23
Monitoring residuals in Workbench-Fluent using Remote Solver fadiga Main CFD Forum 2 October 16, 2013 15:43
judging convergence through residuals MachZero Main CFD Forum 7 December 25, 2012 13:18

Sponsored Links


All times are GMT -4. The time now is 18:33.