Learning Forum

[Alex Chiella]

Hi all!

I've been running simulations with gas flow inside a short sort of y-pipe using the SST k-omega model.

I noticed that more critical runs (higher/lower velocities and whatnot) tend to stabilize their scaled residuals just above the default threshold, using the 1 default Time Scale Factor.
If I were to lower the Time Scale Factor after calculation, those residuals would plummet (see figure below to notice the dip after about 125 iterations). They would at times instantly meet the default convergence criteria, but tend to rise again if I force resume the computation (via lowering some criterion).

If, on the other hand, I start with a low time scale factor, I get nowhere close the results that I obtain using the larger one, and residuals fluctuate pseudo-periodically around higher values.

My understandings were that:
1) if convergence is achieved with a higher Time Scale Factor, same converged results should be guaranteed for lower Time Scale Factors (with a larger number of iterations). The opposite is obviously not true in general.
2) it makes sense to "channel" the simulation towards convergence using a low Time Scale Factor, and then increasing it for faster convergence once the solution has started to take actual shape.

What is happening to me is that lower Factors are not being solution-conservative in regard to 1), and that what is working for me is the exact opposite of what is recommended in 2).

So my question is: might any of this make sense, or my simulations must be necessarily flawed somewhere?

[Rob]

The time scale is a under relaxation function (in simple terms). So a high value will be less stable but will reach the equilibrium result quickly; a low value will be stable but take many iterations to reach the equilibrium result. A sudden drop in any UR value can cause residuals to drop, but that's because the solution change slows, not because it's physically converged. Hence we typically use monitors, fluxes and review images to judge convergence. 

[Alex Chiella]

Thank you Rob, I'll have to stop fixating on residuals then!

[Rob]

Watch residuals too as they're a monitor, it's just to remind you to check using the other tools.