Convergence problem

matrix · Posts: 13

Hi,
I tried to do a CFD simulation, in which there are some supply diffusers on the ceiling and on internal partitions and some extract grilles on the internal partitions that delimit the room.
After checking the in/out fluxes balance, I started the CFD simulation and after thousands iterations every Residual quantity (x/y/z velocity, turbolence KE, turbolence Dissipation...) converges, except Mass.
I've just tried changing time step, relaxation factors... and also using an higher number of cells, but the Mass doesn't converge.
I don't care about thermal profile, but only the velocity vectors in order to try to understand the air flow route.
Attached you can find the .dbs file.
Thank you for helping me.

Steve Potter · Posts: 616 Location: newcastle

Hi,

The problem with your model is that many of the supply and extract edges coincide with the parent surface edges which can cause the air flow to become 'stalled' or 'locked' at the corners of the domain resulting in the mass source residual failing to fall below a minimum level. If you redraw the supply and extract boundaries ensuring a distance of at least 0.1m between the boundary and parent surface edges (as illustrated), you should find that the mass source will continue to fall.

We will add a comment to the on-line help so that others will hopefully not run into the same problem.

Regards,

Steve

matrix · Posts: 13

Hi,
I tried to do simulations after changing extract grilles and supply boundaries as you said, but the problem concerning Mass convergence is still present.
Attached the relevant file .dbs.
Thanks.
Regards.

Steve Potter · Posts: 616 Location: newcastle

Hi,

There is still one small supply boundary on the ceiling which is causing a problem because it is too close to the parent surface edge. If you delete this boundary and incorporate the flow rate within one of the adjacent supply boundaries, you should find that the mass source then continues to reduce.

I have attached a screenshot that highlights the location of the problem supply boundary.

Regards,

Steve

matrix · Posts: 13

Hi,
I tried again following your suggestion but the mass still doesn't converge. Furthermore

it seems that also the velocity components fluctuate.
I saw that even if after 10'000 iterations the quantities continued to fluctuate.

Regards.

Steve Potter · Posts: 616 Location: newcastle

Yes, there is still a mass convergence problem with your model which appears to be caused by the adjacent extract boundaries on the surface indicated on the attached screenshot. While we investigate the underlying reason for this, a workaround is to combine these extracts into a single extract. Generally speaking, it is better to avoid placing ventilation boundaries hard up against each other.

After making this modification, you should find that the solution converges in a relatively stable fashion for coarser grids (> 0.3m grid spacing) but for finer grids (< 0.2m grid spacing), the solution becomes somewhat unstable with high oscillations in the residuals. However, if you use a small grid spacing, you can improve the stability by reducing the velocity false time steps to a value of around 0.005 and you should find that the solution then stabilises after around 200-300 iterations.

Hope this is of help.

Steve

Shortcuts

Login