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.

Instability using PDE module

Hi all,

As beginner with COMSOL I'm having a problem with the stability of my COMSOL results.
In the PDE module I try to solve four coupled differential equations regarding a MHD flow in an annular channel. At one boundary, the velocity profile shows oscillations. Increasing the mash results in stronger oscillations in a smaller region. Does someone have a suggestion how to suppress the oscillations without simply working with a finer mash?

Thanks,

Peter

5 Replies Last Post 11/10/2012, 01:10 GMT-04:00
Posted: 5 years ago 09/10/2012, 15:04 GMT-04:00
Hi

it could be that you do not have enough BC and/or initial conditions and that COMSOL is "oscillating" between two valid solutions.
Normally the solver is made to only work for a fully defined system with ONE solution

--
Good luck
Ivar
Hi it could be that you do not have enough BC and/or initial conditions and that COMSOL is "oscillating" between two valid solutions. Normally the solver is made to only work for a fully defined system with ONE solution -- Good luck Ivar

Posted: 5 years ago 10/10/2012, 03:14 GMT-04:00
I'm shure I have enough boundary conditions regarding my PDE's.

And shouldn't I further than expact the whole solution to be oscillating and not only the region at the boundaries?
I'm shure I have enough boundary conditions regarding my PDE's. And shouldn't I further than expact the whole solution to be oscillating and not only the region at the boundaries?

Posted: 5 years ago 10/10/2012, 07:24 GMT-04:00
Hi

that depends have you tried to change the mesh density, if you oscilalte with the mesh (turn off the smoothing under quality) then there is a fair chance that you have 2 valid solutions and COMSOL cannot decide or alternates, between both.

Or you might change a littleyour initial conditions / BCs and you can then perhaps select one or the other.

ut there might be other issues too ;)

--
Good luck
Ivar
Hi that depends have you tried to change the mesh density, if you oscilalte with the mesh (turn off the smoothing under quality) then there is a fair chance that you have 2 valid solutions and COMSOL cannot decide or alternates, between both. Or you might change a littleyour initial conditions / BCs and you can then perhaps select one or the other. ut there might be other issues too ;) -- Good luck Ivar

Posted: 5 years ago 10/10/2012, 10:48 GMT-04:00
Hi,

I turned off the smoothing, which didn't change the solution at all. I'll simply define a fine mash at the boundaries and hope it will be fine enough to solve the oscillations there.

Thanks anyway
Hi, I turned off the smoothing, which didn't change the solution at all. I'll simply define a fine mash at the boundaries and hope it will be fine enough to solve the oscillations there. Thanks anyway

Posted: 5 years ago 11/10/2012, 01:10 GMT-04:00
Hi

the postprocessing smoothen doesnt change the solution, at best it reveals the "true" calculated points and hiding some artefacts one can obtain in poorly meshed regions (w.r.t. dependent variable and their gradients)
To improve the solution, one must identify the true issue and solve that, but often it comes from meshing issues as this is the "model sampling" used for the calculations. Think of time signal sampling (music) if you want to get the correct tone and pitch, you need to sample up to the Nyquist criteria, and if posible better, with the mesh it's similar, you nee to sample correctly the dependent variables and at to resolve least their first derivative , the gradients

--
Good luck
Ivar
Hi the postprocessing smoothen doesnt change the solution, at best it reveals the "true" calculated points and hiding some artefacts one can obtain in poorly meshed regions (w.r.t. dependent variable and their gradients) To improve the solution, one must identify the true issue and solve that, but often it comes from meshing issues as this is the "model sampling" used for the calculations. Think of time signal sampling (music) if you want to get the correct tone and pitch, you need to sample up to the Nyquist criteria, and if posible better, with the mesh it's similar, you nee to sample correctly the dependent variables and at to resolve least their first derivative , the gradients -- 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.