Converged or not? Indoor velocity points oscillating.
 

I ran indoor RANS using Coupled scheme and RNG k- closure. The mean y* <5 so I used LRNM. I monitored several x-velocity points in the domain. After n iterations, they started looking like so:

https://i.imgur.com/4RicGtw.png

I've applied data sampling to the monitor points but they are still oscillating. However, the amplitude of the oscillations is of the order 10E-5 [m/s]. From a practical standpoint, such variability is negligible. This brings me to my question:

Can we consider the solution converged or does the presence of oscillations, no matter how small, suggest that additional measures should be taken?

What you see has no sense in terms of convergence! You need to check the residuals of the equations you are solving.

Indeed, I have looked at the residuals for the continuity and momentum equations. They have flattened out but are oscillating. This is typical for indoor flows. I am also monitoring velocity at specific points in the domain. As you mention, this is not a direct indication of convergence but I find it noteworthy that all of them are oscillating, instead of gradually increasing or decreasing. I struggle to grasp the implications of this behaviour.

If you have a small range of values in the monitor, the oscillations appears greater than they actaully are.

Show the residuals (normalized) since their initial values and check how many order of magnitude they decrease. Then evaluate the magnitude of the oscillations around the constant residuals.

The oscillations are very small and if I understand you correctly, you are suggesting that they may be small enough to ignore.

https://i.imgur.com/cdCnRjx.png

The normalized residuals will not decrease further and the oscillations are significant. I could play around with the URFs but I suspect that oscillations may persist because I am running RANS for a flow that is inherently transient. I guess the question remains whether the amplitude of the oscillations is acceptable.

My impression is that your solution is very poorly converging ... that could depend on the grid resolution, on the numerical scheme but often is the formulation of the turbulence model to be the problem.

However, I would not accept such solution.

Hmm, I will look into it further and see what I can find. Thank you for your help!