Iterative solver settings problem

Hi all!
I use Feflow 5.41.
In steady flow/steady transport simulation the run simulator shows this warning message “[i]BiCGSATBP matrix solution in 500 steps (> 500 maximum iterations) with errors: res_n=2.848647e+10, res_ini=1.327481e+10[/i]”

I applied the smooth mesh method in “Transform mesh” menu and refined the mesh too.

Thanks for your help !

The residual looks quite large, so that I would think that the reason for non-convergence is somewhere else. Have you re-checked all model parameters?

Thank you for the reply, Peter.
I re-checked all model parameters. I've a different model with the same problems. As heat BC I used the first condition (mean temperature) at top of model; negative heat flux at the bottom of the model (general geothermal gradient, -5616 (J/m2d).
I run the model to obtain the intial conditions.

Hi Peter and Matteo,

I have about the same (2D vertical, trans flow/trans transport) problem:
It's the [b]Horton-Rogers-Lapwood [/b]problem in subcritical rayleigh domain (no convection), but with [u]h=0 upper boundary and q=0.6885 W/m2 at lower bound[/u] (corresponds to 0.03K/m geothermal gradient with Feflow default material parameters).

For [b]initial conditions[/b] I first run the problem with a lower bound constant temperature (20.5 deg C) at steady state. This gives a nice linear geothermal gradient. I then switch to transient/transient with constant q.
I use 'streamline upwinding' and a constant(T), thermal expansion coefficient = 1,5 e-4 /K

[b]Meshing[/b]
Domain: W=2500 m, H=350 m (1 polygon)
Mesh: Triangle, refinement around all polygon borders, gradation 6, target 1m (rest default triangle settings)
The maximum element size in the middle of the model is no more than ~4 m and the model consists of 280k elements.

Even when increasing the BICGSTABP iterations to 2500 I still get residuals of 10e+8, which seems similar to Matteo findings.

The expectation is that I will get the same temperature distribution as the initial conditions and I do, but instability might very possibly lead to artefacts upon complexifying the model.This seems like a pretty straight forward problem, so [b]what do you reckon could cause the instability? [/b]

Thank you very much, keep up the good work!
[u][/u]
Kind regards,
Roeland

Ideas: run a test with finer mesh, use square elements, use uniform mesh, try SAMG with user-specified timestepping, use more string error norm, etc

Certain types of problems seem to be unstable in cross-section mode (boundary condition along the bottom of a section, so you might also try a "3D" model that is set up to simulate a unit thickness cross-section.

Pete