Solving Coupled Flow and Diffusion in COMSOL: Error Analysis

[tommyers]

Hi,

I am not sure if this is the right place to post a COMSOL question. If not I apologise, could you point me to the right place.

I have a 3D geometry in COMSOL shape like a "T" where I want to model flow using navier stokes and a chemical diffusion using convection and diffusion. Both work independantly. However, when I try to solve them both together, because they are coupled. I get this following message:

Error:
Time 0:
Matrix is singular.
Out_of_memory_LU_factorization
Last time step is not converged.

Now I have read what might cause this, suggestions include my mesh type, I have varied this but without success. The other was sharp corners, my geometry has sharp corners, and suggestions have been to fillet the corners, however, I can't figure out how to fillet 3D corners.

Moreover, what could the problem be?

Regards,
Tom

 

[PerennialII]

These can be tough problems to locate and you've probably tried the obvious solutions ... the coupling, and at the very beginning of the analysis at t = 0 ... how have you executed this, is there "something" which can hit infinity at the beginning of the analysis, and should be smoothed out one way or another? And is there something nonlinear (that'd open up some options)?

Often analyzes need to be started in a tad "smoothed" manner to get them running proper, since the initial state isn't numerically really "viable". It may be the corners, but smoothing them out shouldn't always be a necessity (and it can require some effort, and may not really be the underlying source of the problem). Often activating parts of the physics at t>0 is a simple solution and/or adding functions which are "smooth, continuous and stable" + dissipate away as t goes further to define initial and boundary conditions or otherwise unstable parts of the analysis help (of well, something which use a lot when don't want to modify the underlying geometry when working with new implementations).

 

[sameer2blore]

3D fem help in matlab

Hi, i have done 3d fem for solid cylinder. the stiffness matrix is not coming as it should be. there should be 5 lines in sparse matrix. but it is coming 2 more lines in outer side. and the sparse matrix is near singular or showing badly scaled . anybody can help. anybody can help for assembling the local matrices to global.