Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

How to Improve Convergence in Natural Convection

Please login with a confirmed email address before reporting spam

I am wondering how I can improve convergence in my natural convection model. It is a simple model; an isothermal aluminum plate in water.

The model starts out fine but as the run continues (transient model), the plume begins to lose stability causing the problem to reduce its time stepping. I have tried re-meshing as much as my PC can handle (3 mil elements).
I am wondering what else I can do to solve this problem. Should I change tolerances, crosswind or streamline parameters. I am running out of ideas. Any help would be greatly appreciated ( as I have been stuck on this problem for several days now). I have attached the model (with no mesh or solution). Typically, the model solves the first 20-30 seconds, then the solution becomes very strange and results become unreasonable.


8 Replies Last Post Apr 26, 2013, 1:50 a.m. EDT

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Feb 24, 2013, 5:31 p.m. EST
I additionally tried using segregated solvers (2 segregated solvers 1 for pressure and velocity and 1 for temperature). I have read this should help solve highly nonlinear problems. But again after about 30 seconds into the transient solutions, the solution fails.

Edited: I added some of the solutions so readers do not have to open and run the model. Solutions are plotted for times 20, 30 and 40 seconds into the simulation.
I additionally tried using segregated solvers (2 segregated solvers 1 for pressure and velocity and 1 for temperature). I have read this should help solve highly nonlinear problems. But again after about 30 seconds into the transient solutions, the solution fails. Edited: I added some of the solutions so readers do not have to open and run the model. Solutions are plotted for times 20, 30 and 40 seconds into the simulation.


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Feb 25, 2013, 3:22 p.m. EST
Can anyone offer any tips or solution. I am still struggling with this problem. Thanks
Can anyone offer any tips or solution. I am still struggling with this problem. Thanks

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Feb 28, 2013, 2:28 a.m. EST
Hi

for me your model is missing a few things:

1) you have some nodes not defined on any BC
2) you have no inlets (not strictly required)
3) outlets as inlets might act in both directions, but you must define the boundary temperature somewhere
5) if you do not force an inlet your outlet might become an inlet but T is not defined there
6) check carefully the equations for "open Boundary" and see if you cannot use that instead of outlet & outflow
7) if you use 6, note that you then do not have any gauge pressure defined, you need to add a fixed point or BC with a p=0 somewhere, note that p changes with elevation by about 12 Pa/m at sea level

--
Good luck
Ivar
Hi for me your model is missing a few things: 1) you have some nodes not defined on any BC 2) you have no inlets (not strictly required) 3) outlets as inlets might act in both directions, but you must define the boundary temperature somewhere 5) if you do not force an inlet your outlet might become an inlet but T is not defined there 6) check carefully the equations for "open Boundary" and see if you cannot use that instead of outlet & outflow 7) if you use 6, note that you then do not have any gauge pressure defined, you need to add a fixed point or BC with a p=0 somewhere, note that p changes with elevation by about 12 Pa/m at sea level -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Feb 28, 2013, 10:53 a.m. EST

I will try your tips and post the results.
Thanks,
Lucas
I will try your tips and post the results. Thanks, Lucas

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 4, 2013, 9:19 a.m. EST
Here are the results as promised in case there are others with this problem.
I tried to refine the mesh and the problem seemed to have trouble converging. Has anyone experienced this behavior before?
Here are the results as promised in case there are others with this problem. I tried to refine the mesh and the problem seemed to have trouble converging. Has anyone experienced this behavior before?


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Apr 22, 2013, 1:49 p.m. EDT
Hi,

I am also facing the same problem in getting the stationary solution converged for my conj.ht. model (attached). I had been advised to study it in two steps:
1- transient (say 0-1 sec) in order to set reasonable initial values for the stationary solution.
2- stationary with dependent variables on the previous time dependent solution.

But it also did not help. I tried also to set different mesh sizes/strategies, again it did not help. I tried to change solver options, it did not help.

@ Ivar, if you looked to my model you will find it includes all basic BCs/suggestions you advised to Lucas. But it is still not converging.

The model is really not that complicated. If we can not just get the steady state solution, It would be nice if we can get some physical explanation why?

Thanks a lot,
Tamer
Hi, I am also facing the same problem in getting the stationary solution converged for my conj.ht. model (attached). I had been advised to study it in two steps: 1- transient (say 0-1 sec) in order to set reasonable initial values for the stationary solution. 2- stationary with dependent variables on the previous time dependent solution. But it also did not help. I tried also to set different mesh sizes/strategies, again it did not help. I tried to change solver options, it did not help. @ Ivar, if you looked to my model you will find it includes all basic BCs/suggestions you advised to Lucas. But it is still not converging. The model is really not that complicated. If we can not just get the steady state solution, It would be nice if we can get some physical explanation why? Thanks a lot, Tamer


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Apr 22, 2013, 2:00 p.m. EDT
I forgot to clarify that, that my model works fine in transient but the problem is in the stationary solution.

Thanks,
tamer
I forgot to clarify that, that my model works fine in transient but the problem is in the stationary solution. Thanks, tamer

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Apr 26, 2013, 1:50 a.m. EDT
Hi

first of all I'll advise you to check if you cannot upload the latest patch to v4.3a, you do not have the latest version I see

then I would propose that you fillet off all your corners with small radii, your sharp edges makes singularities

Further I'm not sure why you solve first for time and then a stationary, as for t=0 you do already solve the equivalent of a stationary case

T follow what is going on, I recommend to turn ON the plot while solving FORA all solver steps, you will then notice that your fluid is rapidly getting turbulent and that your solver has issues to find steady solutions

Finally I would also recommend that you rebuild your model from scratch, you have disabled a few nodfes, bu I notice that I cannot even delete these, so it's as if your internal pointers in the file have become corrupted. This happens sometimes, when you modify heavily a model, adding and removing nodes and solvers over longer periodes


--
Good luck
Ivar
Hi first of all I'll advise you to check if you cannot upload the latest patch to v4.3a, you do not have the latest version I see then I would propose that you fillet off all your corners with small radii, your sharp edges makes singularities Further I'm not sure why you solve first for time and then a stationary, as for t=0 you do already solve the equivalent of a stationary case T follow what is going on, I recommend to turn ON the plot while solving FORA all solver steps, you will then notice that your fluid is rapidly getting turbulent and that your solver has issues to find steady solutions Finally I would also recommend that you rebuild your model from scratch, you have disabled a few nodfes, bu I notice that I cannot even delete these, so it's as if your internal pointers in the file have become corrupted. This happens sometimes, when you modify heavily a model, adding and removing nodes and solvers over longer periodes -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.