|
[Sponsors] |
February 10, 2016, 01:39 |
Solution Converged But With Oscillations
|
#1 |
Senior Member
|
Dear all. Hope all are doing well. I have developed a model for counter flow double pipe heat exchanger with hot water inside and cold water outside. Even though all my values converge to a variable value less than 10^-4 within forty iterations I am getting oscillations from the start to end. I have checked the mesh. In fact when I changed the relevance center in mesh from medium to fine, only then I got convergence as my solution was not converging previously. Can anybody help how to fix this problem I went to the FAQs on solver related problems. It was stated to increase physical time step. Can anybody tell me how to get the residence time in CFX-Post by placing a streamline and looking at the Time variable on it to get the maximum value?" Also would this solve the problem. Would be extremely grateful. Thanks.
|
|
February 10, 2016, 02:03 |
|
#2 |
Member
Alex
Join Date: Feb 2016
Posts: 81
Rep Power: 10 |
Shomaz, what timescale option do you use and how do you calculate time step (if you use physical time step)?
And what's about maximum residuals comparing to RMS? |
|
February 10, 2016, 02:38 |
|
#3 |
Senior Member
|
Thanks for the reply Red Ember. I am using Physical Timescale of 2 [s] in Timescale Control in Fluid Timescale Control. For Solid Timescale Control I am using Solid Timescale as Auto Timescale. I am also attaching jpegs for both residuals for heat transfer. Would be grateful for your help. Thanks
|
|
February 10, 2016, 02:48 |
|
#4 |
Member
Alex
Join Date: Feb 2016
Posts: 81
Rep Power: 10 |
Both pics represent max residuals if my browser works correctly
Why 2 seconds, how you estimated it? |
|
February 10, 2016, 04:28 |
|
#5 |
Senior Member
|
Sorry about that. I have attached RMS residuals as well. I have two seconds as a guess. In your opinion what value should I take? Thanks.
|
|
February 10, 2016, 05:38 |
|
#6 |
Member
Alex
Join Date: Feb 2016
Posts: 81
Rep Power: 10 |
1. Due to max/rms reiduals grid doesn't provide local convergence problems. Mesh is adequate.
2. Physical timestep seems too big according to view of residuals plot. Try to set it PT = (pipe diameter)/(average velocity) 3. I was told that Solid Timescale should not be automatic, although I can't recommend its value right now. 4. What about domain imbalances and monitor points? |
|
February 10, 2016, 06:10 |
|
#7 |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,701
Rep Power: 143 |
For CHT simulations you really should use imbalances as a convergence criteria. I bet if you look at your imbalances they are nowhere near converged.
For solid time scale factor you can normally use very big numbers - 100, 1000, even more at times. As the solid heat transfer equations are very numerically stable they can generally handle a large amount of acceleration, and the will accelerate the convergence of the imbalances a great deal. |
|
February 13, 2016, 03:57 |
|
#8 |
Senior Member
|
Thanks a lot Glenn. You are right for CHT problems imbalances reflect a good convergence criteria and my imbalances are not converging. Red Ember what diameter should I use, inner pipe diameter or annulus equivalent diameter? I know its a lame question but just want to confirm shouldn't I use annulus equivalent diameter of pipe? Would be grateful for answers. Thanks.
|
|
February 13, 2016, 05:09 |
|
#9 |
Senior Member
|
Also isn't it stated to use a larger physical timestep in case of non convergence. What you are saying Red Ember is to use a small physical timestep. When I used physical timestep=(equivalent diameter)/(average velocity) and solid timescale as auto, heat transfer parameters were stable and straight wrt. accumulated timestep and did not converge within the full limit of iterations of 100. When I was using a value of 2 [s] for fluid at least values were converging even though with oscillations. I am confused what to do. Glenn and Red Ember please guide. Thanks.
|
|
February 13, 2016, 05:19 |
|
#10 |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,701
Rep Power: 143 |
I would not be too precious about setting the time step size for a steady state run. Red Ember's is one approach, there are others (such as average residence time, or a representative dimension divided by a representative velocity).
In my experience you start with one of these estimates, any one, it does not matter much - but then you watch the convergence. If it is converging nicely but slowly you increase the time step size. If it is erratic and jumping about you decrease the size. If it is doing periodic hops you increase it by a large amount (x10 or so). So you start with a rough estimate and you fine tune it from there. That is why the initial estimate does not matter much. By the way: you change it with edit run in progress so you do not have to restart. |
|
February 13, 2016, 05:34 |
|
#11 |
Senior Member
|
Thanks Glenn. I used equivalent diameter of pipe for cold domain in calculation of PT. Is that okay? and what do you mean by "you change it with edit run in progress so you do not have to restart"? What do I change and how? Thanks.
|
|
February 13, 2016, 05:44 |
|
#12 | |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,701
Rep Power: 143 |
Let me quote it:
Quote:
|
||
February 13, 2016, 05:50 |
|
#13 |
Senior Member
|
Thanks a lot Glenn. Understood what you mean't. I can edit solver control parameters while solver is running. I didn't know about this. That is great. So when I change a value while solver is running will the solver restart from start or start from that iteration where I changed solver control parameter? Also while watching convergence do I change physical timestep for fluid if solution is not converging, for solid or for both fluid and solid? I hope you understand what I mean. Thanks.
|
|
February 13, 2016, 05:59 |
|
#14 |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,701
Rep Power: 143 |
Try it and find out.
The solver keeps running with the changed parameter. It does not restart. |
|
February 13, 2016, 06:02 |
|
#15 |
Senior Member
|
Also no matter what I do "Mass and Momentum 1" for Cold water domain oscillate a lot but converge quickly no matter what PT I use for fluid. Even though my parameter of interest are of heat transfer I would like "Mass and Momentum 1" to also converge. That is also bothering me. If my parameters of interest are not changing does one need to change PT in "Convergence Control" or in "Equation Class" for that particular equation that is not converging? Would be grateful. Thanks.
|
|
February 14, 2016, 03:56 |
|
#16 |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,701
Rep Power: 143 |
This is an FAQ: http://www.cfd-online.com/Wiki/Ansys...gence_criteria
|
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Why the solution could only be converged in | aestas | FLUENT | 11 | October 20, 2015 04:31 |
solution not getting converged | shulnak | FLUENT | 0 | July 11, 2014 14:27 |
Naca 0012 (compressible and inviscid) flow convergence problem | bipulsaha | FLUENT | 1 | July 6, 2011 07:51 |
Could anybody help me see this error and give help | liugx212 | OpenFOAM Running, Solving & CFD | 3 | January 4, 2006 18:07 |
Wall functions | Abhijit Tilak | Main CFD Forum | 6 | February 5, 1999 01:16 |