Menu

Discussion Forum

New Discussion
Filter

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 The information provided may be out of date.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.

Convergence behaviour in a stationary solver

Posted Mar 14, 2013, 1:06 p.m. EDT Fluid & Heat, Computational Fluid Dynamics (CFD), Geometry, Mesh, Modeling Tools & Definitions, Parameters, Variables, & Functions, Studies & Solvers Version 4.2a, Version 4.3, Version 4.3a 5 Replies

Miguel Teran
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Dear Comsol gurus,

I have been dealing with a CFD/transport of diluted species (TDS) simulation for a while and I was hoping you could bring some light to the subject.

I am simulating a flow in a micro mixer (related to example: www.comsol.com/showroom/gallery/320/) using laminar flow and transport of diluted species. I want to emulate the behaviour of the so called "chaotic mixers" in which a laminar flow with an initial concentration of a specie "c" is "streched and bent" in order to mix "c" in the entire volume. The chaotic mixer is used in small channels (µm and mm) due to the lack of turbulent forces.

I am not having any results so far. I have uploaded the convergence plots for both step 1: laminar flow, step 2: TDS and my model. At the beginning I wanted to define a lateral (zx in this case) boundary as symmetric boundary for then mirror it. The latter was unsuccessful and I defined it as a wall without results. I have used the default iterative solver and direct solver, I have change the mesh from coarser to fine, I also changed the multigrid mesh coarsening factor from 2 to 1.5 and increase the relative tolerance, all of them at the same time or each one separatelly without any success.

My first guess is that the large "turbulence" made by the bending and streching of the flow might increase each time the error in each iteration leading to a never ending iterative process. But the people involved in the original article was successful in their simulation using periodic boundary conditions but I dont know how to use them.

Sorry for the long post but I wanted to give you a better picture of my situation. Your ideas are more than welcome

My best regards

Miguel
5 Replies Last Post Mar 19, 2013, 4:20 a.m. EDT
Miguel Teran
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 18, 2013, 4:21 a.m. EDT
Hello again,

I have tried to meddle with the Stabilization options (Show>Stabilization, and then at both physics at the "consistent stabilization" menu) by removing one or both the marks of "streamline diffusion" and "Crosswind diffusion" options. Still no success, maybe anyone has a great idea that can help me.

Thank you in advanceand my best regards

Miguel
Hello again, I have tried to meddle with the Stabilization options (Show>Stabilization, and then at both physics at the "consistent stabilization" menu) by removing one or both the marks of "streamline diffusion" and "Crosswind diffusion" options. Still no success, maybe anyone has a great idea that can help me. Thank you in advanceand my best regards Miguel
Julien Givernaud
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 18, 2013, 10:35 a.m. EDT
Hi Miguel,

I have some trouble as you to make spf in 3D converged. I try GMRES with multigrid with Vanka, SOR lines preconditionners with no success. I try spf P1+P1 elements type and a direct solver MUMPS it works but i want to how to compute Navier Stokes (spf P2+P1) with multigrid solver.

Good luck
Hi Miguel, I have some trouble as you to make spf in 3D converged. I try GMRES with multigrid with Vanka, SOR lines preconditionners with no success. I try spf P1+P1 elements type and a direct solver MUMPS it works but i want to how to compute Navier Stokes (spf P2+P1) with multigrid solver. Good luck
Miguel Teran
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 19, 2013, 3:29 a.m. EDT
Hello Julien,

thank you for your answer. I did almost the same, I changed the spf to P1+P1 but I left the Iterative solver (GMRES) and it finally converged. My guess is that maybe the turbulence made by the mixer lead to oscillations in the second order elements of velocities. However, I would like to maintain the P2+P1, do you or someone else know if this can be done?

Edit: I also remove the mark in "crosswind diffusion" at the "Consistent Stabilization" tab.

Thanks in advance and my best regards

Miguel
Hello Julien, thank you for your answer. I did almost the same, I changed the spf to P1+P1 but I left the Iterative solver (GMRES) and it finally converged. My guess is that maybe the turbulence made by the mixer lead to oscillations in the second order elements of velocities. However, I would like to maintain the P2+P1, do you or someone else know if this can be done? Edit: I also remove the mark in "crosswind diffusion" at the "Consistent Stabilization" tab. Thanks in advance and my best regards Miguel
Julien Givernaud
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 19, 2013, 4:07 a.m. EDT
Hi Miguel,

Finally i try many possibilties and it works in my case with this setting:

Stationnary solver : fully coupled
iterative FGMRES
Multigrid F-cycle (with other default settings)
Presmoother Vanka: variable mod.u, nb iteration 1, (with other default settings)
Post smoother SOR lines (default settings): nb iterations 3, relax factor 0.4...
Coarsed solver: direct PARDISO

Julien
Hi Miguel, Finally i try many possibilties and it works in my case with this setting: Stationnary solver : fully coupled iterative FGMRES Multigrid F-cycle (with other default settings) Presmoother Vanka: variable mod.u, nb iteration 1, (with other default settings) Post smoother SOR lines (default settings): nb iterations 3, relax factor 0.4... Coarsed solver: direct PARDISO Julien
Miguel Teran
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 19, 2013, 4:20 a.m. EDT
Hi Julien,

I will try this right away! Thank you!

Miguel
Hi Julien, I will try this right away! Thank you! Miguel

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.

Suggested Content