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

Residuals never converge!

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

Reply
 
LinkBack Thread Tools Display Modes
Old   March 11, 2015, 05:30
Default Residuals never converge!
  #1
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
Hello friend,
I'm a chemical engineer and I'm trying to model the following problem:
air firstly flowing in a pipe duct and then , when it exits from the duct, it flows around a cylinder. This is the scheme:
This is the mesh I've done, simplifying the problem for its geometry as axysimmetryc :




The walls of the pipe duct are adiabatic. Air has got a temperature of 373 K while the whole system has got a temperature of 298 K . The boundary conditions are the following:
PRESSURE OUTLET for the external part of the domain
for the air entering the duct: VELOCITY INLET, with the relative magnitude and axial component of the flow direction =1, the radial 0. Furthermore the Inlet temperature is 373 K.
The walls of the duct are ADIABATIC, while the cylinder at the exit of the pipe duct has got a HEAT FLUX coupled with the air.
The problem is that as soon as the simulation starts, the residual oscillate a lot also when FLUENT reaches 100000 iteration. I set the problem as unsteady, with 0.01 as time step size . The oscillation are these:
What I can do to reach the convergence?? Thanks to everybody! This simulation is my nightmare!!
Attached Images
File Type: jpg scheme.jpg (7.7 KB, 4 views)
Giulia1987 is offline   Reply With Quote

Old   March 11, 2015, 06:37
Default
  #2
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 908
Rep Power: 15
ghost82 will become famous soon enough
Hi Giulia,
I think there's nothing wrong in convergence: when you set an unsteady simulation, convergence is reached for each time step: you can see from your residual chart; at the beginning of each timestep you have "high" residual value, then during iteration residual values decrease and reach convergence for that timestep; this is repeated for each timestep.
__________________
Google is your friend and the same for the search button!
ghost82 is online now   Reply With Quote

Old   March 11, 2015, 06:38
Default
  #3
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
PS : what I've noticed is that leaving properties such as density,viscosity and specific heat not in polynomial form (a0 + a1T + a2 T^2......) but costant (with the default vaules in FLUENT) the residual seems not to oscillate but they decrease...

but If there is a heat exchange between air at 373 K and the pipe duct and the cylinder at 298 K it is a mistake leaving thermal properties constant?
Giulia1987 is offline   Reply With Quote

Old   March 11, 2015, 06:40
Default
  #4
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 908
Rep Power: 15
ghost82 will become famous soon enough
If you noticed "flat" residuals in an unsteady simulation probably it means you reached a steady state solution.
__________________
Google is your friend and the same for the search button!
ghost82 is online now   Reply With Quote

Old   March 11, 2015, 06:55
Default
  #5
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
Quote:
Originally Posted by ghost82 View Post
Hi Giulia,
I think there's nothing wrong in convergence: when you set an unsteady simulation, convergence is reached for each time step: you can see from your residual chart; at the beginning of each timestep you have "high" residual value, then during iteration residual values decrease and reach convergence for that timestep; this is repeated for each timestep.
Excuse me but this is the first time I do an unsteady simulation: why this is repeated for each timestep? So how I can realize if I have reached convergence? Maybe I have to use a smaller time step?
Giulia1987 is offline   Reply With Quote

Old   March 11, 2015, 07:11
Default
  #6
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 908
Rep Power: 15
ghost82 will become famous soon enough
In a steady state simulation you have to reach convergence for the steady state; you can think the steady simulation as a single time step.
You have a single solution for the steady state.

In an unsteady simulation solution changes for each time step, because the solution changes vs. time.
You have multiple solutions, one for each time step, and you have to reach convergence for each time step.

See also attached picture.

Daniele
Attached Images
File Type: png residuals.png (20.9 KB, 10 views)
__________________
Google is your friend and the same for the search button!

Last edited by ghost82; March 11, 2015 at 09:31.
ghost82 is online now   Reply With Quote

Old   March 11, 2015, 07:23
Default
  #7
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
Thank you! So,despite the oscillation, I reach the convergence when in a single time step residuals reach the value previously fixed , for example 10^-6 ?
Giulia1987 is offline   Reply With Quote

Old   March 11, 2015, 07:26
Default
  #8
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 908
Rep Power: 15
ghost82 will become famous soon enough
Correct
Oscillation is quite usual in unsteady simulations: this means that for each time step the solution is changing with time, so the problem is unsteady.
If after some time residuals will be "flat" (horizontal) that means that solution is not changing anymore with time, and you reached a steady state solution by simulating the unsteady problem.
__________________
Google is your friend and the same for the search button!

Last edited by ghost82; March 11, 2015 at 09:32.
ghost82 is online now   Reply With Quote

Old   March 11, 2015, 07:28
Default
  #9
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
Thank you!!
Giulia1987 is offline   Reply With Quote

Old   March 11, 2015, 10:49
Default
  #10
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
But how to evaluate convergence? For example monitoring coefficien drag value and determinating the time when it reaches a constant value?
Giulia1987 is offline   Reply With Quote

Old   March 11, 2015, 10:52
Default
  #11
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 908
Rep Power: 15
ghost82 will become famous soon enough
Yes, for example.
Use the same criteria as the steady state approach: only difference is that your convergence criteria must be reached at each time step.
__________________
Google is your friend and the same for the search button!
ghost82 is online now   Reply With Quote

Old   March 12, 2015, 14:25
Default
  #12
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
another question: the walls of the pipe duct are adiabatic , but in fluent in addiction to the wall of the duct there is its shadow too. So, how to assignate the adiabaticity boundary condition? Both wall and shadow wall are adiabatic or the wall is adiabatic while for its shadow I must set coupled boundary condition????
Giulia1987 is offline   Reply With Quote

Old   March 12, 2015, 14:35
Default
  #13
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 908
Rep Power: 15
ghost82 will become famous soon enough
Set both walls to adiabatic (it should be automatically set for the shadow when you set for wall, however check both walls and set them to zero heat flux).
__________________
Google is your friend and the same for the search button!
ghost82 is online now   Reply With Quote

Old   March 12, 2015, 14:37
Default
  #14
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
thank you!!
Giulia1987 is offline   Reply With Quote

Old   March 14, 2015, 08:20
Unhappy
  #15
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
Hello friends,
my problem now is that when I'm running the simulation Fluent sends an error messages: "reversed flow in faces XXX on outflow 4". Here's a picture of the mesh for the problem and the boundary condition..



This is velocity profile at the current time of the simulation:


and this is pressure profile...

and temperature:

So, what cause this reverse flow??
Giulia1987 is offline   Reply With Quote

Old   March 14, 2015, 11:04
Default
  #16
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 908
Rep Power: 15
ghost82 will become famous soon enough
If I were you I didn't care too much for this warning.
Reversed flow means that fluid is flowing in your domain from your outflow: this could be related to recirculation, which can be quite common.
This warning could also disappear after some time steps.
Just make sure to correctly set your backflow boundary condition (temperature for example): these are estimates you set in the boundary condition panel.
Moreover, if you are interested in the flow near the cylinder, if your outlet is far enough, what happens at the outlet should not interfere too much there.
__________________
Google is your friend and the same for the search button!
ghost82 is online now   Reply With Quote

Old   March 14, 2015, 11:13
Default
  #17
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
I'm sorry: what do you mean for backflow boundary condition?
Giulia1987 is offline   Reply With Quote

Old   March 14, 2015, 11:18
Default
  #18
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 908
Rep Power: 15
ghost82 will become famous soon enough
Here is an example of backflow total temperature panel (it is an outlet boundary condition): this means that if fluid enters the domain from the outlet (reversed-flow) it will enter with a temperature of 288 K.

http://www.arc.vt.edu/ansys_help/flu...n_pressout.png
__________________
Google is your friend and the same for the search button!
ghost82 is online now   Reply With Quote

Old   March 14, 2015, 11:22
Default
  #19
Senior Member
 
ghost82's Avatar
 
Daniele
Join Date: Oct 2010
Location: Italy
Posts: 908
Rep Power: 15
ghost82 will become famous soon enough
Just one thing: I see now that you set an outflow boundary condition: outflow means that the fluid is fully developed: are you sure is it fully developed?
If you have recirculations, and these recirculations are real, then your fluid is not fully developed and you should set a pressure outlet boundary condition.
__________________
Google is your friend and the same for the search button!
ghost82 is online now   Reply With Quote

Old   March 14, 2015, 11:44
Default
  #20
New Member
 
Giulia S.
Join Date: Mar 2015
Posts: 12
Rep Power: 2
Giulia1987 is on a distinguished road
yesterday i put a pressure outlet as a boundary condition but my prof. suggested me to change into outflow... but also pressure outlet gave me a reversed flow..
Giulia1987 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
Continuity convergence issue on a 3d wing Nick R FLUENT 18 February 4, 2015 12:21
motorBike Residuals for SST k-omega... and mine JR22 OpenFOAM Running, Solving & CFD 6 August 1, 2013 09:08
judging convergence through residuals MachZero Main CFD Forum 7 December 25, 2012 13:18
Residuals converge then explode plm OpenFOAM Running, Solving & CFD 2 December 31, 2011 08:35
Convergence - scaled vs unscaled residuals HS FLUENT 1 November 7, 2005 06:45


All times are GMT -4. The time now is 08:19.