SimScale CAE Forum

Simple Valve simulation

Hello, I am preparing a simulation of an irregular water channel. The geometry is quite simple, with a k-epsilon turbulence model, however even by changing the parameters/mesh size I cannot reach the convergence, by checking the different plots.

Here below is a draft solution (not converged) of the velocity field

and here the convergence graphs:




Any advice on how to improve the results ( sim parameters, mesh refining etc) it would be hugely appreciated. thanks!

1 Like

Hi there, thank you for using the forum!

Checking the Residuals tab should be enough. Try increasing the Level of Fineness to 8 and rerun for 1000 loops( 30000 is actually unnecessary in this case). The convergence plot should look good then

Best regards,

I’ll try, thanks fillia!

1 Like

Hi Fillia, Raising the number of elements improved the situation for the residuals but the outlet and domain plots are still performing poorly I believe:

I am not sure on what I need to change, any suggestion for this would be great

No need to check these tabs, you can focus on the Residuals tab only :slight_smile:

Best regards,

@tsite I am basing my analysis from exactly this page:

Outlets convergence plot shows some serious issues. High fluctuations are observed in pressure and velocity values. This is a clear indicator that boundary conditions or the mesh quality might be responsible for numerical issues at the outlets.

Which run are you referring to? :slight_smile:

Hi @tsite , this is the simulaion:


Hi there I hope you had a nice weekend!
I will assume you refer to the latest run then (baseline_8_long). As I mentioned earlier, 1000 loops should be enough for this type of analysis (10000 is an overkill and unnecessary). If you have a steady-state analysis keep running for a longer time than needed, this does not necessarily mean the results will be “better”. Check the outlet’s results after 1000 loops for example:

I am tagging my colleague @RicardoParis in case he has some insights regarding really long steady state analysis runs.


Thanks for your advise on this topic. Yeah 1e04 iterations are not necessary for the residuals but, apparently not enough for the convergence of the different variables. My understanding is that, to be confident in the solution, at least in the domain tab, the variables should have a stable trend, which is not what we see in the screenshot I posted.
Sorry for insisting on this topic, but I am not sure if I am getting this right or not